Cancer Detection by Machine Learning
Cancer Detection by Machine Learning
Cancer Detection by Machine Learning
net/publication/350835536
Article in International Journal of Computer Science and Information Security, · April 2021
CITATIONS READS
5 489
1 author:
SEE PROFILE
All content following this page was uploaded by Md Haris Uddin Sharif on 11 September 2021.
https://doi.org/10.5281/zenodo.4578330 67 https://sites.google.com/site/ijcsis/
ISSN 1947-5500
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 19, No. 2, February 2021
recurring Leukemia and Myeloid Leukemia in babies are grade 2 are malignant and cancerous. The pre-processing
uncommon. The malignancy in leukaemia cell growth and procedure is carried out to eliminate the sound and non-brain
progeny have been generalized to complex molecular tissue from image input for greater precision. There are
approaches [02]. This is known as Leukemia for hematogenic several primary steps for tumor identification. The techniques
cell development. Acute Leukemia induces more than 20 per BSE (brain surface extraction) are used to eliminate non-brain
cent of explosions in the bone marrow (S., 2020). organ. The simple non-local mean (FNLM), partial differential
The liver is a fundamental part of the human body that diffusion (PDDF) and Wiener filter are used for noise
fulfils basic functions, such as drug detoxification, the reduction and contrast depletion for better contrast. Fuzzy
development of blood proteins, and the filtering of the blood Cmean, k-mean and Otsu threshold approaches are the most
from waste components. In place of this, an illness in the common brain tumor segmentation techniques. In the same
human body is highly important, particularly if a cancer way, Net architecture is also one of the popular CNN
diagnosis has been made. Liver cancer is commonly known as architectures for brain cancer segmentation.
hepatic cancer which is the most serious and aggressive Hand-crafted features are collected after the segmentation
ailment. The most prevalent form of liver cancer, causing up process to turn the fractured images into math explanations.
to 80 per cent of death has been Hepatocellular Carcinoma More rigorous methods are currently employed to isolate
(HCC) (S., 2020).. GLOBOCAN estimates that liver cancer is features and then to identify them. The most popular
the world's 6th and 7th leading cause of deaths among both techniques for extraction include histogram orientation
men and women. gradient (HOG), Gabor wavelet transform (GWT), local
In my discussion, having a variety of cancers, I shall now binary patterns (LBP), and form-based functions. Also,
focus on two types of cancers to showcase how machine multiple filtering and reduction methods are used for
learning assists in detecting these diseases. maximizing function selection, such as a genetic algorithm
(GA) and the main component analysis (PCA). At present,
II. SCOPE AND OBJECTIVES CNN architecture is often known as an effective tool for the
diagnosis of brain tumours.
A. Scope and Objectives
A new field of early-detection research on cancer has been
opened up by different forms of cancer detection and
classification using the computer assistance that has shown the
potential to eliminate manual system impairments. The present
survey provides numerous sections on the cutting-edge
methods, analysis and comparisons for F-measurement,
sensitiveness, precision, precise, and accurate data sets for
brain tumours, breast cancer, prostate cancer, liver tumours Fig: 01 [10].
and Leukemia, and skin lesion identification. This diagram
provides a pictorial example of this analysis. To increase image transparency, multiple variations are
B. Benchmark Datasets created. The preprocessing methods, including noise,
complement optimal performance changes, decreasing,
The section discusses data sets commonly used to equalizing histograms, and enhancing tip. In the latest model's
experiment, analyze, and compare state-of-the-art
class, the weighting is used to deal with the topic of class
technologies' cancer tests and rankings. The section also
disparity. Details of validation are inspected in a learned
highlights their source, training sets, test sets, cancer detection
model from the same image set, and each experiment's results
success metrics, segmentation and classification.
are reported. ConvNet, as one score (accuracy) of 75% and
C. Brats 2015 Datasets 80% of the total fusion of 82.29% is provided by an LSTM-
This dataset consists of 274 sets, of 192 training cases (154 based network [08]. The defined model is validated by the
HGG and 38 LGG), 82 test cases and 82 trials cases from the MICCAI Challenge database, including multi-modal brain
Perelman School of the University of Medicine in tumor segmentation (BRATS) 2015, 2016 and 2017,
Pennsylvania (66 HGG and 16 LGG [08]. The training respectively, using the latest science imaging and computer-
pictures include high quality and low glioma with ground- assisted intervention [08]. Such technologies are developed by
level reality. The ground trueness is noted in five codes, for entropy to quickly and accurately detect and interact with
example, 1 for necrosis, 2 for edema, 3 for tumor-free fused vectors to classification units. The study results were
improvement, 4 for tumor enhancement, and 0 for everything 0.99 with 2015 BRATS, 1.00 with 2016 BRATS and 0.99
else. with 2017 BRATS with a coefficient of dice similarity (DSC)
[08]. They did not, though, use other classifiers or their fusion
D. Brain Tumor to check their technique's feasibility.
A brain tumor is an irregular four-degree cell array. Brain
tumors in grade 1 and 2 appear to develop slowly, and cancers
in grades 3 and 4 grow rapidly and difficult to treat. Tumors of
https://doi.org/10.5281/zenodo.4578330 68 https://sites.google.com/site/ijcsis/
ISSN 1947-5500
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 19, No. 2, February 2021
https://doi.org/10.5281/zenodo.4578330 69 https://sites.google.com/site/ijcsis/
ISSN 1947-5500
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 19, No. 2, February 2021
https://doi.org/10.5281/zenodo.4578330 70 https://sites.google.com/site/ijcsis/
ISSN 1947-5500
/,)*.ÿ^ÿ#_*ÿ,**ÿD¢&ÿA*ÿ*0*%&*Yÿ2ÿ&,*ÿ%C*&*Yÿ.2&*IP*ÿ#*3$*ÿ*Dÿ.2&*
"/+"+ÿ N$Y&*Yÿ7ÿU]ÿ898:]ÿ<¡U:ÿ-7 ÿ \543