Nothing Special   »   [go: up one dir, main page]
Next Article in Journal
Image Quality in Adaptive Optics Optical Coherence Tomography of Diabetic Patients
Next Article in Special Issue
Comparative Study of AI Modes in Ultrasound Diagnosis of Breast Lesions
Previous Article in Journal
Development and Validation of a Screening Equation for Obstructive Sleep Apnea
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Diagnostics
Volume 15
Issue 4
10.3390/diagnostics15040428
diagnostics-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Radiomic Analysis of Magnetic Resonance Imaging for Breast Cancer with TP53 Mutation: A Single Center Study

by
Jung Ho Park
1,
Lyo Min Kwon
2,
Hong Kyu Lee
3,
Taeryool Koo
4,
Yong Joon Suh
1,
Mi Jung Kwon
5,* and
Ho Young Kim
6
1
Division of Breast and Endocrine Surgery, Hallym University Sacred Heart Hospital, Anyang 14068, Republic of Korea
2
Department of Radiology, Hallym University Sacred Heart Hospital, Anyang 14068, Republic of Korea
3
Department of Thoracic and Cardiovascular Surgery, Hallym University Sacred Heart Hospital, Anyang 14068, Republic of Korea
4
Department of Radiation Oncology, Hallym University Sacred Heart Hospital, Anyang 14068, Republic of Korea
5
Department of Pathology, Hallym University Sacred Heart Hospital, Anyang 14068, Republic of Korea
6
Department of Internal Medicine, Hallym University Sacred Heart Hospital, Anyang 14068, Republic of Korea
*
Author to whom correspondence should be addressed.
Diagnostics 2025, 15(4), 428; https://doi.org/10.3390/diagnostics15040428
Submission received: 8 January 2025 / Revised: 29 January 2025 / Accepted: 7 February 2025 / Published: 10 February 2025
(This article belongs to the Special Issue Advances in Diagnosis and Prognosis of Breast Cancer)
Browse Figures
Figure 1
<p>Radiomics work-flow of the study. (<b>A</b>) Substraction images were loaded. (<b>B</b>) Tumor area was segmented semi-automatically. (<b>C</b>) Radiomic features were extracted using radiomics module. (<b>D</b>) Logistic regression using radiomics features. (<b>E</b>) <span class="html-italic">TP53</span> mutations were predicted using ROC curve analysis.</p> "> Figure 2
<p>Flow-diagram of the patient selection process.</p> "> Figure 3
<p>Mutational profiles of the tumors shown by Oncoprint.</p> "> Figure 4
<p>Three-dimensional views of the segmented tumors with <span class="html-italic">TP53</span> mutations.</p> ">
Versions Notes

Abstract

:
Background: Radiomics is a non-invasive and cost-effective method for predicting the biological characteristics of tumors. In this study, we explored the association between radiomic features derived from magnetic resonance imaging (MRI) and genetic alterations in patients with breast cancer. Methods: We reviewed electronic medical records of patients with breast cancer patients with available targeted next-generation sequencing data available between August 2018 and May 2021. Substraction imaging of T1-weighted sequences was utilized. The tumor area on MRI was segmented semi-automatically, based on a seeded region growing algorithm. Radiomic features were extracted using the open-source software 3D slicer (version 5.6.1) with PyRadiomics extension. The association between genetic alterations and radiomic features was examined. Results: In total, 166 patients were included in this study. Among the 50 panel genes analyzed, only TP53 mutations were significantly associated with radiomic features. Compared with TP53 wild-type tumors, TP53 mutations were associated with larger tumor size, advanced stage, negative hormonal receptor status, and HER2 positivity. Tumors with TP53 mutations exhibited higher values for Gray Level Non-Uniformity, Dependence Non-Uniformity, and Run Length Non-Uniformity, and lower values for Sphericity, Low Gray Level Emphasis, and Small Dependence Low Gray Level emphasis compared to TP53 wild-type tumors. Six radiomic features were selected to develop a composite radiomics score. Receiver operating characteristic curve analysis showed an area under the curve of 0.786 (95% confidence interval, 0.719–0.854; p < 0.001). Conclusions: TP53 mutations in breast cancer can be predicted using MRI-derived radiomic analysis. Further research is needed to assess whether radiomics can help guide treatment decisions in clinical practice.

1. Introduction

Breast cancer is the most commonly diagnosed cancer type among women globally, with an estimated 2.3 million cases in 2022 [1], and remains the leading cause of cancer-related deaths. The incidence of breast cancer is steadily increasing, particularly among young Asian women [2]. The development of novel anti-cancer drugs has improved breast cancer treatment outcomes over the decades; however, disparities in access to care have limited the availability and use of these advanced treatments.
Breast cancer is a heterogeneous disease, classified into various subtypes based on distinct biologic characteristics. The common classification includes four intrinsic subtypes: luminal A, B, HER2-enriched, and basal-like tumors [3]. Advances in genomic analyses have enhanced subtype classification, enabling more precise treatment approaches [4]. Interestingly, each breast cancer subtype exhibits distinct radiologic phenotypes [5]. For instance, luminal A breast cancers often display an echogenic halo on imaging, reflecting peritumoral edema due to inflammation. HER2-positive breast cancers frequently present microcalcifications on mammography, while triple-negative breast cancers often appear with regular shapes and microlobulated margins on ultrasonography.
The tumor-agnostic approach, which targets genomic alterations shared across various cancer types by anti-cancer agents, has gained acceptance in the field of oncology [6,7]. Newly developed anticancer agents target common biological aberrations shared by various cancer types. For example, trastuzumab was originally developed to target HER2-positive breast cancer. Anti-HER2 agents have also been used for cancers beyond breast cancer, such as gastric cancer [8], cholangiocarcinoma [9], and salivary gland cancers [10]. Beyond HER2 amplification, other molecular targets such as BRAF mutation, RET fusion, NTRK fusion, high tumor mutation burden, or deficient mismatch repair/high microsatellite instability have been identified as biomarkers for various anti-cancer agents [6]. Tumors with known mutational status can be matched to investigational treatments or enrolled in ongoing clinical trials [11].
Targeted next-generation sequencing (NGS) is currently the most widely used method for identifying genetic alterations in clinical practice [6]. Its high efficiency and cost-effectiveness have made it a preferred alternative to conventional sequencing methods. However, the routine application of targeted NGS in clinical practice remains controversial, and its accessibility continues to be limited [12]. In this context, radiomics has emerged as a non-invasive approach for predicting the biological characteristics of tumors through imaging modalities [13]. While conventional image interpretation by radiologists can be subject to observer variability, radiomics extracts digital information from medical images and provides additional insights beyond what is observable by the human eyes. This approach has been actively applied to lung cancers and brain tumors, demonstrating its ability to predict mutation status and survival outcomes [14,15].
Artificial intelligence has significantly improved medical image processing. Deep learning algorithms enable automated medical imaging analysis, demonstrating high performance. Radiomics is anticipated to play an important role in clinical practice, in combination with advancements in artificial intelligence. Radiomics provides standardized and quantifiable features that can be interpreted by human observers, making it widely accessible and feasible, even with relatively small datasets.
Despite these advancements, few studies have explored the associations between radiomics features and genetic alterations in breast cancer. In this study, we aimed to analyze the radiomics features derived from MRI in a breast cancer cohort with known genetic profiles identified by targeted NGS.

2. Materials and Methods

2.1. Patients

We retrospectively reviewed the electronic medical records of patients with breast cancer who underwent surgery at a single tertiary hospital. Patients with available targeted NGS data between August 2018 and May 2021 were included. Clinicopathological characteristics and NGS data were collected.

2.2. Next Generation Sequencing Procedures

Tumor specimens were processed as previously described [16]. Specimens were obtained following surgical resection of breast tumors. Following microscopic examination, a representative tumor area was micro-dissected by an institutional pathologist. Genetic alterations analyzed included single-nucleotide variants, frameshift mutations, small insertions, deletions, and copy number alterations. The institutional gene panel targeted 50 genes including AKT1, ALK, APC, ARID1A, ATRX, BRAF, BRCA1, BRCA2, CDH1, CDK4, CDK6, CDKN2A, CTNNB1, EGFR, HER2, ERBB3, ERBB4, ESR1, FBXW7, FGFR1, FGFR2, FGFR3, FOXA1, GATA3, H3F3A, IDH1, IDH2, KIT, KRAS, MAP2K1, MET, MLH1, MTOR, MYC, MYCN, NRAS, PDGFRA, PIK3CA, PTEN, RB1, RELA, RET, RHOA, RICTOR, ROS1, SMAD4, SMARCB1, SMO, STK11, and TP53. Genomic data were visualized using the web-based Oncoprinter application (https://www.cbioportal.org/oncoprinter (accessed on 15 November 2024)).

2.3. Magnetic Resonance Imaging Acquisition and Tumor Segmentation

Figure 1 illustrates the radiomics workflow. All patients underwent imaging using either of two 3T MRI scanners: Ingenia (Philips Medical Systems, Best, The Netherlands) or MAGNETOM Vida (Siemens Healthinners, Forchheim, Germany). Dynamic contrast-enhanced images were acquired before and after the intravenous injection of a gadolinium-based contrast agent. Post-contrast images were obtained at 60 s intervals following contrast administration. Subtraction images were generated by subtracting pre-contrast T1 images from the first post-contrast T1 image. Digital images were acquired as digital imaging and communication in medicine (DICOM) files from the institutional PACS server. The headers of the DICOM files were de-identified by assigning unique code numbers. The image files were loaded onto the investigator’s workstation. Tumor segmentation was performed semi-automatically, using the open-source software 3D Slicer (version 5.6.1; http://www.slicer.org (accessed on 27 December 2023)). A dedicated breast surgeon manually delineated the tumor and background areas. The tumor areas were refined using a region-growing algorithm. The tumor and background areas were manually trimmed. Only mass regions were included, and non-mass enhancement regions were excluded (Supplementary Figure S1).

2.4. Radiomic Feature Extraction

Radiomics features were extracted using the PyRadiomics extension in 3D slicer (version 3.0.1; Computational Imaging and Bioinformatics Lab, Harvard Medical School). A total of 107 radiomics features, comprising 14 shape-based, 18 first-order, 24 gray level co-occurrence matrix (GLCM), 14 gray level dependence matrix (GLDM), 16 gray level run length matrix (GLRLM), 16 gray level size zone matrix (GLSZM), and five neighboring gray tone difference matrix (NGTDM) features were analyzed (Supplementary Table S1).
We segmented the same tumor corresponding to that used for NGS analysis. In cases of bilateral breast cancer and multifocal or multicentric tumors, we selected only one tumor area with the most advanced stage and largest tumor, respectively.

2.5. Statistical Analysis

Categorical data were summarized as frequencies and percentages, while continuous data were presented as medians and ranges. The chi-squared test was used to compare categorical data. The Kolmogorov–Smirnov test was used to assess the normality of continuous data. Student’s t-tests were used to compare normally distributed continuous data while the Mann–Whitney U test was used to compare non-normally distributed continuous data. Statistical significance was set at p < 0.05.
For the comparison of the 112 radiomics features, Bonferroni correction was applied. Statistical significance was set at p < 5 × 10−4. Multivariable logistic regression with the stepwise backward elimination method was employed to select the radiomics features associated with genetic alterations. Receiver-operator characteristic (ROC) analysis was performed to evaluate the diagnostic performance of each radiomics feature. All statistical analyses were conducted using the Statistical Package for the Social Sciences Version 27.0 (IBM Corporation, Armonk, NY, USA) and R for Windows Version 4.3.1 (http://www.r-project.org (accessed on 16 June 2023)).

3. Results

3.1. Baseline Characteristics

Among the 259 patients with available NGS data, we included 166 in this study (Figure 2) while excluding 93 patients for the following reasons: artifacts interfering with the radiomic analysis (n = 44), unclear tumor boundaries (n = 19), difficult tumor segmentation (n = 18), and failure to acquire an MRI (n = 2). Supplementary Table S2 provides detailed descriptions of the exclusion criteria. The excluded cases were less likely to have TP53 mutations than those included in the study (Supplementary Table S3).
The median age of the patients was 53.5 years (range: 30–91 years). Among them, 63 (38.0%) were premenopausal, 98 (59.0%) postmenopausal, and 5 (3.0%) perimenopausal. A total of 118 (71.1%) underwent breast-conserving surgery while 48 (28.9%) underwent total mastectomy. TNM stage was distributed as follows: I (n = 44, 26.5%), II (n = 69, 41.6%), III (n = 46, 27.7%), and IV (n = 7, 4.2%). Regarding histopathological subtypes, 132 (79.5%) patients had invasive carcinomas of no special type, 10 (6.0%) had lobular carcinoma, 5 (3.0%) had micropapillary carcinoma, 4 (2.4%) had mucinous carcinoma, 4 (2.4%) had medullary carcinoma, and 11 (6.6%) had other subtypes. A total of 102 (61.4%) patients were hormone receptor-positive, and 43 (25.9%) were HER2-positive.
A total of 161 (97.0%) patients received at least one form of systemic treatment. Among the 150 (90.4%) patients who received chemotherapy, 133 (80.1%) received adjuvant chemotherapy, 10 (6.0%) received neoadjuvant chemotherapy, and 7 (4.2%) received palliative chemotherapy. Additionally, 102 (61.4%) patients received endocrine therapy and 38 (22.9%) were administered anti-HER2 therapy.
TP53 mutations were significantly associated with larger tumor size, advanced TNM stage, high body mass index, negative hormone receptor status, and HER2 positivity (Table 1). However, no significant association was observed between age and TP53 status.

BMI, Body Mass Index

Figure 3 summarizes the mutational profiles of the tumors. Among the 166 cases with available NGS data, 148 (89.2%) demonstrated significant genetic alterations, and 72 (43.4%) exhibited copy number alterations. TP53 mutations were identified in 73 cases (44.0%), PIK3CA mutations in 64 cases (38.6%), AKT1 mutations in 15 cases (9.0%), GATA3 mutations in 14 cases (8.4%), PTEN mutations in 12 cases (7.2%), BRCA2 mutations in 9 cases (5.4%), CDH1 mutations in 8 cases (4.8%), and BRCA1 mutations in 6 cases (3.6%). Regarding copy number alterations, ERBB2 amplification was identified in 26 cases (15.7%), MYC amplification in 9 (5.4%), and FGFR1 amplification in 6 cases (3.6%).

3.2. Association Between Genetic Alterations and Radiomic Features

Figure 4 presents the three-dimensional views of the segmented tumors with TP53 mutations. TP53 mutations were significantly associated with 24 radiomics features, including 14 shape-based features, 2 GLCM features, 3 GLDM features, 2 GLSZM features, 2 GLRLM features, and 1 NGTDM feature (Supplementary Table S4). Tumor size differences between TP53 wild-type and mutant tumors were more pronounced in three-dimensional measurements (43.5 vs. 29.6, p = 2.35 × 10−7). TP53-mutant tumors demonstrated significantly higher Gray Level Non-Uniformity values in GLDM (2008.4 vs. 709.4, p = 8.49 × 10−5), GLRLM (1612.9 vs. 503.8, p = 3.21 × 10−5), and GLSZM (408.3 vs. 104.5, p = 6.61 × 10−6) compared with TP53 wild-type tumors. Similarly, Dependence Non-Uniformity (7524.9 vs. 2060.9, p = 1.91 × 10−5) and Run Length Non-Uniformity (32,974.8 vs. 7803.8, p = 5.41 × 10−6) were significantly elevated in TP53-mutant tumors. Conversely, TP53-mutant tumors exhibited significantly lower values for Sphericity (0.50 vs. 0.56, p = 2.86 × 10−4), Low Gray Level Emphasis (0.010 vs. 0.024, p = 3.31 × 10−3), Small Dependence Low Gray Level Emphasis (0.0015 vs. 0.0033, p = 1.90 × 10−4), and Coarseness (0.002 vs. 0.004, p = 2.71 × 10−6) compared with TP53 wild-type tumors.
For PIK3CA mutations, tumors were associated with smaller Maximum 3D Diameter (33.3 vs. 37.2, p = 0.053), as well as lower Gray Level Non-Uniformity of GLDM (927.8 vs. 1502.1, p = 0.064), GLRLM (767.2 vs. 1132.2, p = 0.31), and GLSZM (176.0 vs. 277.0, p = 0.27). However, these associations were statistically not significant (Supplementary Table S5).
For AKT1 mutations, tumors were associated with smaller Total Energy (6.12 × 109 vs. 1.71 × 1010, p = 0.006), Mesh Volume (4614.1 vs. 12,152.3, p = 0.009), Voxel Volume (4704.4 vs. 12,307.7, p = 0.009), Gray Level Non-Uniformity of GLSZM (97.8 vs. 252.0, p = 0.011), Run Length Non-Uniformity (7297.2 vs. 20,022.9, p = 0.012), and Dependence Non-Uniformity (1884.6 vs. 4719.9, p = 0.013). However, these associations were not statistically significant (Supplementary Table S6).
GATA3 mutations were exclusive to luminal breast cancer subtypes and these tumors were associated with higher values of Small Dependence Low Gray Level Emphasis (0.005 vs. 0.002, p = 9 × 10−4), Sphericity (0.62 vs. 0.53, p = 0.002), Coarseness (0.006 vs. 0.003, p = 0.003), Short Run Low Gray Level Emphasis (0.03 vs. 0.01, p = 0.004), Low Gray Level Emphasis (0.03 vs. 0.02, p = 0.005), and Low Gray Level Zone Emphasis (0.06 vs. 0.03, p = 0.006) compared with GATA3 wild-type tumors. Conversely, tumors with GATA3 mutations exhibited lower Maximum 3D Diameter (24.9 vs. 36.7, p = 0.002), Dependence Non-Uniformity (1714.3 vs. 4717.0, p = 0.002), Gray Level Non-Uniformity in GLDM (477.4 vs. 1354.6, p = 0.014), GLRLM (391.5 vs. 1046.8, p = 0.016), and GLSZM (88.2 vs. 252.0, p = 0.002). These associations demonstrated marginal statistical significance (Supplementary Table S7).
No significant associations were observed between radiomics features and other genetic alterations, including mutations in PTEN, CDH1, BRCA1 and BRCA2, or amplifications of ERBB2, MYC, and FGFR1.

3.3. Prediction of Genetic Alterations by Radiomic Features

Receiver operating characteristic (ROC) curve analysis was conducted to assess the predictive power of radiomics features in identifying TP53 mutations. Among the individual radiomics features, the Maximum 3D Diameter yielded the highest predictive accuracy, with an area under the curve (AUC) of 0.734 (Table 2). Following Maximum 3D Diameter, features such as Run Length Non-Uniformity, Gray Level Non-Uniformity of GLDM, Dependence Non-Uniformity, and Gray Level Non-Uniformity of GLRLM also exhibited acceptable predictive performances [17].
We also conducted an ROC curve analysis to evaluate the predictive power of radiomics features for identifying PIK3CA and GATA3 mutations. Coarseness and Strength were associated with PIK3CA mutations, with neither feature demonstrating sufficient predictive ability (Table 2). For predicting GATA3 mutations, Small Dependence Low Gray Level Emphasis yielded the highest predictive power, followed by Sphericity, Coarseness, Low Gray Level Emphasis, and Low Gray Level Run Emphasis.
On multivariable logistic regression, six radiomics features—including Mean, Elongation, Skewness, Low Gray Level Emphasis, Small Dependence High Gray Level Emphasis, Gray Level Non-Uniformity of GLRLM—were selected to generate the composite radiomics score (Supplementary Table S8), which predicted TP53 mutations with an AUC of 0.786 (95% CI, 0.719–0.854; p < 0.001).

4. Discussion

Identifying the mutational status of the tumors is essential for understanding tumor biology and guiding the use of targeted agents in precision medicine. Although targeted NGS is a common approach for determining mutational status, it is costly and requires labor-intensive interpretation. In this study, we explored the potential of radiomics as a non-invasive alternative for identifying somatic mutations in breast cancer by analyzing the association between MRI-derived features and genetic alterations identified through targeted NGS.
Among the 50 genes analyzed, only TP53 mutations were significantly associated with MRI radiomics features. Specifically, TP53 mutations were significantly associated with larger tumor size and higher ‘non-uniformity’ radiomics features. We demonstrated that individual or combined radiomics features could predict TP53 mutations in breast cancer. Among the individual radiomics features, Maximum 3D Diameter was the best predictor of TP53 mutations, posing an AUC of 0.734. When the six radiomics features were combined in a logistic regression model, the prediction of TP53 mutations achieved an AUC of 0.786. Few studies have examined the relationship between TP53 mutations and radiomics features in breast cancer; however, our findings align with those of previous studies. For instance, one study examining the MRI radiomics features of 229 patients with breast cancer demonstrated an AUC of 0.78 for TP53 mutation identification using logistic regression [18]. Another study compared machine learning models using MRI radiomics features in 139 patients with breast cancer and reported an AUC of 0.74 for logistic regression [19].
TP53 is the most commonly mutated gene in breast cancer, as reported by The Cancer Genome Atlas Network [3]. Previous studies have linked TP53 mutations with aggressive characteristics of breast cancer, including larger tumor size, high grade, lymph node positivity, and hormone receptor negativity [20]. TP53 mutations are considered poor prognostic markers and are associated with treatment resistance [21,22]. It has been suggested that the prognostic significance of TP53 mutations differs according to breast cancer subtypes [23]. They were associated with poor survival in estrogen receptor-positive breast cancers and HER2-enriched subtypes but not with survival in basal-like subtypes. Routine testing for TP53 is not performed in clinical practice because no drug directly targets TP53 mutations [24]. However, preclinical studies have shown the potential of targeting TP53 mutations through the reactivation of mutant p53 proteins [25].
The PI3K-Akt-mTOR pathway is the most frequently altered signaling pathway in human cancer [26]. Targeting this pathway is an active area of investigation in breast cancer. In our study, PIK3CA and AKT1 mutations were associated with smaller tumor size and more uniform texture, although these associations did not meet the strict criteria for significance after Bonferroni correction.
PIK3CA is the second most commonly mutated gene in breast cancer [3]. PIK3CA mutations are established oncogenic drivers, with their clinical significance varying across breast cancer subtypes [27]. These mutations are prevalent in luminal subtypes, and are often associated with a favorable prognosis. In contrast, they account for less than 10% of triple-negative breast cancer, with the highest prevalence in the luminal androgen receptor subtype [28]. Alpelisib is the first PI3K inhibitor approved for the treatment of hormone receptor-positive, HER2-negative breast cancer [29]. AKT1, another key gene in the PI3K-Akt-mTOR pathway, functions as a downstream target of PIK3CA. Capivasertib, an AKT inhibitor, has demonstrated a significant survival benefit against hormone receptor-positive breast cancer [30].
GATA3 mutations were associated with smaller tumor size and uniform textural features, reflecting the radiologic characteristics of luminal breast cancer. Multiple radiomics features, such as Small Dependence Low Gray Level Emphasis, Sphericity, Coarseness, Low Gray Level Emphasis, and Low Gray Level Run Emphasis had the potential to predict GATA3 mutations. GATA3 mutations are frequently identified in small luminal breast cancers with homogeneous enhancing patterns, indicative of a favorable prognosis. They are almost exclusively identified in luminal breast cancers, whereas TP53 mutations are more commonly identified in hormone receptor-negative breast cancers. This finding is consistent with a previous study that reported mutual exclusivity between GATA3 mutations and TP53 mutations [3].
Mutations in BRCA1 and BRCA2 are the primary cause of hereditary breast and ovarian cancer syndrome. Patients with BRCA mutations can be candidates for genetic counseling, risk-reducing mastectomy, risk-reducing salpingo-oophrectomy, and treatment with poly (ADP-ribose) polymerase inhibitors [31]. On imaging, tumors with BRCA1 mutations are more likely to appear as round and well-circumscribed lesions on imaging [32]. However, in our study, radiomic analysis alone could not differentiate BRCA mutational status, reinforcing that genetic sequencing remains the gold standard for identifying BRCA mutations. Nevertheless, a recent study suggested that radiomic analysis of contralateral breast tissue may aid in identifying BRCA1 mutated tumors [33].
Radiomic analysis is a valuable tool for quantifying intratumoral heterogeneity, which has been linked to treatment resistance and poor prognosis [34]. In our study, TP53 mutations were associated with higher value of ‘non-uniformity’ radiomics features, underscoring their potential as markers of intratumoral heterogeneity. Entropy measured by radiomic analysis has been proposed in previous studies as a marker of intratumoral heterogeneity [35], although this finding was not confirmed in our study. Nevertheless, our findings remain consistent with those of previous studies, demonstrating that intratumoral heterogeneity assessed via MRI can serve as a prognostic marker for breast cancer. Specifically, homogeneous enhancement patterns on MRI have been associated with a favorable prognosis of breast cancer [36]. Intratumoral heterogeneity quantified by radiomic analysis has shown predictive value for treatment response to neoadjuvant chemotherapy in breast cancer [37]. Further validation is needed to determine whether intratumoral heterogeneity measured at the radiologic level corresponds to that observed at the molecular level.
To the best of our knowledge, this is the first study to compare radiomic features of breast cancer with genetic alterations identified through targeted NGS. Our findings highlight the complementary role of radiomics in molecular diagnostics. A limitation of clinical NGS is its lack of spatial information [38]. Multi-regional sequencing and spatial transcriptomics address this limitation [39,40]; however, they are not yet available in routine clinical practice. Contrastingly, MRI provides valuable spatial information, capturing histopathological patterns such as peritumoral edema, fibrosis, and neovascularization [41]. In the era of precision medicine, the ability to non-invasively characterize tumors through imaging is increasingly relevant.
This study had several limitations. First, it was a single-center, retrospective cohort study with a limited sample size, which may have hindered the detection of possible correlations between radiomics features and genetic alterations. Additionally, our study population consisted exclusively of South Koreans, restricting the generalizability of our conclusions to other ethnic groups. Second, a significant proportion of patients were excluded because of the inapplicability of radiomic analysis, underscoring the need for consistent selection criteria in future studies utilizing radiomic analysis for breast cancer. Third, tumor segmentation was performed semi-automatically, introducing interobserver variability [42]. The use of emerging artificial intelligence-based segmentation models that provide fully automated segmentation may help address this limitation.

5. Conclusions

Our study demonstrated that TP53 mutations in breast cancer can be predicted using MRI-derived radiomic analysis. This approach provides a new strategy for precision medicine in breast cancer. Further research is needed to assess whether radiomics can be utilized to guide treatment decisions in clinical practice.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/diagnostics15040428/s1, Figure S1: Detailed description of the semi-automatic segmentation process; Table S1: Radiomics features used in the study; Table S2: Detailed descriptions of the reasons for case exclusion; Table S3: Comparisons between included and excluded patients; Table S4: Radiomics features associated with TP53 mutations; Table S5: Radiomics features associated with PIK3CA mutations; Table S6: Radiomics features associated with AKT1 mutations; Table S7: Radiomics features associated with GATA3 mutations; Table S8: Multivariable logistic regression using six radiomics features.

Author Contributions

Software, L.M.K.; validation, Y.J.S.; investigation, H.K.L.; resources, M.J.K.; writing—original draft preparation, J.H.P.; writing—review and editing, T.K.; supervision, H.Y.K.; funding acquisition, M.J.K. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

The study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Review Board of Hallym University Sacred Heart Hospital (IRB number HALLYM 2023-08-004 and date of approval is 6 August 2024).

Informed Consent Statement

Patient consent was waived due to the retrospective nature of the study design.

Data Availability Statement

The raw data supporting the conclusions of this article will be made available by the authors on request.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:
AUCArea Under the Curve
DICOMDigital Imaging and Communication in Medicine
GLCMGray Level Co-occurrence Matrix
GLDMGray Level Dependence Matrix
GLRLMGray Level Run Length Matrix
GLSZMGray Level Size Zone Matrix
HER2Human epidermal growth receptor 2
LASSOLeast Absolute Shrinkage and Selection Operator
MRIMagnetic Resonance Imaging
NGSNext Generation Sequencing
NGTDMNeighboring Gray Tone Difference Matrix
PACSPicture Archiving and Communication System
ROCReceiver Operating Characteristic

References

  1. Filho, A.M.; Laversanne, M.; Ferlay, J.; Colombet, M.; Pineros, M.; Znaor, A.; Parkin, D.M.; Soerjomataram, I.; Bray, F. The GLOBOCAN 2022 cancer estimates: Data sources, methods, and a snappshot of the cancer burden worldwide. Int. J. Cancer 2024, 156, 1336–1346. [Google Scholar] [CrossRef] [PubMed]
  2. Giaquinto, A.N.; Sung, H.; Newman, L.A.; Freedman, R.A.; Smith, R.A.; Star, J.; Jemal, A.; Siegel, R.L. Breast cancer statistics 2024. CA Cancer J. Clin. 2024, 74, 477–495. [Google Scholar] [CrossRef] [PubMed]
  3. Cancer Genome Atlas Network. Comprehensive molecular portraits of human breast tumours. Nature 2012, 490, 61–70. [Google Scholar] [CrossRef]
  4. Wallden, B.; Storhoff, J.; Nielsen, T.; Dowidar, N.; Schaper, C.; Ferree, S.; Liu, S.; Leung, S.; Geiss, G.; Snider, J.; et al. Development and verification of the PAM50-based Prosigna breast cancer gene signature assay. BMC Med. Genom. 2015, 8, 54. [Google Scholar] [CrossRef]
  5. Dobruch-Sobczak, K.; Gumowska, M.; Mączewska, J.; Kolasińska-Ćwikła, A.; Guzik, P. Immunohistochemical subtypes of the breast cancer in the ultrasound and clinical aspect–literature review. J. Ultrason. 2022, 22, 93–99. [Google Scholar] [CrossRef]
  6. Gouda, M.A.; Nelson, B.E.; Buschhorn, L.; Wahida, A.; Subbiah, V. Tumor-Agnostic Precision Medicine from the AACR GENIE Database: Clinical Implications. Clin. Cancer Res. 2023, 29, 2753–2760. [Google Scholar] [CrossRef]
  7. Subbiah, V.; Gouda, M.A.; Ryll, B.; Burris, H.A.; Kurzrock, R. The evolving landscape of tissue-agnostic therapies in precision oncology. CA Cancer J. Clin. 2024, 74, 433–452. [Google Scholar] [CrossRef]
  8. Zhu, Y.X.; Zhu, X.D.; Wei, X.W.; Tang, C.J.; Zhang, W.W. HER2-targeted therapies in gastric cancer. Biochim. Biophys. Acta (BBA) Rev. Cancer 2021, 1876, 188549. [Google Scholar] [CrossRef]
  9. ten Haaft, B.H.E.A.; Pedregal, M.; Prato, J.; Klümpen, H.J.; Moreno, V.; Lamarca, A. Revolutionizing anti-HER2 therapies for extrahepatic cholangiocarcinoma and gallbladder cancer: Current advancements and future perspectives. Eur. J. Cancer 2024, 199, 113564. [Google Scholar] [CrossRef]
  10. Filippini, D.M.; Pagani, R.; Tober, N.; Lorini, L.; Riefolo, M.; Molinari, G.; Burato, A.; Alfieri, S.; Bossi, P.; Presutti, L. HER2-targeted therapies for salivary gland cancers. Oral. Oncol. 2024, 148, 106612. [Google Scholar] [CrossRef]
  11. Kato, S.; Kim, K.H.; Lim, H.J.; Boichard, A.; Nikanjam, M.; Weihe, E.; Kuo, D.J.; Eskander, R.N.; Goodman, A.; Galanina, N.; et al. Real-world data from a molecular tumor board demonstrates improved outcomes with a precision N-of-One strategy. Nat. Commun. 2020, 11, 4965. [Google Scholar] [CrossRef] [PubMed]
  12. Lee, Y.J.; Hwang, Y.S.; Kim, J.; Ahn, S.H.; Son, B.H.; Kim, H.J.; Ko, B.S.; Kim, J.; Chung, I.Y.; Lee, J.W.; et al. A nomogram for predicting probability of low risk of MammaPrint results in women with clinically high-risk breast cancer. Sci. Rep. 2021, 11, 23509. [Google Scholar] [CrossRef] [PubMed]
  13. Bodalal, Z.; Trebeschi, S.; Nguyen-Kim, T.D.L.; Schats, W.; Beets-Tan, R. Radiogenomics: Bridging imaging and genomics. Abdom. Radiol. 2019, 44, 1960–1984. [Google Scholar] [CrossRef]
  14. Avanzo, M.; Stancanello, J.; Pirrone, G.; Sartor, G. Radiomics and deep learning in lung cancer. Strahlenther. Onkol. 2020, 196, 879–887. [Google Scholar] [CrossRef]
  15. Beig, N.; Bera, K.; Tiwari, P. Introduction to radiomics and radiogenomics in neuro-oncology: Implications and challenges. Neurooncol. Adv. 2020, 2, iv3–iv14. [Google Scholar] [CrossRef]
  16. Park, J.H.; Kwon, M.J.; Seo, J.; Kim, H.Y.; Min, S.K.; Kim, L.S. Somatic Mutations of TP53 Identified by Targeted Next-Generation Sequencing Are Poor Prognostic Factors for Primary Operable Breast Cancer: A Single- Center Study. J. Breast Cancer 2022, 25, 379–386. [Google Scholar] [CrossRef]
  17. Mandrekar, J.N. Receiver operating characteristic curve in diagnostic test assessment. J. Thorac. Oncol. 2010, 5, 1315–1316. [Google Scholar] [CrossRef]
  18. Moon, W.K.; Chen, H.H.; Shin, S.U.; Han, W.; Chang, R.F. Evaluation of TP53/PIK3CA mutations using texture and morphology analysis on breast MRI. Magn. Reson. Imaging 2019, 63, 60–69. [Google Scholar] [CrossRef]
  19. Sun, K.; Zhu, H.; Chai, W.M.; Yan, F.H. TP53 Mutation Estimation Based on MRI Radiomics Analysis for Breast Cancer. J. Magn. Reson. Imaging 2023, 57, 1095–1103. [Google Scholar] [CrossRef]
  20. Olivier, M.; Langerod, A.; Carrieri, P.; Bergh, J.; Klaar, S.; Eyfjord, J.; Theillet, C.; Rodriguez, C.; Lidereau, R.; Bieche, I.; et al. The clinical value of somatic TP53 gene mutations in 1,794 patients with breast cancer. Clin. Cancer Res. 2006, 12, 1157–1167. [Google Scholar] [CrossRef]
  21. Langerod, A.; Zhao, H.; Borgan, O.; Nesland, J.M.; Bukholm, I.R.; Ikdahl, T.; Karesen, R.; Borresen-Dale, A.L.; Jeffrey, S.S. TP53 mutation status and gene expression profiles are powerful prognostic markers of breast cancer. Breast Cancer Res. 2007, 9, R30. [Google Scholar] [CrossRef] [PubMed]
  22. Coradini, D.; Biganzoli, E.; Ardoino, I.; Ambrogi, F.; Boracchi, P.; Demicheli, R.; Daidone, M.G.; Moliterni, A. p53 status identifies triple-negative breast cancer patients who do not respond to adjuvant chemotherapy. Breast 2015, 24, 294–297. [Google Scholar] [CrossRef] [PubMed]
  23. Silwal-Pandit, L.; Vollan, H.K.M.; Chin, S.F.; Rueda, O.M.; McKinney, S.; Osako, T.; Quigley, D.A.; Kristensen, V.N.; Aparicio, S.; Borresen-Dale, A.L.; et al. TP53 Mutation Spectrum in Breast Cancer Is Subtype Specific and Has Distinct Prognostic Relevance. Clin. Cancer Res. 2014, 20, 3569–3580. [Google Scholar] [CrossRef] [PubMed]
  24. Andre, F.; Ismaila, N.; Allison, K.H.; Barlow, W.E.; Collyar, D.E.; Damodaran, S.; Henry, N.L.; Jhaveri, K.; Kalinsky, K.; Kuderer, N.M.; et al. Biomarkers for Adjuvant Endocrine and Chemotherapy in Early-Stage Breast Cancer: ASCO Guideline Update. J. Clin. Oncol. 2022, 40, 1816–1837. [Google Scholar] [CrossRef]
  25. Tuval, A.; Strandgren, C.; Heldin, A.; Palomar-Siles, M.; Wiman, K.G. Pharmacological reactivation of p53 in the era of precision anticancer medicine. Nat. Rev. Clin. Oncol. 2024, 21, 106–120. [Google Scholar] [CrossRef]
  26. Glaviano, A.; Foo, A.S.C.; Lam, H.Y.; Yap, K.C.H.; Jacot, W.; Jones, R.H.; Eng, H.; Nair, M.G.; Makvandi, P.; Geoerger, B.; et al. PI3K/AKT/mTOR signaling transduction pathway and targeted therapies in cancer. Mol. Cancer 2023, 22, 138. [Google Scholar] [CrossRef]
  27. Reinhardt, K.; Stückrath, K.; Hartung, C.; Kaufhold, S.; Uleer, C.; Hanf, V.; Lantzsch, T.; Peschel, S.; John, J.; Pöhler, M.; et al. PIK3CA-mutations in breast cancer. Breast Cancer Res. Treat. 2022, 196, 483–493. [Google Scholar] [CrossRef]
  28. Elfgen, C.; Reeve, K.; Moskovszky, L.; Güth, U.; Bjelic-Radisic, V.; Fleisch, M.; Tausch, C.; Varga, Z. Prognostic impact of PIK3CA protein expression in triple negative breast cancer and its subtypes. J. Cancer Res. Clin. 2019, 145, 2051–2059. [Google Scholar] [CrossRef]
  29. Andre, F.; Ciruelos, E.M.; Juric, D.; Loibl, S.; Campone, M.; Mayer, I.A.; Rubovszky, G.; Yamashita, T.; Kaufman, B.; Lu, Y.S.; et al. Alpelisib plus fulvestrant for PIK3CA-mutated, hormone receptor-positive, human epidermal growth factor receptor-2-negative advanced breast cancer: Final overall survival results from SOLAR-1. Ann. Oncol. 2021, 32, 208–217. [Google Scholar] [CrossRef]
  30. Turner, N.C.; Oliveira, M.; Howell, S.J.; Dalenc, F.; Cortes, J.; Gomez Moreno, H.L.; Hu, X.; Jhaveri, K.; Krivorotko, P.; Loibl, S.; et al. Capivasertib in Hormone Receptor-Positive Advanced Breast Cancer. N. Engl. J. Med. 2023, 388, 2058–2070. [Google Scholar] [CrossRef]
  31. Arun, B.; Couch, F.J.; Abraham, J.; Tung, N.; Fasching, P.A. BRCA-mutated breast cancer: The unmet need, challenges and therapeutic benefits of genetic testing. Br. J. Cancer 2024, 131, 1400–1414. [Google Scholar] [CrossRef] [PubMed]
  32. Ha, S.M.; Chae, E.Y.; Cha, J.H.; Kim, H.H.; Shin, H.J.; Choi, W.J. Association of BRCA Mutation Types, Imaging Features, and Pathologic Findings in Patients With Breast Cancer With BRCA1 and BRCA2 Mutations. AJR Am. J. Roentgenol. 2017, 209, 920–928. [Google Scholar] [CrossRef] [PubMed]
  33. Pecchi, A.; Bozzola, C.; Beretta, C.; Besutti, G.; Toss, A.; Cortesi, L.; Balboni, E.; Nocetti, L.; Ligabue, G.; Torricelli, P. DCE-MRI Radiomic analysis in triple negative ductal invasive breast cancer. Comparison between BRCA and not BRCA mutated patients: Preliminary results. Magn. Reason. Imaging 2024, 113, 110214. [Google Scholar] [CrossRef] [PubMed]
  34. Eloyan, A.; Yue, M.S.; Khachatryan, D. Tumor heterogeneity estimation for radiomics in cancer. Stat. Med. 2020, 39, 4704–4723. [Google Scholar] [CrossRef] [PubMed]
  35. Lee, J.Y.; Lee, K.S.; Seo, B.K.; Cho, K.R.; Woo, O.H.; Song, S.E.; Kim, E.K.; Lee, H.Y.; Kim, J.S.; Cha, J. Radiomic machine learning for predicting prognostic biomarkers and molecular subtypes of breast cancer using tumor heterogeneity and angiogenesis properties on MRI. Eur. Radiol. 2022, 32, 650–660. [Google Scholar] [CrossRef]
  36. Wu, J.; Cui, Y.; Sun, X.L.; Cao, G.H.; Li, B.L.; Ikeda, D.M.; Kurian, A.W.; Li, R.J. Unsupervised Clustering of Quantitative Image Phenotypes Reveals Breast Cancer Subtypes with Distinct Prognoses and Molecular Pathways. Clin. Cancer Res. 2017, 23, 3334–3342. [Google Scholar] [CrossRef]
  37. Shi, Z.; Huang, X.; Cheng, Z.; Xu, Z.; Lin, H.; Liu, C.; Chen, X.; Liu, C.; Liang, C.; Lu, C.; et al. MRI-based Quantification of Intratumoral Heterogeneity for Predicting Treatment Response to Neoadjuvant Chemotherapy in Breast Cancer. Radiology 2023, 308, e222830. [Google Scholar] [CrossRef]
  38. Ke, R.; Mignardi, M.; Hauling, T.; Nilsson, M. Fourth Generation of Next-Generation Sequencing Technologies: Promise and Consequences. Hum. Mutat. 2016, 37, 1363–1367. [Google Scholar] [CrossRef]
  39. Li, Q.Y.; Zhang, X.Y.; Ke, R.Q. Spatial Transcriptomics for Tumor Heterogeneity Analysis. Front. Genet. 2022, 13, 906158. [Google Scholar] [CrossRef]
  40. Yates, L.R.; Gerstung, M.; Knappskog, S.; Desmedt, C.; Gundem, G.; Van Loo, P.; Aas, T.; Alexandrov, L.B.; Larsimont, D.; Davies, H.; et al. Subclonal diversification of primary breast cancer revealed by multiregion sequencing. Nat. Med. 2015, 21, 751–759. [Google Scholar] [CrossRef]
  41. Mayrhofer, R.M.; Ng, H.P.; Putti, T.C.; Kuchel, P.W. Magnetic resonance in the detection of breast cancers of different histological types. Magn. Reson. Insights 2013, 6, 33–49. [Google Scholar] [CrossRef]
  42. Granzier, R.W.Y.; Verbakel, N.M.H.; Ibrahim, A.; van Timmeren, J.E.; van Nijnatten, T.J.A.; Leijenaar, R.T.H.; Lobbes, M.B.I.; Smidt, M.L.; Woodruff, H.C. MRI-based radiomics in breast cancer: Feature robustness with respect to inter-observer segmentation variability. Sci. Rep. 2020, 10, 14163. [Google Scholar] [CrossRef]
Diagnostics 15 00428 g001
Figure 1. Radiomics work-flow of the study. (A) Substraction images were loaded. (B) Tumor area was segmented semi-automatically. (C) Radiomic features were extracted using radiomics module. (D) Logistic regression using radiomics features. (E) TP53 mutations were predicted using ROC curve analysis.
Figure 1. Radiomics work-flow of the study. (A) Substraction images were loaded. (B) Tumor area was segmented semi-automatically. (C) Radiomic features were extracted using radiomics module. (D) Logistic regression using radiomics features. (E) TP53 mutations were predicted using ROC curve analysis.
Diagnostics 15 00428 g001
Diagnostics 15 00428 g002
Figure 2. Flow-diagram of the patient selection process.
Figure 2. Flow-diagram of the patient selection process.
Diagnostics 15 00428 g002
Diagnostics 15 00428 g003
Figure 3. Mutational profiles of the tumors shown by Oncoprint.
Figure 3. Mutational profiles of the tumors shown by Oncoprint.
Diagnostics 15 00428 g003
Diagnostics 15 00428 g004
Figure 4. Three-dimensional views of the segmented tumors with TP53 mutations.
Figure 4. Three-dimensional views of the segmented tumors with TP53 mutations.
Diagnostics 15 00428 g004
Table 1. Comparisons between TP53 mutant versus wild-type tumors.
Table 1. Comparisons between TP53 mutant versus wild-type tumors.
TP53 Wild-Type (n = 93)TP53 Mutant (n = 73)p
Median age (range), years54 (36–83)53 (30–91)0.854
Menopausal status 0.751
  Premenopausal40 (43.0%)29 (39.7%)
  Postmenopausal53 (57.0%)44 (60.3%)
BMI, kg/m223.7 (17.8–35.8)25.3 (18.0–44.8)0.035
Median tumor size (range), cm2.0 (0.4–8.5)2.5 (0.5–9.5)0.014
TNM Stage <0.001
  I33 (35.5%)11 (15.1%)
  II42 (45.2%)27 (37.0%)
  III15 (16.1%)31 (42.5%)
  IV3 (3.2%)4 (5.5%)
Hormone receptor <0.001
  Negative17 (18.3%)45 (61.6%)
  Positive76 (81.7%)28 (38.4%)
HER2 status 0.001
  Negative78 (83.9%)45 (61.6%)
  Positive15 (16.1%)28 (38.4%)
Table 2. Predictive powers of radiomics features for identifying mutations.
Table 2. Predictive powers of radiomics features for identifying mutations.
Radiomic FeaturesAUC95% CIp
TP53
  Composite radiomics score0.7860.719–0.854<0.001
  Maximum 3D Diameter0.7340.659–0.810<0.001
  Run Length Non-Uniformity0.7060.627–0.785<0.001
  Gray Level Non-Uniformity of GLDM0.7040.625–0.783<0.001
  Dependence Non-Uniformity0.6940.614–0.773<0.001
  Gray Level Non-Uniformity of GLRLM0.6880.607–0.770<0.001
PIK3CA
  Coarseness0.5840.493–0.6760.072
  Strength0.5830.494–0.6710.067
GATA3
  Small Dependence Low Gray Level Emphasis0.7680.621–0.915<0.001
  Sphericity0.7420.604–0.881<0.001
  Coarseness0.7370.592–0.8820.001
  Low Gray Level Emphasis0.7250.580–0.8700.002
  Low Gray Level Run Emphasis0.7240.580–0.8690.002
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Park, J.H.; Kwon, L.M.; Lee, H.K.; Koo, T.; Suh, Y.J.; Kwon, M.J.; Kim, H.Y. Radiomic Analysis of Magnetic Resonance Imaging for Breast Cancer with TP53 Mutation: A Single Center Study. Diagnostics 2025, 15, 428. https://doi.org/10.3390/diagnostics15040428

AMA Style

Park JH, Kwon LM, Lee HK, Koo T, Suh YJ, Kwon MJ, Kim HY. Radiomic Analysis of Magnetic Resonance Imaging for Breast Cancer with TP53 Mutation: A Single Center Study. Diagnostics. 2025; 15(4):428. https://doi.org/10.3390/diagnostics15040428

Chicago/Turabian Style

Park, Jung Ho, Lyo Min Kwon, Hong Kyu Lee, Taeryool Koo, Yong Joon Suh, Mi Jung Kwon, and Ho Young Kim. 2025. "Radiomic Analysis of Magnetic Resonance Imaging for Breast Cancer with TP53 Mutation: A Single Center Study" Diagnostics 15, no. 4: 428. https://doi.org/10.3390/diagnostics15040428

APA Style

Park, J. H., Kwon, L. M., Lee, H. K., Koo, T., Suh, Y. J., Kwon, M. J., & Kim, H. Y. (2025). Radiomic Analysis of Magnetic Resonance Imaging for Breast Cancer with TP53 Mutation: A Single Center Study. Diagnostics, 15(4), 428. https://doi.org/10.3390/diagnostics15040428

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop