JP3626974B2 - Image processing device for medical diagnosis - Google Patents

Description

によるフイルタリング処理が行われる。ここでＸ（ｉ，ｊ） は入力画像Ｘの位置（ｉ，ｊ） の座標値を表し、Ｙ（ｉ，ｊ） は空間フイルタ８より出力される画像の位置（ｉ，ｊ） を表す。またａ_ｄｉｄｊは２次元のフイルタ係数を表す。フイルタ係数としては、例えば図３や図４に示すようなフイルタ係数を用いる。図３に示すようなフイルタ係数を有するフイルタは低域通過型フイルタであり、この低域通過型フイルタを用いた場合、後述の境界線抽出で得られる境界線は低域成分の等高線になる。また図４に示すようなフイルタ係数を有するフイルタはラプラシアンＧと呼ばれる平滑化微分フイルタであり、ガウス関数の２階微分をデイジタル化したものである。この平滑化微分フイルタによつて得られる境界線は、輝度変化の大きなエツジに相当する。
【００２３】
空間フイルタ８の出力は２値化器９に入力され、２値化器９では、空間フイルタ８の出力を、予め設定されているしきい値ＴＨ１によつて２値化する。すなわち入力値がしきい値ＴＨ１よりも大きな値の場合には「１」、小さな値の場合には「０」に変換される。この出力値はフレームメモリ３の所定の位置に上書きされる。
【００２４】
このようにして入力画像全体に対する２値化処理が終了すると、図５に示すように、マイクロコンピユータ４は境界線抽出処理ＳＰ１、輪郭線抽出処理ＳＰ２及び輪郭線形状修正処理ＳＰ３を実行する。以下にマイクロコンピユータ４によるこれらの処理について説明する。ここで後述するように、メモリ１０及び１１には、マイクロコンピユータ４による計算結果が格納される。すなわちメモリ１０には選択された輪郭線を構成する点の座標系列が格納され、メモリ１１には輪郭線の形状を修正するための変曲点の情報が格納される。
【００２５】
まずマイクロコンピユータ４による境界線抽出処理ＳＰ１の処理手順について説明する。図６に示すように、境界線抽出処理ＳＰ１の第１の処理としての境界線開始点の検出処理ＳＰ１０において、マイクロコンピユータ４は図２に示す順序でフレームメモリ３を走査し、各位置で次式
【数２】

の関係が満たされているか否かを調べる。（２）式において、Ｘ（ｉ，ｊ） はフレームメモリ３上の位置（ｉ，ｊ） における座標値を表す。フレームメモリ３を走査しているとき、最初に（２）式を満足する位置（ｉ，ｊ） を境界線の開始位置とする。なお以下において、フレームメモリ３上の位置を表す場合にｐを用い、例えばその点の座標が（ｉ，ｊ） である場合には、ｐ＝（ｉ，ｊ） と表す。また検出された境界開始点をｐ_０ とする。
【００２６】
境界開始点ｐ_０ が検出されると、この位置における２値画像の値（「０」又は「１」）がレジスタｒ_０ に保存され、新たなラベルＬが生成されてフレームメモリ３上の境界開始点ｐ_０ の位置に値「Ｌ」が保存される。ここでラベル番号Ｌは、予め格納されている２値画像と混同しないように「２」以上の整数を用いる。境界線抽出処理ＳＰ１は、境界開始点を検出するための走査が画像の右下に到達した時点で終了する。
【００２７】
次にマイクロコンピユータ４は、境界線抽出処理ＳＰ１の第２の処理としての境界線追跡処理ＳＰ１１において、検出された境界開始点ｐ_０ を含む境界線を構成する点列（境界点）ｐ_ｋ （ｋ＝０、１…）を近傍探索によつて順次追跡する。この近傍探索処理を図７を用いて説明する。フレームメモリ３上の２値画像を表す図７において、いま境界点ｐ_ｋ までが確定しているとする。このとき、境界点ｐ_ｋ の近傍８画素の値を時計回りに順に調べ、レジスタｒ_０ と同じ値（すなわち境界点ｐ_ｋ と同じ値）をもつ最初の位置を次の境界点ｐ_ｋ＋１ とする。境界点ｐ_ｋ の近傍８画素のいずれから探索を開始するかは、境界点ｐ_ｋ の１つ前に確定した境界点ｐ_ｋ−１ と境界点ｐ_ｋ との位置関係によつて決まる。
【００２８】
すなわち境界点ｐ_ｋ−１ を中心とした近傍８画素の中で、時計回りに見た場合に境界点ｐ_ｋ よりも順序が１つ前に当たる位置を、境界点ｐ_ｋ＋１ を求める際の探索開始点とする。図７においては、境界点ｐ_ｋ 上に位置する画素が近傍探索の開始点となり、右隣の画素が境界点ｐ_ｋ＋１ として検出される。フレームメモリ３上のｐ_ｋ＋１ の位置にはラベル番号Ｌを保存する。以上の操作を、（Ａ）次の境界点が境界開始点ｐ_０ と一致するか、又は（Ｂ）次の境界点が１つ前の境界点ｐ_ｋ−１ と一致するかの２つの条件のいずれかが満たされるまで繰り返す。
【００２９】
条件（Ａ）は境界線が閉じたことを意味しており、１つの境界線の追跡が完了したことになる。この場合、次の境界線の開始点を検出するため、現在の境界開始点ｐ_０ の次の走査位置から再び境界開始点検出処理ＳＰ１０を行う。条件（Ｂ）は、追跡中の境界線が閉じていないことを意味する。この場合図８に示すように、境界開始点ｐ_０ に戻り、近傍探索を反時計回りに行なうことで、反対方向に境界線を追跡する。反対方向への追跡が終了した時点で、境界開始点ｐ_０ の次の走査位置から境界開始点Ｐ_０ の検出を再開する。
【００３０】
このようにして境界線を抽出すると、マイクロコンピユータ４は輪郭線抽出処理ＳＰ２を開始する。図９に示すように、まずマイクロコンピユータ４は輪郭線選択処理ＳＰ２０において、上述の方法によつて抽出した各境界線毎にフレームメモリ３を走査し、ラベルＬを値としてもつ位置（ｉ，ｊ） とメモリ７に保存されている基準位置（ｉ_ｃ， ｊ_ｃ ） との距離ｄを次式
【数３】

によつて計算する。ラベルＬをもつすべての位置に対して距離ｄを計算し、図１０に示すように、その最小値をラベルＬの境界線と基準位置との距離ｄ_ｌ とする。全ての境界線の中で、基準位置との距離ｄ_ｌ が最も小さい境界線を輪郭線として選択する。なお各境界線の基準位置との距離ｄ_ｌ は、上述の境界線の追跡において計算しておくこともできる。
【００３１】
次にマイクロコンピユータ４は、輪郭線座標系列生成処理ＳＰ２１において、フレームメモリ３で選択された輪郭線を構成する点の座標を順次メモリ１０に保存して座標系列を生成する。この処理手順は、図６に示す境界線抽出処理と同じであるが、境界線開始点の検出処理では、各位置において（２）式を評価する代わりに、その位置のラベルが、選択された輪郭線のものであるか否かを調べる。すなわち走査中、最初に輪郭線のラベルに遭遇した位置を境界線開始点ｐ_０ とする。また追跡すべき対象は１つしかないので境界線開始点の検出は１度でけである。
【００３２】
この座標系列の生成において、上述の追跡停止条件（Ｂ）すなわち次の境界点が１つ前の境界点ｐ_ｋ−１ と一致した場合には、この時点までに生成した座標系列の順序を逆転させた後、逆方向の追跡を開始する。逆方向の追跡が完了した段階で、再び全座標系列の順序を逆転させる。これにより、生成された座標系列が輪郭線を時計回りになぞる順序になることが保証される。
【００３３】
このようにして輪郭線抽出処理ＳＰ２が終了すると、マイクロコンピユータ４は図１１に示す処理手順に従つて輪郭線形状修正処理ＳＰ３を開始する。この場合、心臓のようなほぼ楕円形の器官の輪郭線を抽出する場合について説明し、図１２及び図１３に示すような輪郭形状を修正する場合について説明する。図１２に示すような輪郭線が生成される原因としては、空間平滑化のための隣接領域の融合が考えられる。また図１３に示すようなくぼみは、所望器官に他の器官が重なつている場合に発生すると考えられ、心臓の超音波画像における乳頭筋がその１例である。
【００３４】
まずマイクロコンピユータ４は輪郭線形状修正処理ＳＰ３として、変曲点検出処理ＳＰ３０において、図１４に示すように、メモリ１０に格納されている座標系列の各点に対して曲率ρを次式
【数４】

によつて計算する。ここで座標（ｉ_ｎ， ｊ_ｎ ） はメモリ１０においてｎ番目に格納されている座標であり、ｋは注目点前後の輪郭線の傾きを計算するための定数である。曲率ρの絶対値が予め設定されたしきい値ＴＨ２よりも大きい場合にはこの点を変曲点とみなし、メモリ１１のｎ番目に曲率ρの値を保存する。曲率ρの絶対値がしきい値ＴＨ２以下の場合には、メモリ１１のｎ番目には「０」を保存する。ここで図１４に示すように、（４）式の符号は変曲点の種類を表す。すなわち（４）式によつて得られる値が正の値となる場合には、この変曲点は輪郭線の中心部に対して凹と考えられ、逆に負である場合には凸であると考えられる。
【００３５】
ここで融合領域の除去処理ＳＰ３１は、図１５に示すように、曲率ρの特に大きな凹の変曲点で挟まれた部分を除去することによつて実現する。すなわちメモリ１１の中で正の値を有し、その絶対値が予め設定されたしきい値ＴＨ３よりも大きい点を融合接点とする。しきい値ＴＨ３は、変曲点検出処理ＳＰ３０で用いたしきい値ＴＨ２よりも大きな値に設定される。融合接点が２つ以上検出された場合には、各融合接点に対して、隣の融合接点までの画像上での直線距離を計算する。隣接する融合接点間の距離が予め設定されたしきい値ＴＨ４よりも小さい場合には、この２つの融合接点に挟まれたメモリ１１上の各位置に特別な値ρ∞を代入し、輪郭線より除去する。図１６に示すように、輪郭線が閉じている場合には、２つの融合接点によつて分けられる輪郭線の２つの部分についてそれぞれ重心を計算し、基準位置から遠い方を除去する。
【００３６】
（４）式において、定数ｋやしきい値ＴＨ４の選び方によつては、実質的に１つの融合接点であるにもかかわらず、連結した複数の点が融合接点として検出される場合がある。この場合、連結した融合接点の中から１つを代表点として選び、代表点に挟まれた部分を除去する。例えば代表点として、連結する融合接点の中で曲率ρの絶対値が最大のものを選ぶことができる。
【００３７】
次にマイクロコンピユータ４は、くぼみ部分の除去処理ＳＰ３２において、図１７に示すように、しきい値ＴＨ２で検出された２つの凸変曲点の間に、同じしきい値ＴＨ２で検出された凹変曲点が存在する場合に、２つの凸変曲点で挟まれた部分を除去することによつて実現する。すなわち、まず融合領域の除去処理が施されたメモリ１１の値を順に調べ、正の値を有する凹変曲点を検出する。次にこの凹変曲点から両方向に、負の値を有する凸変曲点を探索し、各方向において最初に遭遇した凸変曲点を検出する。これらの２つの凸変曲点に挟まれたメモリ１１上の各位置に特別な値ρ∞を代入して輪郭線から除去する。
【００３８】
この場合上述の融合接点と同様に、凹変曲点や凸変曲点も連結して検出される場合があるが、この場合にも代表点を用いればよい。また心臓の超音波画像における乳頭筋のように、これによつて生ずるくぼみの輪郭線全体に対する相対的な位置がおおよそ予測できる場合には、凹変曲点、凸変曲点の出現位置を予め制限することができる。すなわち、ある制限された範囲に出現した変曲点に対してのみ、くぼみ部分を除去することができる。
【００３９】
補間処理は、上述の２つの除去処理によつて輪郭線に発生した非連結部分を補間するための処理であり、連結した新しい輪郭線の座標系列を生成する。図１８に示すように、ある部分を除去した結果発生した輪郭線の端点座標をｐ_ｓ ＝（ｉ_ｓ， ｊ_ｓ ） 、ｐ_ｅ ＝（ｉ_ｅ ， ｊ_ｅ ） とすると、例えばこの端点ｐ_ｓ 及びｐ_ｅ 間を次式
【数５】

ような２次関数で補間する。（５）式において、ｔは媒介変数であり、ｐ_ｍ ＝（ｉ_ｍ ， ｊ_ｍ ） は、上述の２つの除去処理の後に残つている輪郭線構成点の重心ｐ_ｇ から図１８のように定まる点である。すなわち点ｐ_ｍ は、端点ｐ_ｓ を通り重心ｐ_ｇ と端点ｐ_ｓ を結ぶ線分に直交する直線と、端点ｐ_ｅ を通り重心ｐ_ｇ と端点ｐ_ｅ を結ぶ線分に直交する直線との交点である。（５）式によつて、端点ｐ_ｓ と端点ｐ_ｅ を結ぶ補間座標系列を発生させるために、ｐ_０ ＝ｐ_ｓ として次の処理を繰り返す。
【００４０】
いま点ｐ_ｎ までの補間座標系列が確定しているとすると、補間点ｐ_ｎ に隣接するいくつかの近傍点ｐ＝（ｉ，ｊ） において次式
【数６】

を計算し、ｓの値が最小となる点を次の補間点ｐ_ｎ＋１ とする。ここで（６）式を計算する近傍は、図１９に示すように補間点ｐ_ｎ と１つ前に確定した点ｐ_ｎ−１ との位置関係によつて決まる。すなわち補間点ｐ_ｎ をはさんで点ｐ_ｎ−１ の反対側５画素（図１９の斜線で示す部分）で（６）式を計算する。また（６）式のｔ_１ 及びｔ_２ はそれぞれ実数であり、かつ０．０ 以上１．０ 以下の値となる必要があり、図１９に示した近傍点のなかで、さらにこの条件を満たすものの中から（６）式を最小にするものを選ぶ。補間点ｐ_ｎ＋１ が端点ｐ_ｅ と一致した時点で、補間座標系列が完成する。メモリ１１上の端点ｐ_ｓ の次から、端点ｐ_ｅ の１つ前までの座標を消去し、消去した座標の代わりに、生成した補間点の座標系列を挿入する。このような補間処理を、除去処理によつて生じた全ての端点間に適用して新たな輪郭線の座標系列を生成する。この場合、必要であれば、この座標系列中の各点をフレームメモリ３に書き込み、画像表示装置５に表示することもできる。
【００４１】
以上の構成によれば、入力された超音波画像の医療診断用静止画像を空間的に平滑化した後複数の境界線を抽出して当該境界線のうち基準位置との距離ｄが最も小さい境界線を輪郭線として選択し、当該輪郭線の形状を修正して所望器官の輪郭線として抽出したことにより、所望器官の輪郭線を自動的に抽出することができるので医療診断を迅速かつ正確に行うことができる。
【００４２】
（２）第２実施例
図１との対応部分に同一符号を付して示す図２０において、２０は全体として本発明の第２実施例による医療診断用画像処理装置を示している。医療診断用画像処理装置２０では、入力された超音波画像の医療診断用動画像を所望器官の動きの繰り返し周期毎にサンプリングして同期加算して時間方向のノイズを低減した後、この時間平滑化された画像を空間的に平滑化して複数の境界線を抽出する。当該境界線より所望器官の輪郭に対応する境界線を輪郭線として抽出し、当該輪郭線の形状を修正して所望器官の輪郭線として抽出するようになされている。
【００４３】
ここでこの実施例では、医療診断用動画像として周期的な動きのある心臓の動画像を用い、また後述する第３実施例及び第４実施例においても医療診断用動画像として心臓の動画像を用いる。
【００４４】
遅延素子２１は画像サンプリング器２２において心電図波形の解析に必要な時間だけ入力データを遅延させる。フレーム同期信号生成器２３は遅延素子２１で遅延される前のフレーム開始のタイミングを示す信号を画像サンプリング器２２に出力する。画像サンプリング器２２は、画像と同期して心電図モニタ装置２４より送出される心電図波形に基づいて、各フレームの入力開始時にそのフレームをサンプリングするか否かを判断し、サンプリングされた画像の時間平均値画像をフレームメモリ３上に生成する。予め設定された周期分の時間平均が行なわれた後、第１実施例と同様の処理により輪郭線が抽出される。フレームメモリ３の値は測定開始時にリセツトされる。また必要であれば、抽出された輪郭線をフレームメモリ３に書き込み、画像表示装置５に表示することもできる。
【００４５】
画像サンプリング器２２の構成を図２１に示す。図２１に示すように、画像サンプリング器２２は、基準点検出器２２Ａ、タイマ２２Ｂ、時相検出器２２Ｃ及び同期加算器２２Ｄによつて構成されている。基準点検出器２２Ａは心電図モニタ装置２４より時々刻々入力される心電図波形の基準点を検出する。基準点は、心電図波形にその繰り返し周期毎に現われる点であり、例えば図２２に示すように、Ｒ波のピーク（各周期において最大値となる位置）を基準点として用いている。基準点が検出されると、タイマ２２Ｂがリセツトされ、リセツト後タイマ２２Ｂは次にリセツトされるまで一定の時間間隔で値をインクリメントする。
【００４６】
時相検出器２２Ｃはフレーム同期信号生成器２３よりフレームの開始を示す信号を受け取ると、タイマ２２Ｂの値を読み取り、予め設定されている値ｔ（以下これを時相と呼ぶ）との差を計算する。この値ｔがしきい値ＴＨ５以下である場合だけ、次に入力されるフレームデータが同期加算器２２Ｄに送出され、これ以外の場合には入力データは破棄される。同期加算器２２Ｄでは、入力データに対応するフレームメモリ３上のデータが読み出され、入力データとの平均値を次式
【数７】

によつて計算する。
【００４７】
ここでＸ_ｎ （ｉ， ｊ）は画像をサンプリングし始めた時点からｎ番目のサンプリング画像における位置（ｉ， ｊ）のデータ値を表し、Ｘ′_ｎ−１ （ｉ， ｊ）はｎ−１番目までの平均画像における位置（ｉ， ｊ）のデータ値、すなわちｎ番目の画像がサンプリングされた時点でフレームメモリ３に保存されているデータの値を表している。（７）式によつて得られる計算結果はフレームメモリ３の同じ位置に保存される。ただし測定開始直後は入力データとフレームメモリ３上の値との平均値計算は行われず、入力データがそのままフレームメモリ３に書き込まれる。
【００４８】
ここで画像をサンプリングする時相を、基準点からの時間ではなく、繰り返し周期における位相として指定することもできる。すなわち基準点検出器２２Ａでは、ある基準点から次の基準点までの時間Ｔを検出し、はじめの基準点からＴ／ａの時刻に入力されたフレームをサンプリングする。この方法によれば、各繰り返しにおける周期のゆらぎを吸収することができるので、一段と正確に同期加算することができる。
【００４９】
以上の構成によれば、入力された心臓の動画像を、心臓の動きの繰り返し周期毎にサンプリングして同期加算して時間方向のノイズを低減した後、この時間平滑化された画像をさらに空間的に平滑化して複数の境界線を抽出し、当該境界線のうち基準位置との距離ｄが最も小さい境界線を輪郭線として選択し、この輪郭線の形状を修正して心臓の輪郭線として抽出するようにしたことにより、心臓の輪郭線を自動的に抽出することができるので医療診断を迅速かつ正確に行うことができる。
【００５０】
（３）第３実施例
図２０との対応部分に同一符号を付して示す図２３において、３０は全体として本発明の第３実施例による画像処理装置を示している。この画像処理装置３０では、入力された超音波画像の医療診断用動画像を、所望器官の繰り返し周期よりも短い間隔でサンプリングし、当該器官の同じ状態にある画像を同期加算して時間方向のノイズを低減した後、所望器官の状態毎に時間平滑化された各画像をそれぞれ空間的に平滑化して時相毎に複数の境界線を抽出する。複数の境界線より所望器官の輪郭に対応する境界線を輪郭線として時相毎に選択し、当該各輪郭線の形状を修正して所望器官の輪郭線を時相毎に抽出するようになされている。
【００５１】
画像サンプリング器３１は、画像サンプリング器２２とほぼ同じであるが、画像をサンプリングする時相として複数の値ｔ０、ｔ１、……が予め設定されている。画像サンプリング器３１において各時相でサンプリングされた画像は時相毎に平均画像が生成され、各時相に対応するフレームメモリ３２に出力される。ここでフレームメモリ３２はサンプリング時相の数と同数用意されており、各フレームメモリ３２には同時相の平均画像が格納される。このように、予め設定された周期分の時間平均が行なわれた後、各フレームメモリ３２上の平均画像に対して第１実施例と同様の処理によつて輪郭線を抽出する。
【００５２】
なお必要であれば、時相の異なる複数の輪郭線を、フレームメモリ３２のいづれか１つのフレームメモリ３３に重ね書きし、画像表示装置５によつて静止画像として表示させてもよい。また各時相の輪郭線を異なるフレームメモリ３２に書き込み、これらを時間的に切り替えて画像表示装置５に表示させるようにしてもよい。
【００５３】
以上の構成によれば、入力された心臓の動画像を、心臓の動きの繰り返し周期において、予め設定した複数の時相でサンプリングし、同時相毎に同期加算して時間方向のノイズを低減した後、異なる時相毎に時間平滑化された画像をそれぞれ空間的に平滑化して各画像より複数の境界線を抽出し、当該各複数の境界線のうち基準位置との距離ｄが最も小さい境界線を輪郭線として抽出し、当該各輪郭線の形状を修正して心臓の輪郭線として抽出するようにしたことにより、心臓の輪郭線を時相毎に自動的に抽出することができるので、心臓の輪郭線の時間的変化を観察することができ、これにより医療診断を迅速かつ正確に行うことができる。また抽出された輪郭線より、輪郭線の速度等診断に必要なパラメータを計算することができる。
【００５４】
（４）第４実施例
図２３との対応部分に同一符号を付して示す図２４において、４０は全体として本発明の第４実施例による画像処理装置を示している。この画像処理装置４０では、１枚の２値画像を保存するための２値画像用フレームメモリ４１が設けられており、空間フイルタとして図３に示す低域通過型フイルタ４２を用いている。
【００５５】
すなわち画像処理装置４０では、画像処理装置３０と同様に時間平均の施された画像に対して、空間フイルタ４２による空間的平滑化処理が行われ、空間フイルタ４２の出力は直接フレームメモリ３２に書き込まれる。マイクロコンピユータ４では、空間的平滑化処理が完了した各画像に対して図２５に示すように閉曲線の生成、輪郭線の抽出及び輪郭線形状の修正を行つて輪郭線を抽出する。すなわちマイクロコンピユータ４は、閉曲線の生成処理ＳＰ１において所望器官の輪郭線の内側に存在する閉曲線を生成し、輪郭線の抽出処理ＳＰ２において、得られた閉曲線を外側に向けて拡大することによつておおよその輪郭線形状を抽出し、輪郭線形状の修正処理ＳＰ３において、抽出された輪郭線の形状を第１実施例で述べた輪郭線形状修正処理（図１１）と同じ方法で修正する。
【００５６】
閉曲線の生成処理（ＳＰ１）の処理手順を図２６に示す。まずマイクロコンピユータ４は、しきい値の初期化処理ＳＰ１０において、座標入力装置６によつて指定された基準位置における画素値をフレームメモリ３２より読み出し、これをしきい値として設定する。次に２値化処理ＳＰ１１において、フレームメモリ３２上の画像データを読み出し、設定されているしきい値によつて２値化を行い、その結果を２値画像用フレームメモリ４１に保存する。次に境界線の抽出処理ＳＰ１２において、この２値画像の境界線を第１実施例で述べた境界線抽出処理（図６）と同じ方法で抽出する。
【００５７】
次にマイクロコンピユータ４は、境界線の選択処理ＳＰ１３において、境界線の抽出処理ＳＰ１２で得られた境界線のうち、座標入力装置６によつて指定された基準位置との距離ｄが最も小さい境界線を選択する。この処理は第１実施例で述べた輪郭線抽出処理（図９）と同じである。次に閉曲線の判定処理ＳＰ１４において、境界線の選択処理ＳＰ１３で選択された境界線に対して次の２つの条件、すなわち（Ａ）開始点と終了点が隣接しているか、（Ｂ）境界線の長さが予め設定されたしきい値ＴＨ６より小さいかについて調べる。
【００５８】
条件（Ａ）又は（Ｂ）のいずれかの条件を満たされない場合には、マイクロコンピユータ４はしきい値の修正処理ＳＰ１５に進み、現在のしきい値を一段と小さい値に設定し直して、２値化処理ＳＰ１１からしきい値の修正処理ＳＰ１５までの処理ループを実行する。この２つの条件がともに満足されている場合には、マイクロコンピユータ４は抽出された境界線を閉曲線と判定し、その座標系列を閉曲線の修正処理ＳＰ１６に送る。
【００５９】
閉曲線の修正処理ＳＰ１６において、マイクロコンピユータ４は、抽出された閉曲線の形状を変曲点に応じて修正する。この処理は第１実施例で述べた輪郭線形状修正処理（図１１）と同じである。ただしここでは、不用なふくらみが生ずることを回避するためにくぼみ部分の補間を（５）式の２次曲線ではなく、次式
【数８】

に示す直線で行うことも可能である。直線補間による補間座標系列を発生させるためには、（６）式の代わりに次式
【数９】

を満たす近傍点ｐ＝（ｉ，ｊ） を追跡すればよい。
【００６０】
次に輪郭線の抽出処理（図２５のＳＰ２）の処理手順を図２７に示す。まずマイクロコンピユータ４は、閉曲線の重心位置の算出処理ＳＰ２０において、形状の修正された閉曲線（以下単に閉曲線と呼ぶ）の重心点ｐ_ｇ ＝（ｉ_ｇ ， ｊ_ｇ ） を計算する。次に輪郭点の検出処理ＳＰ２１において、図２８に示すように、閉曲線を構成する各点ｐ_ｋ ＝（ｉ_ｋ ，ｊ_ｋ ） に対して、重心ｐ_ｇ から各点に向かう直線に沿つて入力画像の１次微分を計算し、点ｐ_ｋ から見て最初に次式
【数１０】

の条件を満たす点を、真の輪郭線を構成する点ｐ_ｋ ′＝（ｉ_ｋ ′， ｊ_ｋ ′） として検出する。
【００６１】
ここでＸｋ_ｍ は重心点ｐ_ｇ から閉曲線上の点ｐ_ｋ に向かう直線に沿つて入力画像の値を並べた１次元データ列で、Ｘｋ_０ が重心位置における画素値に相当する。 ｄｅｖ_ｍ は、この配列における１次微分の絶対値である。またＴＨ１、ＴＨ２、ＴＨ３は予め設定された正の値である。（１０）式は、レベルそのものが増加していると同時に、図２９に示すように１次微分の絶対値がしきい値ＴＨ３よりも大きく、かつその変化が増加から減少に転じる点を表しており、点ｐ_ｋ から見て最初にこの条件を満足する点が輪郭点として検出される。
【００６２】
最後に、輪郭点の補間処理ＳＰ２２において、マイクロコンピユータ４は検出された輪郭点の間を補間して輪郭線を構成する。この補間処理は第１実施例で述べた補間処理（図１８）とほぼ同じである。ただし第４実施例では、図３０に示すように、隣接する輪郭点の間を全て補間する。またこの場合補間する区間が短いために、（８）式に示すような直線補間を行うことができる。なおこの輪郭点の補間処理の前に、検出された輪郭点の位置を次のように平滑化することもできる。
【００６３】
例えば、検出されたｋ番目の輪郭点と重心点との距離ｌ_ｋ を（３）式のように計算し、これをｋの方向に平滑化する。この平滑化処理には、例えば、１次元の平均化フイルタやメデイアンフイルタ等を用いる。平滑化されたｌ_ｋ ′を用いて新しい輪郭点の位置を（ｉ_ｇ ＋（ｉ_ｋ − ｉ_ｇ ）ｋ′／ ｌ_ｓｋ， ｊ_ｇ ＋（ｊ_ｋ − ｊ_ｇ ）ｋ′／ ｌ_ｓｋ） とすることにより、最終的に滑らかな輪郭線を得ることができる。ここでｌ_ｓｋ′は閉曲線上のｋ番目の点と重心点との距離を表す。
【００６４】
以上の構成によれば、入力された心臓の動画像を、心臓の動きの繰り返し周期よりも短い間隔でサンプリングし、心臓の同じ状態にある画像を同期加算して時間方向のノイズを低減した後、心臓の状態毎に時間平滑化された各画像をそれぞれ空間的に平滑化して、全ての点が抽出すべき輪郭線の内側に存在するように閉曲線を各画像毎に生成し、この各閉曲線を外側に向けて拡大することによつて心臓のおおよその輪郭線を抽出し、当該各輪郭線の形状を修正して心臓の輪郭線として抽出するようにしたことにより、心臓の輪郭線を時相毎に自動的に抽出することができるので、心臓の輪郭線の時間的変化を観察することができ、これにより医療診断を迅速かつ正確に行うことができる。またこれらの輪郭線より輪郭線の速度など診断に必要なパラメータを計算することができる。
【００６５】
（５）第５実施例
この実施例では、心臓の単軸断面の超音波画像より心内膜の輪郭線を抽出する場合について説明する。心臓の単軸断面画像には、図３１に示すように同心円状に並ぶレベルの高い２つの領域Ａ、Ｂが存在し、外側のリング状領域Ａでは、当該領域Ａの上部及び下部に相当する部分Ｃが特にレベルが高くなる傾向がある。抽出すべき心内膜は、内側のリング状領域Ｂの内側面に相当する。
【００６６】
この実施例の医療診断用画像処理装置の構成（図示せず）は画像処理装置４０とほぼ同じであるが、２値画像を保存するための２値画像用フレームメモリ４１は設けられていない。またこの実施例におけるマイクロコンピユータ４による処理の流れは図２５に示したものと同じであるが、各処理の内容は異なる。すわなちマイクロコンピユータ４は、閉曲線の生成処理ＳＰ１では、所望器官の輪郭線の外側に存在する閉曲線として円を生成し、輪郭線の抽出処理ＳＰ２では、生成された内側に向けて縮小することによつておおよその輪郭線形状を抽出する。輪郭線形状の修正処理ＳＰ３は第４実施例と全く同じように処理する。
【００６７】
閉曲線の生成処理（図２５のＳＰ１）の処理手順を図３２に示す。まずマイクロコンピユータ４は、極小点の検出処理ＳＰ１０では、空間的平滑化処理の施されたフレームメモリ３２上の画像から垂直方向の１次元データ列を読み出し、そのレベル変化を解析する。ここで垂直方向のレベル変化の解析を図３３に示す。図３３において、ｉは現在注目している画像上の水平座標を示し、図中の曲線は、水平座標ｉに対応する垂直方向の画素値（レベル）を表している。
【００６８】
このデータ例において、まずｊ＜ｊ_Ｃ の範囲においてレベルが最大となる座標ｊ_Ｌ と、ｊ＞ｊ_Ｃ の範囲においてレベルが最大となる座標ｊ_Ｕ を検出する。ただしｊ_Ｃ は座標入力装置６によつて指定された基準位置ｐ_Ｃ ＝（ｉ_Ｃ 、 ｊ_Ｃ ） の垂直座標である。次にｊ_Ｌ 及びｊ_Ｕ からそれぞれに向けてレベル変化を調べ、最初に遭遇する極小点の座標ｊ_ＬＰ及びｊ_ＵＰを検出する。検出された２つの垂直座標ｊ_ＬＰ及びｊ_ＵＰは画像上での点ｐ_Ｌ ＝（ｉ，ｊ_ＬＰ） 、点ｐ_Ｕ ＝（ｉ，ｊ_ｕＰ） に対応する。これらの点は、上述した同心円状に並ぶ２つの領域に挟まれた部分に存在すると考えられ、抽出すべき心内膜よりも外側に位置する。
【００６９】
次に次式
【数１１】

を満たす全ての水平座標に対して同じ解析を行い、（２Δｉ ＋１）×２個の極小点を検出する。なおΔｉ は予め設定された正の定数である。
【００７０】
次に円近似処理ＳＰ１１において、マイクロコンピユータ４は、検出された（２Δｉ ＋１）個の極小点を１つの円で近似し、その中心座標（ａ， ｂ）及び半径ｒを求める。ここで、検出された複数の極小点を適当な順序に並べ、ｋ番目の点の座標を（ｉ_ｋ ， ｊ_ｋ ） と表す。これらを円の方程式に代入すると、次式
【数１２】

となる。ここでｋ＞０に対応する式よりｋ＝０の式を引くことにより、次式
【数１３】

を得ることができる。これを最小自乗法を用いて解くことにより中心座標（ａ， ｂ）を求める。半径ｒは、得られた中心座標を（１２）式の全ての式に代入し、それぞれから求まるｒを平均することによつて算出する。
【００７１】
次に円の生成処理ＳＰ１２において、マイクロコンピユータ４は、算出された中心座標（ａ， ｂ）と半径ｒとを用いて、円を構成する座標系列Ｓを生成する。これは例えば座標（ａ ， ｂ−ｒ） を開始点として次式
【数１４】

の値を最小とする近傍点の追跡処理を行うことにより生成することができる。近傍点の追跡は第１実施例における補間座標系列の生成と同様に行うことができる。
【００７２】
次に輪郭線の抽出処理（図２５のＳＰ２）の処理手順は、図２７に示す処理手順とほぼ同じであるが、閉曲線の重心位置は円の中心として既に求められているので、重心位置の算出処理（図２７のＳＰ２０）は必要ない。輪郭点の検出処理（図２７のＳＰ２１）において、マイクロコンヒユータ４は、図３４に示すように、円を構成する各点ｐ_ｋ ＝（ｉ_ｋ， ｊ_ｋ ） から重心点（円の中心）ｐ_ｇ （ａ ， ｂ）に向かう直線に沿つて入力画像の１次微分を計算し、最初に次式
【数１５】

を満たす点を、真の輪郭線を構成する点ｐ′_ｋ ＝（ｉ′_ｋ ，ｊ ′_ｋ ） として検出する。（１５）式の各記号は（１０）式で述べたものと同じである。ただしＸｋ_ｍ は円上の点ｐ_ｋ からその中心ｐ_ｇ に向かう直線に沿つて画像値を並べた１次元データ列であり、Ｘｋ_０ は円上の点の画素値に相当する。（１５）式は、レベルそのものが減少していると同時に、図３５に示すように１次微分の絶対値がしきい値ＴＨ３よりも大きく、かつその変化が増加から減少に転じる点を表しており、円上の点から見て最初にこの条件を満足する点が輪郭点として検出される。
【００７３】
以上の構成によれば、入力された心臓の単軸断面の超音波画像を、心臓の動きの繰り返し周期よりも短い間隔でサンプリングし、心臓の同じ状態にある画像を同期加算して時間方向のノイズを低減した後、心臓の状態毎に時間平滑化された各画像をそれぞれ空間的に平滑化して、全ての点が抽出すべき心内膜の輪郭線の外側に存在するように閉曲線を各画像毎に生成し、この各閉曲線を内側に向けて縮小することによつて心内膜のおおよその輪郭線を抽出し、当該各輪郭線の形状を修正して心内膜の輪郭線として抽出するようにしたことにより、心内膜の輪郭線を時相毎に自動的に抽出することができるので、心内膜の輪郭線の時間的変化を観察することができ、これにより迅速かつ正確に医療診断を行うことができる。またこれらの輪郭線より輪郭線の速度など診断に必要なパラメータを計算することができる。
【００７４】
（６）他の実施例
なお上述の実施例においては、医療診断用動画像として心臓の動画像を本発明の医療診断用画像処理装置に適用した場合について述べたが、本発明はこれに限らず、他の周期的な動きのある器官の動画像を本発明の医療診断用画像処理装置に適用しても同様の効果を得ることができる。
【００７５】
また上述の第１実施例においては、入力された静止画像において、境界線の抽出処理（図５）を行うことによつて所望器官の輪郭線を抽出する場合について述べたが、本発明はこれに限らず、第４実施例及び第５実施例のように、閉曲線の生成処理（図２５）によつて所望器官の輪郭線を抽出するようにしてもよい。
【００７６】
この場合、入力された静止画像を空間的に平滑化した後、抽出すべき所望器官の輪郭線に囲まれた領域内の任意の点をその内側に含む閉曲線を生成し、当該閉曲線を平滑化画像のレベル変化に基づいて変形することによつて所望器官のおおよその輪郭線を抽出し、当該輪郭線の形状を修正して所望器官の輪郭線として抽出する。
【００７７】
また上述の第２実施例においては、入力された動画像において、境界線の抽出処理（図５）を行うことによつて所望器官の輪郭線を抽出する場合について述べたが、本発明はこれに限らず、第４実施例及び第５実施例のように、閉曲線の生成処理（図２５）を行うことによつて所望器官の輪郭線を抽出するようにしてもよい。
【００７８】
この場合、入力された動画像を、所望器官の動きの繰り返し周期毎にサンプリングして同期加算して時間方向のノイズを低減した後、この時間平滑化された画像を空間的に平滑化して、抽出すべき所望器官の輪郭線に囲まれた領域内の任意の点をその内側に含む閉曲線を生成し、この閉曲線を平滑化画像のレベル変化に基づいて変形することによつて所望器官のおおよその輪郭線を抽出し、当該輪郭線の形状を修正して所望器官の輪郭線として抽出する。
【００７９】
また上述の実施例においては、医療診断用画像として超音波画像を本発明の医療診断用画像処理装置に適用した場合について述べたが、本発明はこれに限らず、他の医療診断用画像を本発明の医療診断用画像処理装置に適用してもよい。
【００８０】
【発明の効果】
上述のように本発明によれば、入力された医療診断用静止画像を空間的に平滑化した後複数の境界線を抽出し、当該複数の境界線より所望器官の輪郭に対応する境界線を輪郭線として抽出し、当該輪郭線を構成する各点において接線の傾きが大きく変化する変曲点を検出して得られた該変曲点に基づいて輪郭線上の不要な点を除去するようにして当該輪郭線の形状を修正し、所望器官の輪郭線として抽出するようにしたことにより、所望器官の輪郭線を自動的に抽出することができる医療診断用画像処理装置を実現し得る。
【００８１】
また本発明によれば、入力された医療診断用動画像を所望器官の動きの繰り返し周期毎にサンプリングして同期加算して時間方向のノイズを低減した後、この時間平滑化された画像を空間的に平滑化して複数の境界線を抽出し、当該複数の境界線より所望器官の輪郭に対応する境界線を輪郭線として抽出し、当該輪郭線を構成する各点において接線の傾きが大きく変化する変曲点を検出して得られた該変曲点に基づいて輪郭線上の不要な点を除去するようにして当該輪郭線の形状を修正し、所望器官の輪郭線として抽出するようにしたことにより、所望器官の輪郭線を自動的に抽出することができる医療診断用画像処理装置を実現し得る。
【００８２】
さらに本発明によれば、入力された医療診断用動画像を所望器官の動き繰り返し周期より短い間隔でサンプリングし、所望器官の同じ状態にある画像を同期加算して時間方向のノイズを低減した後、所望器官の状態毎に時間平滑化された画像をそれぞれ空間的に平滑化して各画像毎に複数の境界線を抽出し、当該複数の境界線より所望器官の輪郭に対応する境界線を輪郭線として各画像毎に抽出し、当該輪郭線を構成する各点において接線の傾きが大きく変化する変曲点を検出して得られた該変曲点に基づいて輪郭線上の不要な点を除去するようにして当該各輪郭線の形状を修正し、所望器官の輪郭線として抽出するようにしたことにより、所望器官の輪郭線を時相毎に自動的に抽出することができる医療診断用画像処理装置を実現し得る。
【００８３】
さらに本発明によれば、入力された医療診断用静止画像を空間的に平滑化した後、抽出すべき所望器官の輪郭線に囲まれた領域内の任意の点をその内側に含む閉曲線を生成し、この閉曲線を平滑化された画像のレベル変化に基づいて変形して所望器官のおおよその輪郭線を抽出し、当該輪郭線を構成する各点において接線の傾きが大きく変化する変曲点を検出して得られた該変曲点に基づいて輪郭線上の不要な点を除去するようにして当該輪郭線の形状を修正し、所望器官の輪郭線として抽出するようにしたことにより、所望器官の輪郭線を自動的に抽出することができる医療診断用画像処理装置を実現し得る。
【００８４】
さらに本発明によれば、入力された医療診断用動画像を所望器官の動きの繰り返し周期毎にサンプリングして同期加算して時間方向のノイズを低減した後、この時間平滑化された画像を空間的に平滑化して、抽出すべき所望器官の輪郭線に囲まれた領域内の任意の点をその内側に含む閉曲線を生成し、この閉曲線を平滑化された画像のレベル変化に基づいて変形して所望器官のおおよその輪郭線を抽出し、当該輪郭線を構成する各点において接線の傾きが大きく変化する変曲点を検出して得られた該変曲点に基づいて輪郭線上の不要な点を除去するようにして当該輪郭線の形状を修正し、所望器官の輪郭線として抽出するようにしたことにより、所望器官の輪郭線を自動的に抽出することができる医療診断用画像処理装置を実現し得る。
【００８５】
さらに本発明によれば、入力された医療診断用動画像を所望器官の動きの繰り返し周囲よりも短い間隔でサンプリングし、所望器官の同じ状態にある画像を同期加算して時間方向のノイズを低減した後、所望器官の状態毎に時間平滑化された画像をそれぞれ空間的に平滑化して、抽出すべき所望器官の輪郭線に囲まれた領域内の任意の点をその内側に含む閉曲線を各画像毎に生成し、これらの各閉曲線を平滑化された画像のレベル変化に基づいて変形して所望器官のおおよその輪郭線を抽出し、当該輪郭線を構成する各点において接線の傾きが大きく変化する変曲点を検出して得られた該変曲点に基づいて輪郭線上の不要な点を除去するようにして当該輪郭線の形状を修正し、所望器官の輪郭線として抽出するようにしたことにより、所望器官の輪郭線を時相毎に自動的に抽出することができる医療診断用画像処理装置を実現し得る。
【図面の簡単な説明】
【図１】本発明の第１実施例による医療診断用画像処理装置の構成を示すブロツク図である。
【図２】画像の走査手順を示す略線図である。
【図３】空間フイルタのフイルタ係数の一例を示す図表である。
【図４】空間フイルタのフイルタ係数の一例を示す図表である。
【図５】マイクロコンピユータの処理の流れの説明に供するフローチヤートである。
【図６】境界線抽出処理の処理手順の説明に供するフローチヤートである。
【図７】境界線追跡のための近傍探索の説明に供する略線図である。
【図８】境界線の追跡方向の説明に供する略線図である。
【図９】輪郭線抽出処理の処理手順の説明に供するフローチヤートである。
【図１０】輪郭線の選択処理の説明に供する略線図である。
【図１１】輪郭線形状修正処理の処理手順の説明に供するフローチヤートである。
【図１２】隣接領域が融合した場合の輪郭形状の一例を示す略線図である。
【図１３】くぼみを有する輪郭形状の一例を示す略線図である。
【図１４】変曲点の検出の説明に供する略線図である。
【図１５】融合領域の除去の説明に供する略線図である。
【図１６】輪郭線が閉じている場合の融合領域の除去の説明に供する略線図である。
【図１７】くぼみの除去の説明に供する略線図である。
【図１８】補間処理の説明に供する略線図である。
【図１９】補間点の検出の説明に供する略線図である。
【図２０】本発明の第２実施例による医療診断用画像処理装置の構成を示すブロツク図である。
【図２１】画像サンプリング器の構成を示すブロツク図である。
【図２２】心電図波形における基準点の説明に供する波形図である。
【図２３】本発明の第３実施例による医療診断用画像処理装置の構成を示すブロツク図である。
【図２４】本発明の第４実施例による医療診断用画像処理装置の構成を示すブロツク図である。
【図２５】第４実施例におけるマイクロコンピユータの処理の流れを示すフローチヤートである。
【図２６】第４実施例における閉曲線生成処理の処理手順を示すフローチヤートである。
【図２７】第４実施例における輪郭線抽出処理の処理手順を示すフローチヤートである。
【図２８】第４実施例における輪郭点の探索方向の説明に供する略線図である。
【図２９】第４実施例における輪郭点の検出の説明に供する波形図である。
【図３０】第４実施例における輪郭点の補間の説明に供する略線図である。
【図３１】心臓の単軸断面画像を示す略線図である。
【図３２】第５実施例における閉曲線生成処理の処理手順を示すフローチヤートである。
【図３３】極小点の検出処理の説明に供する波形図である。
【図３４】第５実施例における輪郭点の探索方向の説明に供する略線図てある。
【図３５】第５実施例における輪郭点の検出の説明に供する波形図である。
【符号の説明】
１、２０、３０、４０……医療診断用画像処理装置、２……アナログデイジタル変換回路（Ａ／Ｄ変換器）、３、３２……フレームメモリ、４……マイクロコンピユータ、５……画像表示装置、６……座標入力装置、７、１０、１１……メモリ、８……空間フイルタ、９……２値化器、２１……遅延素子、２２、３１……画像サンプリング器、２３……フレーム同期信号生成器、２４……心電図モニタ装置。４１……２値画像用フレームメモリ、４２……低域通過型空間フイルタ。[0001]
【table of contents】
The present invention will be described in the following order.
Industrial application fields
Conventional technology
Problems to be solved by the invention
Means for solving the problems (FIGS. 1, 20, 21, 23, and 24)
Action
Example
(1) 1st Example (FIGS. 1-19)
(2) Second embodiment (FIGS. 20 to 22)
(3) Third embodiment (FIG. 23)
(4) Fourth embodiment (FIGS. 24 to 30)
(5) Fifth embodiment (FIGS. 31 to 35)
(6) Other embodiments
The invention's effect
[0002]
[Industrial application fields]
The present invention relates to an image processing apparatus for medical diagnosis, and is suitable for application to medical diagnosis using, for example, an ultrasonic image or an NMR (nuclear magnetic resonance) image.
[0003]
[Prior art]
2. Description of the Related Art Today, in the medical field, as one method of medical diagnosis, a doctor uses, for example, an ultrasonic image obtained by an ultrasonic diagnostic apparatus, and a doctor uses a cardiovascular system such as a patient's endocardium, heart valve, and arterial intima It is known to perform medical diagnosis by capturing organ movements. In this case, a doctor generally performs a medical diagnosis by a method of “tracing” the boundary of each organ while visually observing a still image.
[0004]
[Problems to be solved by the invention]
However, since the ultrasonic image obtained by the ultrasonic diagnostic apparatus is unclear, there is a large error in the method in which the doctor makes a diagnosis by tracing the boundary of each organ while visually observing the still image as described above. Therefore, there is a problem that an accurate diagnosis cannot be performed and the diagnosis takes time.
[0005]
The present invention has been made in view of the above points, and an object of the present invention is to propose a medical image processing apparatus that can automatically extract the contour of a desired organ.
[0006]
[Means for Solving the Problems]
In order to solve such a problem, in the present invention, in the medical diagnostic image processing apparatus (1) for extracting the contour line of the desired organ from the medical diagnostic still image, the inputted medical diagnostic still image is spatially smoothed. Space filtering means (8), boundary line extraction means (4) for extracting a plurality of boundary lines from the image obtained by the space filtering means (8), and boundary line extraction means (4) A contour line extracting means (4) for extracting a boundary line corresponding to the contour of the desired organ as a contour line from the plurality of boundary lines, and correcting the shape of the contour line extracted by the contour line extracting means (4) A contour line shape correcting means (4), and the contour line shape correcting means (4) greatly changes the slope of the tangent line at each point constituting the contour line extracted by the contour line extracting means (4). Detecting an inflection point to be It was followed by removal of unnecessary points on the contour line based on.
[0007]
In the present invention, in the medical diagnosis image processing apparatus (20) for extracting the contour line of a desired organ having a periodic motion from the medical diagnosis moving image, the medical diagnosis moving image is generated every repetition cycle of the desired organ motion. An image sampling means (22) for inputting an image, a synchronous addition means (22D) for adding a plurality of images obtained by the image sampling means (22) for each pixel to generate a time-smoothed image, and synchronization Spatial filtering means (8) for spatially smoothing the image obtained by the adding means (22D), and boundary line extraction means (4) for extracting a boundary line from the image obtained by the spatial filtering means (8). ), And a contour line extracting unit (4) for extracting a boundary line corresponding to the contour of the desired organ as a contour line from the boundary line extracted by the boundary line extracting unit (4), and a contour line extracting unit (4) According to Contour shape correcting means (4) for correcting the shape of the contour line formed, and the contour line shape correcting means (4) constitutes the contour lines extracted by the contour line extracting means (4). An inflection point at which the tangent slope changes greatly at a point is detected, and unnecessary points on the contour line are removed based on the inflection point.
[0008]
Furthermore, in the present invention, in the medical diagnosis image processing apparatus (30) for extracting the contour line of the desired organ having a periodic motion from the medical diagnostic moving image, the medical diagnosis is performed at an interval shorter than the repetition cycle of the desired organ motion. An image sampling means (31) for inputting a moving image, and images corresponding to the same state of the desired organ are collected from the images obtained by the image sampling means (31) and added for each pixel, and added for each state of the desired organ. A synchronous addition means (22D) for generating a time-smoothed image, and a spatial filtering means (8) for spatially smoothing an image for each desired organ state obtained by the synchronous addition means (22D). , A boundary line extraction means (4) for extracting a boundary line from each image obtained by the spatial filtering means (8), and a desired organ ring from the boundary line extracted by the boundary line extraction means (4) A contour line extracting unit (4) for extracting a boundary line corresponding to the contour line as a contour line, a contour line shape correcting unit (4) for correcting the shape of each contour line extracted by the contour line extracting unit (4), and The contour line shape correcting means (4) detects an inflection point at which the inclination of the tangent line greatly changes at each point constituting the contour line extracted by the contour line extracting means (4). Unnecessary points on the contour line are removed based on the curved points.
[0009]
Furthermore, in the present invention, in the medical diagnostic image processing apparatus (1) for extracting a contour line of a desired organ from a medical diagnostic still image, spatial filtering means for spatially smoothing the inputted medical diagnostic still image. (8), closed curve generating means (4) for generating a closed curve including an arbitrary point in the region surrounded by the contour line of the desired organ to be extracted, and smoothing by the spatial filtering means (8) Based on the level change of the image, the closed curve generated by the closed curve generating means (4) is deformed, and the closed curve deforming means (4) for extracting the approximate contour of the desired organ, and the closed curve deforming means (4). Contour line shape correcting means (4) for correcting the shape of the obtained contour line is provided, and the contour line shape correcting means (4) constitutes the contour line extracted by the contour line extracting means (4). Tangential slope at each point Detecting an inflection point that varies greatly, and so as to remove unnecessary points on the contour line on the basis of the inflection point.
[0010]
Furthermore, in the present invention, in the medical diagnosis image processing apparatus (20) for extracting a contour line of a desired organ having a periodic movement from a moving image for medical diagnosis, a moving image for medical diagnosis for each repetition cycle of the movement of the desired organ An image sampling means (22) for inputting an image, a synchronous addition means (22D) for adding a plurality of images obtained by the image sampling means (22) for each pixel to generate a time-smoothed image, and synchronization Spatial filtering means (8) for spatially smoothing the image obtained by the adding means (22D), and a closed curve including an arbitrary point within the region surrounded by the contour line of the desired organ to be extracted Based on the level change of the image spatially smoothed by the closed curve generating means (4) and the spatial filtering means (8), the closed curve obtained by the closed curve generating means (4) is deformed and desired vessel A closed curve deforming means (4) for extracting an approximate contour line, and a contour shape correcting means (4) for correcting the shape of the contour line obtained by the closed curve deforming means (4). (4) detects an inflection point at which the slope of the tangent line greatly changes at each point constituting the contour line extracted by the contour line extraction means (4), and on the contour line based on the inflection point. Unnecessary points were removed.
[0011]
Furthermore, in the present invention, in the medical diagnosis image processing apparatus (40) for extracting the contour line of a desired organ having a periodic movement from the medical diagnosis moving image, the medical treatment is performed at an interval shorter than the repetition cycle of the movement of the desired organ. An image sampling means (31) for inputting a diagnostic moving image, and images corresponding to the same state of the desired organ are collected from the images obtained by the image sampling means (31), and are added for each pixel, for each state of the desired organ. Synchronous addition means (22D) for generating a time-smoothed image and spatial filtering means (8) for spatially smoothing the images for each desired organ state obtained by the synchronous addition means (22D). Closed curve generating means (4) for generating for each image a closed curve including an arbitrary point in the region surrounded by the contour line of the desired organ to be extracted, and spatial filtering means (8) Closed curve deformation means (4) for deforming each closed curve obtained by the closed curve generation means (4) based on the level change of each smoothed image and extracting an approximate contour line of the desired organ for each image; The contour shape correcting means (4) for correcting the shape of each contour line obtained by the closed curve deforming means (4) is provided. The contour shape correcting means (4) is provided by the contour line extracting means (4). Thus, an inflection point where the tangent slope changes greatly at each point constituting the extracted contour line is detected, and unnecessary points on the contour line are removed based on the inflection point.
[0012]
[Action]
A plurality of boundary lines are extracted after spatially smoothing the input still image for medical diagnosis. A boundary line corresponding to the contour of the desired organ is extracted as a contour line from the plurality of boundary lines, and the shape of the contour line is corrected and extracted as a contour line of the desired organ. When correcting the shape of the contour line, an inflection point at which the inclination of the tangent changes greatly at each point constituting the contour line is detected, and unnecessary points on the contour line are removed based on the inflection point. I made it. As a result, the contour line of the desired organ can be automatically extracted, so that medical diagnosis can be performed quickly and accurately.
[0013]
In addition, the input medical diagnostic moving image is sampled at every repetition period of the desired organ movement and synchronously added to reduce noise in the time direction, and then the time-smoothed image is spatially smoothed. Extract multiple borders. A boundary line corresponding to the contour of the desired organ is extracted as a contour line from the plurality of boundary lines, and the shape of the contour line is corrected and extracted as a contour line of the desired organ. When correcting the shape of the contour line, an inflection point at which the inclination of the tangent changes greatly at each point constituting the contour line is detected, and unnecessary points on the contour line are removed based on the inflection point. I made it. As a result, the contour line of the desired organ can be automatically extracted, so that medical diagnosis can be performed quickly and accurately.
[0014]
In addition, the input video images for medical diagnosis are sampled at intervals shorter than the desired organ movement cycle, and the images in the same state of the desired organ are synchronously added to reduce noise in the time direction, and then the state of the desired organ Each time-smoothed image is spatially smoothed to extract a plurality of boundary lines for each image. A boundary line corresponding to the contour of the desired organ is extracted from each of the plurality of boundary lines as a contour line for each image, and the shape of each contour line is corrected and extracted as the contour line of the desired organ. When correcting the shape of the contour line, an inflection point at which the inclination of the tangent changes greatly at each point constituting the contour line is detected, and unnecessary points on the contour line are removed based on the inflection point. I made it. Thereby, since the contour line of the desired organ can be automatically extracted for each time phase, the temporal change of the contour line of the desired organ can be observed, and medical diagnosis can be performed quickly and accurately. Also, parameters necessary for diagnosis such as the speed of the contour line can be calculated from these contour lines.
[0015]
Further, after spatially smoothing the inputted still image for medical diagnosis, a closed curve including an arbitrary point in the region surrounded by the contour line of the desired organ to be extracted is generated. The closed curve is deformed based on the level change of the smoothed image to extract an approximate contour line of the desired organ, and the shape of the contour line is corrected and extracted as the contour line of the desired organ. When correcting the shape of the contour line, an inflection point at which the inclination of the tangent changes greatly at each point constituting the contour line is detected, and unnecessary points on the contour line are removed based on the inflection point. I made it. Thereby, since the outline of a desired organ can be extracted automatically, medical diagnosis can be performed quickly and accurately.
[0016]
Further, the input moving image for medical diagnosis is sampled at every repetition cycle of the desired organ movement and synchronously added to reduce noise in the time direction. After the time-smoothed image is spatially smoothed, a closed curve including an arbitrary point in the region surrounded by the contour line of the desired organ to be extracted is generated. The closed curve is deformed based on the level change of the smoothed image to extract an approximate contour line of the desired organ, and the shape of the contour line is corrected and extracted as the contour line of the desired organ. When correcting the shape of the contour line, an inflection point at which the inclination of the tangent changes greatly at each point constituting the contour line is detected, and unnecessary points on the contour line are removed based on the inflection point. I made it. Thereby, since the outline of a desired organ can be extracted automatically, medical diagnosis can be performed quickly and accurately.
[0017]
In addition, the input moving image for medical diagnosis is sampled at intervals shorter than the surroundings where the desired organ moves repeatedly, and images in the same state of the desired organ are synchronously added to reduce noise in the time direction. Next, each time-smoothed image of each desired organ state is spatially smoothed, and each closed curve including an arbitrary point within the region surrounded by the contour line of the desired organ to be extracted is included. Generated for each image. Each of these closed curves is deformed based on the level change of the smoothed image to extract an approximate contour line of the desired organ. When correcting the shape of the contour line, an inflection point at which the inclination of the tangent changes greatly at each point constituting the contour line is detected, and unnecessary points on the contour line are removed based on the inflection point. I made it. As a result, since the contour line of the desired organ can be automatically extracted for each time phase, the temporal change of the contour line of the desired organ can be observed, so that medical diagnosis can be performed quickly and accurately. . Also, parameters necessary for diagnosis such as the speed of the contour line can be calculated from these contour lines.
[0018]
【Example】
Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.
[0019]
(1) First embodiment
In FIG. 1, reference numeral 1 denotes an image processing apparatus for medical diagnosis according to a first embodiment of the present invention as a whole. In this medical diagnosis image processing apparatus 1, a plurality of boundary lines are extracted after spatially smoothing a still image for medical diagnosis of an input ultrasonic image, and a boundary corresponding to the contour of a desired organ is extracted from the boundary line. A line is selected as a contour line, and the shape of the contour line is corrected and extracted as a contour line of a desired organ.
[0020]
The input image signal input to the analog digital converter (A / D converter) 2 is digitized and then stored in the frame memory 3. Here, the input image is assumed to be a grayscale image having only a luminance component. The image stored in the frame memory 3 is displayed on the image display device 5 under the instruction of the microcomputer 4.
[0021]
The coordinate input device 6 is a pointing device such as a mouse, for example, and has coordinates (i) of a substantially central position (hereinafter referred to as a reference position) of a desired organ region. _c, j _c ) And is controlled by the microcomputer 4. Coordinates can be input by designating a certain position on the image display device 5 with the coordinate input device 6, and the designated coordinates are stored in the memory 7. When the coordinate input is completed by the coordinate input device 6, the image data on the frame memory 3 is read out in the scanning order shown in FIG. 2 and input to the spatial filter 8.
[0022]
In the spatial filter 8, the following formula
[Expression 1]

The filtering process is performed. Here, X (i, j) represents the coordinate value of the position (i, j) of the input image X, and Y (i, j) represents the position (i, j) of the image output from the spatial filter 8. Also a _didj Represents a two-dimensional filter coefficient. As the filter coefficient, for example, a filter coefficient as shown in FIG. 3 or FIG. 4 is used. A filter having a filter coefficient as shown in FIG. 3 is a low-pass filter, and when this low-pass filter is used, a boundary line obtained by boundary line extraction described later is a contour line of a low-frequency component. A filter having a filter coefficient as shown in FIG. 4 is a smoothed differential filter called Laplacian G, which is a digitalized second-order derivative of a Gaussian function. The boundary line obtained by the smoothing differential filter corresponds to an edge having a large luminance change.
[0023]
The output of the space filter 8 is input to the binarizer 9, and the binarizer 9 binarizes the output of the space filter 8 with a preset threshold value TH1. That is, when the input value is larger than the threshold value TH1, it is converted to “1”, and when it is smaller, it is converted to “0”. This output value is overwritten at a predetermined position in the frame memory 3.
[0024]
When the binarization process for the entire input image is completed in this way, as shown in FIG. 5, the micro computer 4 executes a boundary line extraction process SP1, a contour line extraction process SP2, and a contour line shape correction process SP3. These processes by the microcomputer 4 will be described below. As will be described later, the memory 10 and 11 store the calculation results by the microcomputer 4. That is, the memory 10 stores the coordinate series of the points constituting the selected contour line, and the memory 11 stores the information of the inflection point for correcting the shape of the contour line.
[0025]
First, the processing procedure of the boundary line extraction process SP1 by the microcomputer 4 will be described. As shown in FIG. 6, in the boundary line start point detection process SP10 as the first process of the boundary line extraction process SP1, the micro computer 4 scans the frame memory 3 in the order shown in FIG. formula
[Expression 2]

Check whether the relationship is satisfied. In the equation (2), X (i, j) represents a coordinate value at the position (i, j) on the frame memory 3. When scanning the frame memory 3, the position (i, j) satisfying the expression (2) is first set as the boundary line start position. In the following, p is used to represent the position on the frame memory 3, and for example, when the coordinates of the point are (i, j), p = (i, j). The detected boundary start point is p ₀ And
[0026]
Boundary start point p ₀ Is detected, the value (“0” or “1”) of the binary image at this position is stored in the register r. ₀ And a new label L is generated and the boundary start point p on the frame memory 3 is generated. ₀ The value “L” is stored at the position of. Here, an integer of “2” or more is used as the label number L so as not to be confused with a binary image stored in advance. The boundary line extraction process SP1 ends when scanning for detecting the boundary start point reaches the lower right of the image.
[0027]
Next, the microcomputer 4 detects the boundary start point p detected in the boundary line tracking process SP11 as the second process of the boundary line extraction process SP1. ₀ Point sequence (boundary point) p constituting the boundary line including _k (K = 0, 1,...) Are sequentially tracked by neighborhood search. This neighborhood search process will be described with reference to FIG. In FIG. 7 representing the binary image on the frame memory 3, now the boundary point p _k Is determined. At this time, the boundary point p _k The values of 8 pixels in the vicinity of are sequentially checked clockwise, and the register r ₀ The same value (ie boundary point p _k The first position with the same value) as the next boundary point p _{k + 1} And Boundary point p _k Which of the 8 neighboring pixels starts the search depends on the boundary point p _k Boundary point p established one before _k-1 And boundary point p _k It is determined by the positional relationship.
[0028]
That is, the boundary point p _k-1 Boundary point p when viewed clockwise among the 8 neighboring pixels centered on _k The position where the order is one before the boundary point p _{k + 1} As the search start point when In FIG. 7, the boundary point p _k The pixel located above becomes the start point of the neighborhood search, and the pixel on the right is the boundary point p _{k + 1} Detected as P on the frame memory 3 _{k + 1} The label number L is stored at the position of. (A) The next boundary point is the boundary start point p. ₀ Or (B) the next boundary point is the previous boundary point p _k-1 Repeat until one of the two conditions is met.
[0029]
Condition (A) means that the boundary line is closed, and tracking of one boundary line is completed. In this case, the current boundary start point p is used to detect the start point of the next boundary line. ₀ The boundary start point detection processing SP10 is performed again from the next scanning position. Condition (B) means that the boundary line being tracked is not closed. In this case, as shown in FIG. ₀ Returning to, the boundary line is traced in the opposite direction by performing the neighborhood search counterclockwise. Boundary start point p when tracking in the opposite direction is finished ₀ Boundary start point P from the next scan position ₀ Restart detection.
[0030]
When the boundary line is extracted in this way, the microcomputer 4 starts the contour line extraction process SP2. As shown in FIG. 9, first, in the outline selection processing SP20, the micro computer 4 scans the frame memory 3 for each boundary line extracted by the above-described method, and positions (i, j ) And a reference position (i _c, j _c ) And the distance d
[Equation 3]

To calculate. The distance d is calculated for all the positions having the label L, and the minimum value is calculated as the distance d between the boundary line of the label L and the reference position as shown in FIG. _l And The distance d from the reference position among all the boundary lines _l The boundary line with the smallest is selected as the contour line. The distance d from the reference position of each boundary line _l Can also be calculated in the boundary tracking described above.
[0031]
Next, in the contour coordinate series generation processing SP21, the microcomputer 4 sequentially stores the coordinates of the points constituting the contour selected in the frame memory 3 in the memory 10 to generate a coordinate series. This processing procedure is the same as the boundary line extraction process shown in FIG. 6, but in the boundary line start point detection process, instead of evaluating the formula (2) at each position, the label at that position is selected. Check if it is a contour line. In other words, the position where the outline label is first encountered during scanning is the boundary start point p. ₀ And Since there is only one target to be tracked, the boundary line start point can be detected only once.
[0032]
In the generation of the coordinate series, the tracking stop condition (B) described above, that is, the next boundary point is the previous boundary point p. _k-1 If they match, the order of the coordinate series generated up to this point is reversed, and then backward tracking is started. When the backward tracking is completed, the order of all coordinate series is reversed again. This ensures that the generated coordinate series is in the order of tracing the contour line clockwise.
[0033]
When the contour line extraction process SP2 is completed in this way, the microcomputer 4 starts the contour line shape correction process SP3 according to the processing procedure shown in FIG. In this case, the case of extracting the outline of an almost elliptical organ such as the heart will be described, and the case of correcting the outline shape as shown in FIGS. 12 and 13 will be described. A possible cause of the generation of the contour line as shown in FIG. 12 is the fusion of adjacent regions for spatial smoothing. Further, as shown in FIG. 13, the dent is considered to occur when another organ overlaps the desired organ, and the papillary muscle in the ultrasonic image of the heart is one example.
[0034]
First, the microcomputer 4 performs an inflection point detection process SP30 as the contour shape correction process SP3, and calculates the curvature ρ for each point of the coordinate series stored in the memory 10 as shown in FIG.
[Expression 4]

To calculate. Where coordinates (i _n, j _n ) Is the nth coordinate stored in the memory 10, and k is a constant for calculating the inclination of the outline before and after the point of interest. When the absolute value of the curvature ρ is larger than a preset threshold value TH2, this point is regarded as an inflection point, and the value of the curvature ρ is stored in the nth of the memory 11. When the absolute value of the curvature ρ is equal to or less than the threshold value TH2, “0” is stored in the nth of the memory 11. Here, as shown in FIG. 14, the sign of the equation (4) represents the type of inflection point. That is, when the value obtained by equation (4) is a positive value, this inflection point is considered concave with respect to the center of the contour, and conversely, when it is negative, it is convex. it is conceivable that.
[0035]
Here, the fusion region removal processing SP31 is realized by removing a portion sandwiched between concave inflection points having a particularly large curvature ρ, as shown in FIG. That is, a point having a positive value in the memory 11 and having an absolute value larger than a preset threshold value TH3 is set as a fusion contact. The threshold value TH3 is set to a value larger than the threshold value TH2 used in the inflection point detection process SP30. When two or more fusion contacts are detected, the linear distance on the image to the next fusion contact is calculated for each fusion contact. When the distance between adjacent fusion contacts is smaller than a preset threshold TH4, a special value ρ∞ is substituted for each position on the memory 11 between the two fusion contacts, and the contour line Remove more. As shown in FIG. 16, when the contour line is closed, the center of gravity is calculated for each of the two portions of the contour line divided by the two fusion contacts, and the one far from the reference position is removed.
[0036]
In the equation (4), depending on how to select the constant k and the threshold value TH4, there are cases where a plurality of connected points are detected as fusion contacts, although they are substantially one fusion contact. In this case, one of the connected fusion contacts is selected as a representative point, and a portion sandwiched between the representative points is removed. For example, as the representative point, the one having the maximum absolute value of the curvature ρ can be selected from the fusion contacts to be connected.
[0037]
Next, in the indentation removal processing SP32, the micro computer 4, as shown in FIG. 17, between the two convex inflection points detected at the threshold TH2, the concave detected at the same threshold TH2. When an inflection point exists, it is realized by removing a portion sandwiched between two convex inflection points. That is, first, the values of the memory 11 that has been subjected to the fusion region removal process are sequentially checked to detect a concave inflection point having a positive value. Next, a convex inflection point having a negative value is searched in both directions from the concave inflection point, and the first inflection point encountered in each direction is detected. A special value ρ∞ is substituted for each position on the memory 11 between these two convex inflection points and removed from the contour line.
[0038]
In this case, as in the case of the above-described fusion contact, a concave inflection point or a convex inflection point may be detected by being connected. In this case, a representative point may be used. If the relative position of the indentation resulting from this, such as the papillary muscle in the ultrasound image of the heart, can be roughly predicted, the inflection point and the appearance position of the convex inflection point should Can be limited. That is, the indented portion can be removed only for the inflection point that appears in a certain limited range.
[0039]
The interpolation process is a process for interpolating the unconnected portion generated in the contour line by the two removal processes described above, and generates a coordinate series of a new connected contour line. As shown in FIG. 18, the end point coordinates of the contour line generated as a result of removing a certain part are expressed as p. _s = (I _s, j _s ), P _e = (I _e , J _e ) For example, this end point p _s And p _e The following formula
[Equation 5]

Interpolate with a quadratic function. In the formula (5), t is a parameter and p _m = (I _m , J _m ) Is the center of gravity p of the contour line configuration point remaining after the above two removal processes _g To be determined as shown in FIG. That is, point p _m Is the end point p _s Through the center of gravity p _g And end point p _s A straight line perpendicular to the line segment connecting the two and the end point p _e Through the center of gravity p _g And end point p _e Is an intersection with a straight line perpendicular to the line segment connecting the two. According to the equation (5), the end point p _s And end point p _e In order to generate an interpolated coordinate series connecting ₀ = P _s Repeat the following process.
[0040]
Now point p _n If the interpolation coordinate series up to is fixed, the interpolation point p _n At several neighboring points p = (i, j)
[Formula 6]

And the point where the value of s is minimum is determined as the next interpolation point p. _{n + 1} And Here, the neighborhood where the equation (6) is calculated is the interpolation point p as shown in FIG. _n And the point p that was previously determined _n-1 It is determined by the positional relationship. That is, the interpolation point p _n Point p between _n-1 (6) is calculated with 5 pixels on the opposite side (part indicated by hatching in FIG. 19). Also, t in equation (6) ₁ And t ₂ Each must be a real number and must be a value between 0.0 and 1.0, and among the neighboring points shown in FIG. Choose one. Interpolation point p _{n + 1} Is the end point p _e The interpolated coordinate series is completed at the point in time that End point p on the memory 11 _s The end point p after _e The previous coordinates are deleted, and the generated coordinate series of interpolation points is inserted instead of the deleted coordinates. Such an interpolation process is applied between all end points generated by the removal process to generate a new contour line coordinate series. In this case, if necessary, each point in the coordinate series can be written into the frame memory 3 and displayed on the image display device 5.
[0041]
According to the above configuration, a boundary for which the distance d from the reference position is the smallest among the boundary lines by extracting a plurality of boundary lines after spatially smoothing the medical diagnostic still image of the input ultrasonic image By selecting a line as the contour line, correcting the shape of the contour line and extracting it as the contour line of the desired organ, the contour line of the desired organ can be automatically extracted, so medical diagnosis can be performed quickly and accurately It can be carried out.
[0042]
(2) Second embodiment
In FIG. 20, in which parts corresponding to those in FIG. 1 are assigned the same reference numerals, 20 denotes an image processing apparatus for medical diagnosis according to a second embodiment of the present invention as a whole. In the medical diagnostic image processing apparatus 20, the medical diagnostic moving image of the input ultrasonic image is sampled at every repetition cycle of the movement of the desired organ and synchronously added to reduce the noise in the time direction. A plurality of boundary lines are extracted by spatially smoothing the converted image. A boundary line corresponding to the contour of the desired organ is extracted from the boundary line as a contour line, and the shape of the contour line is corrected and extracted as the contour line of the desired organ.
[0043]
Here, in this embodiment, a moving image of the heart having a periodic motion is used as a moving image for medical diagnosis, and a moving image of the heart is used as a moving image for medical diagnosis in the third and fourth embodiments described later. Is used.
[0044]
The delay element 21 delays the input data by a time necessary for the analysis of the electrocardiogram waveform in the image sampling unit 22. The frame synchronization signal generator 23 outputs a signal indicating the start timing of the frame before being delayed by the delay element 21 to the image sampler 22. The image sampler 22 determines whether or not to sample each frame at the start of input of each frame based on the electrocardiogram waveform sent from the electrocardiogram monitor device 24 in synchronization with the image, and time averages the sampled image. A value image is generated on the frame memory 3. After time averaging for a preset period is performed, a contour line is extracted by the same processing as in the first embodiment. The value in the frame memory 3 is reset at the start of measurement. If necessary, the extracted contour line can be written in the frame memory 3 and displayed on the image display device 5.
[0045]
The configuration of the image sampler 22 is shown in FIG. As shown in FIG. 21, the image sampling device 22 includes a reference point detector 22A, a timer 22B, a time phase detector 22C, and a synchronous adder 22D. The reference point detector 22A detects the reference point of the electrocardiogram waveform inputted from the electrocardiogram monitor device 24 every moment. The reference point is a point that appears in the electrocardiogram waveform at each repetition period, and for example, as shown in FIG. When the reference point is detected, the timer 22B is reset, and after reset, the timer 22B increments the value at a constant time interval until the next reset.
[0046]
When the time phase detector 22C receives a signal indicating the start of a frame from the frame synchronization signal generator 23, the time phase detector 22C reads the value of the timer 22B and calculates the difference from a preset value t (hereinafter referred to as a time phase). calculate. Only when the value t is equal to or less than the threshold value TH5, the next frame data to be input is sent to the synchronous adder 22D. In other cases, the input data is discarded. In the synchronous adder 22D, the data on the frame memory 3 corresponding to the input data is read, and the average value with the input data is expressed by the following equation:
[Expression 7]

To calculate.
[0047]
Where X _n (I, j) represents the data value at the position (i, j) in the nth sampled image from the time when the image starts to be sampled, and X ′ _n-1 (I, j) represents the data value of the position (i, j) in the average image up to the (n-1) th, that is, the value of the data stored in the frame memory 3 when the nth image is sampled. Yes. The calculation result obtained by the equation (7) is stored in the same position in the frame memory 3. However, immediately after the start of measurement, the average value of the input data and the value on the frame memory 3 is not calculated, and the input data is written in the frame memory 3 as it is.
[0048]
Here, the time phase at which the image is sampled can be specified not as the time from the reference point but as the phase in the repetition period. That is, the reference point detector 22A detects the time T from one reference point to the next reference point, and samples the frame input at the time T / a from the first reference point. According to this method, fluctuations in the period at each repetition can be absorbed, so that synchronous addition can be performed more accurately.
[0049]
According to the above configuration, after the input heart moving image is sampled at every repetition period of the heart motion and synchronously added to reduce noise in the time direction, the time-smoothed image is further spatialized. A plurality of boundary lines are extracted by smoothing, a boundary line having the shortest distance d from the reference position is selected as the contour line, and the shape of the contour line is corrected to obtain a heart contour line. By extracting, the outline of the heart can be automatically extracted, so that medical diagnosis can be performed quickly and accurately.
[0050]
(3) Third embodiment
In FIG. 23, in which parts corresponding to those in FIG. 20 are assigned the same reference numerals, 30 denotes an image processing apparatus according to the third embodiment of the present invention as a whole. In this image processing device 30, the medical diagnostic moving image of the input ultrasonic image is sampled at an interval shorter than the repetition cycle of the desired organ, and the images in the same state of the organ are synchronously added to the time direction. After reducing the noise, each time-smoothed image for each desired organ state is spatially smoothed to extract a plurality of boundary lines for each time phase. A boundary line corresponding to the contour of the desired organ is selected from a plurality of boundary lines as a contour line for each time phase, and the shape of each contour line is modified to extract the contour line of the desired organ for each time phase. ing.
[0051]
The image sampler 31 is substantially the same as the image sampler 22, but a plurality of values t0, t1,... Are preset as time phases for sampling an image. For the images sampled at each time phase in the image sampling unit 31, an average image is generated for each time phase and output to the frame memory 32 corresponding to each time phase. Here, the same number of frame memories 32 as the number of sampling time phases are prepared, and each frame memory 32 stores an average image of simultaneous phases. As described above, after the time averaging for a preset period is performed, a contour line is extracted from the average image on each frame memory 32 by the same process as in the first embodiment.
[0052]
If necessary, a plurality of contour lines having different time phases may be overwritten in one frame memory 33 of the frame memory 32 and displayed as a still image by the image display device 5. Further, the contour lines of the respective time phases may be written in different frame memories 32, and these may be temporally switched and displayed on the image display device 5.
[0053]
According to the above configuration, the input moving image of the heart is sampled at a plurality of preset time phases in the repetition period of the heart motion, and the noise in the time direction is reduced by synchronous addition for each simultaneous phase. After that, the time-smoothed images at different time phases are spatially smoothed to extract a plurality of boundary lines from each image, and the boundary having the smallest distance d from the reference position among the plurality of boundary lines By extracting a line as a contour line, and correcting the shape of each contour line and extracting it as a heart contour line, the heart contour line can be automatically extracted for each time phase. A temporal change in the outline of the heart can be observed, whereby a medical diagnosis can be performed quickly and accurately. Further, parameters necessary for diagnosis such as the speed of the contour line can be calculated from the extracted contour line.
[0054]
(4) Fourth embodiment
In FIG. 24, in which parts corresponding to those in FIG. 23 are assigned the same reference numerals, 40 denotes an image processing apparatus according to the fourth embodiment of the present invention as a whole. The image processing apparatus 40 is provided with a binary image frame memory 41 for storing one binary image, and uses a low-pass filter 42 shown in FIG. 3 as a spatial filter.
[0055]
That is, in the image processing device 40, as in the image processing device 30, a spatial smoothing process is performed on the image subjected to time averaging by the spatial filter 42, and the output of the spatial filter 42 is directly written in the frame memory 32. It is. The microcomputer 4 extracts a contour line by generating a closed curve, extracting a contour line, and correcting a contour shape as shown in FIG. 25 for each image for which the spatial smoothing process has been completed. That is, the microcomputer 4 generates a closed curve existing inside the contour line of the desired organ in the closed curve generation process SP1, and expands the obtained closed curve outward in the contour line extraction process SP2. The approximate contour shape is extracted, and the contour shape correction processing SP3 corrects the extracted contour shape by the same method as the contour shape correction processing (FIG. 11) described in the first embodiment.
[0056]
FIG. 26 shows a processing procedure of the closed curve generation process (SP1). First, in the threshold value initialization process SP10, the microcomputer 4 reads out the pixel value at the reference position designated by the coordinate input device 6 from the frame memory 32, and sets this as the threshold value. Next, in the binarization processing SP11, the image data on the frame memory 32 is read out, binarized according to the set threshold value, and the result is stored in the binary image frame memory 41. Next, in the boundary line extraction process SP12, the boundary line of the binary image is extracted by the same method as the boundary line extraction process (FIG. 6) described in the first embodiment.
[0057]
Next, in the boundary line selection process SP13, the microcomputer 4 has the smallest distance d from the reference position designated by the coordinate input device 6 among the boundary lines obtained in the boundary line extraction process SP12. Select a line. This process is the same as the outline extraction process (FIG. 9) described in the first embodiment. Next, in the closed curve determination process SP14, the following two conditions are satisfied with respect to the boundary line selected in the boundary line selection process SP13, that is, (A) the start point and the end point are adjacent, or (B) the boundary line It is checked whether the length of is less than a preset threshold value TH6.
[0058]
If either of the conditions (A) or (B) is not satisfied, the microcomputer 4 proceeds to the threshold value correction processing SP15, sets the current threshold value to a smaller value, and sets 2 A processing loop from the value processing SP11 to the threshold value correction processing SP15 is executed. If both of the two conditions are satisfied, the microcomputer 4 determines that the extracted boundary line is a closed curve, and sends the coordinate series to the closed curve correction process SP16.
[0059]
In the closed curve correction process SP16, the micro computer 4 corrects the shape of the extracted closed curve in accordance with the inflection point. This processing is the same as the contour shape correction processing (FIG. 11) described in the first embodiment. However, here, in order to avoid the occurrence of unnecessary bulging, the interpolation of the indented portion is not the quadratic curve of equation (5), but the following equation:
[Equation 8]

It is also possible to carry out by the straight line shown in FIG. In order to generate an interpolated coordinate series by linear interpolation, the following formula is used instead of formula (6):
[Equation 9]

Neighboring points p = (i, j) satisfying
[0060]
Next, FIG. 27 shows the processing procedure of the contour line extraction processing (SP2 in FIG. 25). First, the microcomputer 4 calculates the center of gravity p of the closed curve whose shape has been corrected (hereinafter simply referred to as the closed curve) in the calculation process SP20 of the center of gravity of the closed curve. _g = (I _g , J _g ) Is calculated. Next, in the contour point detection process SP21, as shown in FIG. _k = (I _k , J _k ) For the center of gravity p _g Calculate the first derivative of the input image along the straight line from _k First, the following formula
[Expression 10]

The point p that constitutes the true contour line is the point that satisfies the condition of _k ′ = (I _k ′, J _k ′) Is detected.
[0061]
Where Xk _m Is the center of gravity p _g To point p on the closed curve _k Xk is a one-dimensional data sequence in which the values of the input image are arranged along a straight line toward ₀ Corresponds to the pixel value at the position of the center of gravity. dev _m Is the absolute value of the first derivative in this array. TH1, TH2, and TH3 are positive values set in advance. Equation (10) represents the point that the level itself increases and at the same time the absolute value of the first derivative is larger than the threshold value TH3 and the change starts from increasing to decreasing as shown in FIG. Point p _k First, a point that satisfies this condition is detected as a contour point.
[0062]
Finally, in the contour point interpolation processing SP22, the microcomputer 4 constructs a contour line by interpolating between the detected contour points. This interpolation processing is almost the same as the interpolation processing (FIG. 18) described in the first embodiment. However, in the fourth embodiment, as shown in FIG. 30, all the points between adjacent contour points are interpolated. In this case, since the section to be interpolated is short, linear interpolation as shown in the equation (8) can be performed. Note that, before the contour point interpolation processing, the position of the detected contour point can be smoothed as follows.
[0063]
For example, the distance l between the detected kth contour point and the center of gravity point _k Is calculated as shown in Equation (3) and smoothed in the k direction. For the smoothing process, for example, a one-dimensional averaging filter or a median filter is used. Smoothed l _k ′ Is used to determine the position of the new contour point (i _g + (I _k -I _g ) K '/ l _sk , J _g + (J _k − J _g ) K '/ l _sk By doing so, a smooth outline can be finally obtained. Where l _sk 'Represents the distance between the k-th point on the closed curve and the centroid point.
[0064]
According to the above configuration, the input heart moving image is sampled at an interval shorter than the heart motion repetition period, and the images in the same state of the heart are synchronously added to reduce noise in the time direction. Each of the images smoothed in time for each state of the heart is spatially smoothed, and a closed curve is generated for each image so that all points are inside the contour to be extracted. The outline of the heart is extracted by enlarging it outward, and the outline of the heart is extracted as a heart outline by modifying the shape of each outline. Since the phase can be automatically extracted for each phase, it is possible to observe temporal changes in the outline of the heart, thereby making it possible to perform a medical diagnosis quickly and accurately. Also, parameters necessary for diagnosis such as the speed of the contour line can be calculated from these contour lines.
[0065]
(5) Fifth embodiment
In this embodiment, a case where an endocardial contour line is extracted from an ultrasonic image of a uniaxial cross section of the heart will be described. In the uniaxial cross-sectional image of the heart, there are two regions A and B having high levels arranged concentrically as shown in FIG. 31, and the outer ring-shaped region A corresponds to the upper and lower portions of the region A. Part C tends to have a particularly high level. The endocardium to be extracted corresponds to the inner surface of the inner ring-shaped region B.
[0066]
The configuration (not shown) of the medical diagnostic image processing apparatus of this embodiment is almost the same as that of the image processing apparatus 40, but the binary image frame memory 41 for storing the binary image is not provided. In addition, the flow of processing by the microcomputer 4 in this embodiment is the same as that shown in FIG. 25, but the contents of each processing are different. In other words, in the closed curve generation process SP1, the micro computer 4 generates a circle as a closed curve that exists outside the contour line of the desired organ, and in the contour line extraction process SP2, it reduces toward the generated inner side. Thus, the approximate contour shape is extracted. The contour line shape correction processing SP3 is performed in exactly the same manner as in the fourth embodiment.
[0067]
FIG. 32 shows a processing procedure of closed curve generation processing (SP1 in FIG. 25). First, in the minimum point detection process SP10, the micro computer 4 reads a one-dimensional data string in the vertical direction from the image on the frame memory 32 subjected to the spatial smoothing process, and analyzes the level change. Here, the analysis of the level change in the vertical direction is shown in FIG. In FIG. 33, i represents the horizontal coordinate on the image of interest, and the curve in the figure represents the vertical pixel value (level) corresponding to the horizontal coordinate i.
[0068]
In this data example, first, j <j _C Coordinate j with the maximum level in the range _L And j> j _C Coordinate j with the maximum level in the range _U Is detected. Where j _C Is the reference position p designated by the coordinate input device 6 _C = (I _C , J _C ) Is the vertical coordinate. Then j _L And j _U The level change from the first to the next, the coordinates of the minimal point encountered first _LP And j _UP Is detected. The two detected vertical coordinates j _LP And j _UP Is the point p on the image _L = (I, j _LP ), Point p _U = (I, j _uP ). These points are considered to exist in a portion sandwiched between the two regions arranged concentrically as described above, and are located outside the endocardium to be extracted.
[0069]
Then the following formula
[Expression 11]

The same analysis is performed for all horizontal coordinates satisfying (2Δi + 1) × 2 minimum points. Δi is a preset positive constant.
[0070]
Next, in the circle approximation processing SP11, the microcomputer 4 approximates the detected (2Δi + 1) minimum points with one circle, and obtains the center coordinates (a, b) and the radius r. Here, the detected minimum points are arranged in an appropriate order, and the coordinates of the k-th point are expressed as (i _k , J _k ) Substituting these into the equation of the circle gives
[Expression 12]

It becomes. Here, by subtracting the equation of k = 0 from the equation corresponding to k> 0,
[Formula 13]

Can be obtained. The center coordinates (a, b) are obtained by solving this using the method of least squares. The radius r is calculated by substituting the obtained center coordinates into all the equations (12) and averaging r obtained from each.
[0071]
Next, in the circle generation processing SP12, the microcomputer 4 generates a coordinate series S constituting the circle using the calculated center coordinates (a, b) and the radius r. This is, for example, the following equation with coordinates (a, br) as the starting point:
[Expression 14]

Can be generated by performing a tracking process of neighboring points that minimize the value of. The tracking of the neighboring points can be performed in the same manner as the generation of the interpolation coordinate series in the first embodiment.
[0072]
Next, the processing procedure of the contour extraction process (SP2 in FIG. 25) is almost the same as the processing procedure shown in FIG. 27, but the center of gravity position of the closed curve has already been obtained as the center of the circle. The calculation process (SP20 in FIG. 27) is not necessary. In the contour point detection process (SP21 in FIG. 27), the micro computer 4 makes each point p forming a circle as shown in FIG. _k = (I _k, j _k ) To the center of gravity (the center of the circle) p _g Calculate the first derivative of the input image along the straight line going to (a, b).
[Expression 15]

A point p ′ that constitutes a true contour line _k = (I ' _k , J ′ _k ) To detect. Each symbol in equation (15) is the same as that described in equation (10). Xk _m Is the point p on the circle _k From its center p _g Is a one-dimensional data sequence in which image values are arranged along a straight line toward ₀ Corresponds to the pixel value of a point on the circle. Equation (15) represents that the level itself is decreasing, and at the same time, as shown in FIG. 35, the absolute value of the first derivative is larger than the threshold value TH3, and the change changes from increasing to decreasing. Thus, the first point that satisfies this condition as seen from the point on the circle is detected as the contour point.
[0073]
According to the above configuration, the input ultrasonic image of the single axis cross section of the heart is sampled at an interval shorter than the repetition period of the heart motion, and the images in the same state of the heart are synchronously added to the time direction. After reducing noise, each time-smoothed image for each heart condition is spatially smoothed to create a closed curve so that all points are outside the endocardial contour to be extracted. An outline of the endocardium is extracted by generating each image and shrinking each closed curve inward, and the shape of each outline is corrected and extracted as an endocardial outline. By doing so, it is possible to automatically extract the endocardial contour line for each time phase, so it is possible to observe the temporal change of the endocardial contour line, thereby enabling quick and accurate Medical diagnosis can be performed. Also, parameters necessary for diagnosis such as the speed of the contour line can be calculated from these contour lines.
[0074]
(6) Other embodiments
In the above-described embodiment, the case where a heart moving image is applied as the medical diagnosis moving image to the medical diagnosis image processing apparatus of the present invention has been described. However, the present invention is not limited to this, and other periodic images are used. The same effect can be obtained even when a moving image of a moving organ is applied to the medical diagnostic image processing apparatus of the present invention.
[0075]
In the first embodiment described above, the case where the contour line of the desired organ is extracted from the input still image by performing the boundary line extraction process (FIG. 5) has been described. However, the contour line of the desired organ may be extracted by the closed curve generation process (FIG. 25) as in the fourth and fifth embodiments.
[0076]
In this case, the input still image is spatially smoothed, and then a closed curve including an arbitrary point in the region surrounded by the outline of the desired organ to be extracted is generated, and the closed curve is smoothed. By deforming based on the level change of the image, an approximate outline of the desired organ is extracted, and the shape of the outline is corrected and extracted as the outline of the desired organ.
[0077]
In the second embodiment described above, the case where the contour line of the desired organ is extracted from the input moving image by performing the boundary line extraction process (FIG. 5) has been described. However, the contour line of the desired organ may be extracted by performing closed curve generation processing (FIG. 25) as in the fourth and fifth embodiments.
[0078]
In this case, the input moving image is sampled at every repetition period of the movement of the desired organ and synchronously added to reduce noise in the time direction, and then the time-smoothed image is spatially smoothed, A closed curve including an arbitrary point in the region surrounded by the contour line of the desired organ to be extracted is generated, and the closed curve is deformed based on the level change of the smoothed image to approximate the desired organ. The contour line of the desired organ is extracted by correcting the shape of the contour line.
[0079]
In the above-described embodiments, the case where an ultrasonic image is applied as a medical diagnostic image to the medical diagnostic image processing apparatus of the present invention has been described. However, the present invention is not limited thereto, and other medical diagnostic images may be used. You may apply to the image processing apparatus for medical diagnosis of this invention.
[0080]
【The invention's effect】
As described above, according to the present invention, a plurality of boundary lines are extracted after spatially smoothing an input still image for medical diagnosis, and a boundary line corresponding to the contour of a desired organ is extracted from the plurality of boundary lines. It is extracted as a contour line, and unnecessary points on the contour line are removed based on the inflection point obtained by detecting the inflection point where the inclination of the tangent line greatly changes at each point constituting the contour line. Thus, by correcting the shape of the contour line and extracting it as the contour line of the desired organ, it is possible to realize an image processing apparatus for medical diagnosis that can automatically extract the contour line of the desired organ.
[0081]
According to the present invention, the input medical diagnosis moving image is sampled at every repetition period of the desired organ movement, and synchronously added to reduce noise in the time direction. Smoothing and extracting a plurality of boundary lines, extracting a boundary line corresponding to the contour of the desired organ from the plurality of boundary lines as a contour line, and the inclination of the tangent line greatly changes at each point constituting the contour line Based on the inflection point obtained by detecting the inflection point, unnecessary points on the contour line are removed, the shape of the contour line is corrected, and the contour line of the desired organ is extracted. Accordingly, an image processing apparatus for medical diagnosis that can automatically extract the contour line of a desired organ can be realized.
[0082]
Further, according to the present invention, the input medical diagnostic moving image is sampled at intervals shorter than the desired organ movement repetition period, and the images in the same state of the desired organ are synchronously added to reduce noise in the time direction. The images smoothed in time for each desired organ state are spatially smoothed to extract a plurality of boundary lines for each image, and the boundary lines corresponding to the contours of the desired organ are contoured from the plurality of boundary lines. A line is extracted for each image, and unnecessary points on the contour line are removed based on the inflection point obtained by detecting the inflection point where the inclination of the tangent line greatly changes at each point constituting the contour line. Thus, by correcting the shape of each contour line and extracting it as the contour line of the desired organ, it is possible to automatically extract the contour line of the desired organ for each time phase. A processing device can be realized.
[0083]
Furthermore, according to the present invention, after the input still image for medical diagnosis is spatially smoothed, a closed curve including an arbitrary point within the region surrounded by the contour line of the desired organ to be extracted is generated. Then, the closed curve is deformed based on the level change of the smoothed image to extract an approximate contour line of the desired organ, and an inflection point where the inclination of the tangent line greatly changes at each point constituting the contour line is obtained. By removing unnecessary points on the contour line based on the inflection points obtained by detection, the shape of the contour line is corrected and extracted as the contour line of the desired organ. It is possible to realize a medical diagnostic image processing apparatus that can automatically extract the contour line of the image.
[0084]
Further, according to the present invention, the input medical diagnostic moving image is sampled at every repetition period of the desired organ movement and synchronously added to reduce the noise in the time direction. Smoothed to generate a closed curve that includes any point within the region surrounded by the contour of the desired organ to be extracted, and deforms this closed curve based on the level change of the smoothed image The outline of the desired organ is extracted, and an unnecessary point on the outline is detected based on the inflection point obtained by detecting the inflection point at which the inclination of the tangent line greatly changes at each point constituting the outline. An image processing apparatus for medical diagnosis that can automatically extract the contour of the desired organ by correcting the shape of the contour by removing the points and extracting the contour as the contour of the desired organ Can be realized.
[0085]
Furthermore, according to the present invention, the input moving image for medical diagnosis is sampled at intervals shorter than the circumference of the desired organ movement repeatedly, and images in the same state of the desired organ are synchronously added to reduce noise in the time direction. After that, each time-smoothed image for each desired organ state is spatially smoothed, and each closed curve including an arbitrary point in the region surrounded by the contour line of the desired organ to be extracted Each image is generated for each image, and each of these closed curves is deformed based on the level change of the smoothed image to extract an approximate contour line of the desired organ, and the tangent slope is large at each point constituting the contour line. Based on the inflection point obtained by detecting the changing inflection point, unnecessary points on the contour line are removed, the shape of the contour line is corrected, and the contour line of the desired organ is extracted. The desired device Capable of realizing a medical diagnostic image processing apparatus capable of automatically extracted for each time phase the contour of the.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of an image processing apparatus for medical diagnosis according to a first embodiment of the present invention.
FIG. 2 is a schematic diagram illustrating an image scanning procedure.
FIG. 3 is a chart showing an example of a filter coefficient of a spatial filter.
FIG. 4 is a chart showing an example of a filter coefficient of a spatial filter.
FIG. 5 is a flowchart for explaining the flow of processing of a micro computer.
FIG. 6 is a flowchart for explaining the processing procedure of boundary line extraction processing;
FIG. 7 is a schematic diagram for explaining a neighbor search for boundary line tracking;
FIG. 8 is a schematic diagram for explaining a boundary tracking direction;
FIG. 9 is a flowchart for explaining a processing procedure of contour line extraction processing;
FIG. 10 is a schematic diagram for explaining an outline selection process;
FIG. 11 is a flowchart for explaining a processing procedure of contour shape correction processing;
FIG. 12 is a schematic diagram illustrating an example of a contour shape when adjacent regions are merged.
FIG. 13 is a schematic diagram illustrating an example of a contour shape having a depression.
FIG. 14 is a schematic diagram for explaining detection of an inflection point;
FIG. 15 is a schematic diagram for explaining removal of a fusion region.
FIG. 16 is a schematic diagram for explaining removal of a fusion region when a contour line is closed;
FIG. 17 is a schematic diagram for explaining the removal of a dent.
FIG. 18 is a schematic diagram for explaining interpolation processing;
FIG. 19 is a schematic diagram for explaining detection of an interpolation point.
FIG. 20 is a block diagram illustrating a configuration of an image processing apparatus for medical diagnosis according to a second embodiment of the present invention.
FIG. 21 is a block diagram showing a configuration of an image sampler.
FIG. 22 is a waveform diagram for explaining a reference point in an electrocardiogram waveform;
FIG. 23 is a block diagram showing a configuration of an image processing apparatus for medical diagnosis according to a third embodiment of the present invention.
FIG. 24 is a block diagram showing the configuration of an image processing apparatus for medical diagnosis according to a fourth embodiment of the present invention.
FIG. 25 is a flow chart showing the flow of processing of the micro computer in the fourth embodiment.
FIG. 26 is a flowchart showing a processing procedure of closed curve generation processing in the fourth embodiment.
FIG. 27 is a flowchart showing a processing procedure of contour extraction processing in the fourth embodiment.
FIG. 28 is a schematic diagram for explaining an outline point search direction in the fourth embodiment;
FIG. 29 is a waveform diagram for explaining detection of contour points in the fourth embodiment.
FIG. 30 is a schematic diagram for explaining contour point interpolation in the fourth embodiment;
FIG. 31 is a schematic diagram showing a single-axis cross-sectional image of the heart.
FIG. 32 is a flowchart showing a processing procedure of closed curve generation processing in the fifth embodiment.
FIG. 33 is a waveform diagram for explanation of detection processing of local minimum points.
FIG. 34 is a schematic diagram for explaining an outline point search direction in the fifth embodiment;
FIG. 35 is a waveform diagram for explanation of detection of contour points in the fifth embodiment.
[Explanation of symbols]
1, 20, 30, 40 ... Medical diagnosis image processing device, 2 ... Analog digital conversion circuit (A / D converter), 3, 32 ... Frame memory, 4 ... Microcomputer, 5 ... Image display Device, 6 ... coordinate input device, 7, 10, 11 ... memory, 8 ... spatial filter, 9 ... binarizer, 21 ... delay element, 22, 31 ... image sampler, 23 ... Frame synchronization signal generator, 24 ... ECG monitor device. 41... Binary image frame memory, 42... Low-pass spatial filter.

Claims (6)

In an image processing apparatus for medical diagnosis that extracts a contour line of a desired organ from a still image for medical diagnosis,
Spatial filtering means for spatially smoothing the input still image for medical diagnosis;
Boundary line extracting means for extracting a plurality of boundary lines from the image obtained by the spatial filtering means;
Contour extraction means for extracting a boundary line corresponding to the contour of the desired organ as a contour line from the plurality of boundary lines extracted by the boundary line extraction means;
A contour shape correcting means for correcting the shape of the contour line extracted by the contour line extracting means ,
The contour shape correcting means is
An inflection point where the inclination of the tangent changes greatly at each point constituting the contour line extracted by the contour line extraction means is detected, and unnecessary points on the contour line are removed based on the inflection point. An image processing apparatus for medical diagnosis characterized by: In an image processing apparatus for medical diagnosis that extracts a contour line of a desired organ having a periodic movement from a moving image for medical diagnosis,
Image sampling means for inputting the medical diagnostic moving image for each repetition period of the movement of the desired organ;
A synchronous addition means for adding a plurality of images obtained by the image sampling means for each pixel to generate a time-smoothed image;
Spatial filtering means for spatially smoothing the image obtained by the synchronous addition means;
Boundary line extraction means for extracting a boundary line from the image obtained by the spatial filtering means;
Contour extraction means for extracting a boundary corresponding to the contour of the desired organ as a contour from the boundary extracted by the boundary extraction means;
A contour shape correcting means for correcting the shape of the contour line extracted by the contour line extracting means ,
The contour shape correcting means is
An inflection point where the inclination of the tangent changes greatly at each point constituting the contour line extracted by the contour line extraction means is detected, and unnecessary points on the contour line are removed based on the inflection point. An image processing apparatus for medical diagnosis characterized by: In an image processing apparatus for medical diagnosis that extracts a contour line of a desired organ having a periodic movement from a moving image for medical diagnosis,
Image sampling means for inputting the medical diagnostic moving image at an interval shorter than the repetition cycle of the desired organ movement;
Synchronous addition means for collecting images corresponding to the same state of the desired organ from the images obtained by the image sampling means and adding them for each pixel to generate a time-smoothed image for each state of the desired organ;
Spatial filtering means for spatially smoothing each image of the desired organ state obtained by the synchronous addition means;
Boundary line extraction means for extracting a boundary line from each of the images obtained by the spatial filtering means;
Contour extraction means for extracting a boundary corresponding to the contour of the desired organ as a contour from the boundary extracted by the boundary extraction means;
Contour shape correcting means for correcting the shape of each contour line extracted by the contour line extracting means ,
The contour shape correcting means is
An inflection point where the inclination of the tangent changes greatly at each point constituting the contour line extracted by the contour line extraction means is detected, and unnecessary points on the contour line are removed based on the inflection point. An image processing apparatus for medical diagnosis characterized by: In an image processing apparatus for medical diagnosis that extracts a contour line of a desired organ from a still image for medical diagnosis,
Spatial filtering means for spatially smoothing the input still image for medical diagnosis;
Closed curve generation means for generating a closed curve including an arbitrary point within the region surrounded by the outline of the desired organ to be extracted;
Closed curve deforming means for deforming the closed curve generated by the closed curve generating means based on the level change of the image smoothed by the spatial filtering means and extracting an approximate contour line of the desired organ;
A contour shape correcting means for correcting the shape of the contour line obtained by the closed curve deforming means ,
The contour shape correcting means is
An inflection point where the inclination of the tangent changes greatly at each point constituting the contour line extracted by the contour line extraction means is detected, and unnecessary points on the contour line are removed based on the inflection point. An image processing apparatus for medical diagnosis characterized by: In an image processing apparatus for medical diagnosis that extracts a contour line of a desired organ having a periodic movement from a moving image for medical diagnosis,
Image sampling means for inputting the medical diagnostic moving image for each repetition period of the movement of the desired organ;
A synchronous addition means for adding a plurality of images obtained by the image sampling means for each pixel to generate a time-smoothed image;
Spatial filtering means for spatially smoothing the image obtained by the synchronous addition means;
Closed curve generation means for generating a closed curve including an arbitrary point within the region surrounded by the outline of the desired organ to be extracted;
Closed curve deforming means for deforming the closed curve obtained by the closed curve generating means based on the level change of the image spatially smoothed by the spatial filtering means and extracting an approximate contour line of the desired organ; ,
A contour shape correcting means for correcting the shape of the contour line obtained by the closed curve deforming means ,
The contour shape correcting means is
An inflection point where the inclination of the tangent changes greatly at each point constituting the contour line extracted by the contour line extraction means is detected, and unnecessary points on the contour line are removed based on the inflection point. An image processing apparatus for medical diagnosis characterized by: In an image processing apparatus for medical diagnosis that extracts a contour line of a desired organ having a periodic movement from a moving image for medical diagnosis,
Image sampling means for inputting the medical diagnostic moving image at an interval shorter than a repetition cycle of the desired organ movement;
Synchronous addition means for collecting images corresponding to the same state of the desired organ from the images obtained by the image sampling means and adding them for each pixel to generate a time-smoothed image for each state of the desired organ;
Spatial filtering means for spatially smoothing each image of the desired organ state obtained by the synchronous addition means;
Closed curve generating means for generating, for each image, a closed curve including an arbitrary point in the region surrounded by the outline of the desired organ to be extracted;
Based on the level change of each image smoothed by the space filtering means, the closed curve obtained by the closed curve generating means is deformed, and an approximate contour line of the desired organ is extracted for each image. Closed curve deforming means,
A contour shape correcting means for correcting the shape of each contour line obtained by the closed curve deforming means ,
The contour shape correcting means is
An inflection point where the inclination of the tangent changes greatly at each point constituting the contour line extracted by the contour line extraction means is detected, and unnecessary points on the contour line are removed based on the inflection point. An image processing apparatus for medical diagnosis characterized by:

Images