A machine learning approach to radiogenomics of breast cancer: a study of 922 subjects and 529 DCE-MRI features

Development and validation of an interpretable machine learning model for diagnosing pathologic complete response in breast cancer

BACKGROUND

Pathologic complete response (pCR) following neoadjuvant chemotherapy (NACT) is a critical prognostic marker for patients with breast cancer, potentially allowing surgery omission. However, noninvasive and accurate pCR diagnosis remains a significant challenge due to the limitations of current imaging techniques, particularly in cases where tumors completely disappear post-NACT.

METHODS

We developed a novel framework incorporating Dimensional Accumulation for Layered Images (DALI) and an Attention-Box annotation tool to address the unique challenge of analyzing imaging data where target lesions are absent. These methods transform three-dimensional magnetic resonance imaging into two-dimensional representations and ensure consistent target tracking across time-points. Preprocessing techniques, including tissue-region normalization and subtraction imaging, were used to enhance model performance. Imaging features were extracted using radiomics and pretrained deep-learning models, and machine-learning algorithms were integrated into a stacked ensemble model. The approach was developed using the I-SPY 2 dataset and validated with an independent Tangshan People's Hospital cohort.

RESULTS

The stacked ensemble model achieved superior diagnostic performance, with an area under the receiver operating characteristic curve of 0.831 (95 % confidence interval, 0.769-0.887) on the test set, outperforming individual models. Tissue-region normalization and subtraction imaging significantly enhanced diagnostic accuracy. SHAP analysis identified variables that contributed to the model predictions, ensuring model interpretability.

CONCLUSION

This innovative framework addresses challenges of noninvasive pCR diagnosis. Integrating advanced preprocessing techniques improves feature quality and model performance, supporting clinicians in identifying patients who can safely omit surgery. This innovation reduces unnecessary treatments and improves quality of life for patients with breast cancer.

Cross-site Validation of AI Segmentation and Harmonization in Breast MRI

This work aims to perform a cross-site validation of automated segmentation for breast cancers in MRI and to compare the performance to radiologists. A three-dimensional (3D) U-Net was trained to segment cancers in dynamic contrast-enhanced axial MRIs using a large dataset from Site 1 (n = 15,266; 449 malignant and 14,817 benign). Performance was validated on site-specific test data from this and two additional sites, and common publicly available testing data. Four radiologists from each of the three clinical sites provided two-dimensional (2D) segmentations as ground truth. Segmentation performance did not differ between the network and radiologists on the test data from Sites 1 and 2 or the common public data (median Dice score Site 1, network 0.86 vs. radiologist 0.85, n = 114; Site 2, 0.91 vs. 0.91, n = 50; common: 0.93 vs. 0.90). For Site 3, an affine input layer was fine-tuned using segmentation labels, resulting in comparable performance between the network and radiologist (0.88 vs. 0.89, n = 42). Radiologist performance differed on the common test data, and the network numerically outperformed 11 of the 12 radiologists (median Dice: 0.85-0.94, n = 20). In conclusion, a deep network with a novel supervised harmonization technique matches radiologists' performance in MRI tumor segmentation across clinical sites. We make code and weights publicly available to promote reproducible AI in radiology.

Accurate phenotyping of luminal A breast cancer in magnetic resonance imaging: A new 3D CNN approach

Breast cancer (BC) remains a predominant and deadly cancer in women worldwide. By 2040, projections indicate that more than 3 million new cases of breast cancer will emerge annually, culminating in more than 1 million deaths worldwide. Early detection and accurate diagnosis of BC are critical factors that influence treatment success and patient outcomes. During the past three decades, several medical imaging modalities, such as X-ray Mammography (MG), Ultrasound (US), Computer Tomography (CT), Magnetic Resonance Imaging (MRI), and Digital Tomosynthesis (DT) have been explored to support radiologists/physicians in clinical decision-making workflows for the detection, diagnosis, treatment, and monitoring of BC. Magnetic resonance imaging (MRI) is an advanced imaging modality that provides detailed information on the structure and function of breast tissue. In particular, MRI may be crucial to discern the phenotype of BC, as each subtype has a different prognosis and requires different treatment strategies. This study aims to explore deep learning models for classifying/diagnosing BC phenotypes. As a main contribution, we propose a new 3D convolutional neural network (CNN) model based on quantitative medical imaging biomarkers (QIB) obtained from MRI data to diagnose the luminal A subtype (LA) of BC. LA is a subtype characterized by positive hormone receptor expression and negative HER2 expression. It uses a binary classification strategy to distinguish between pathological luminal A and non-luminal A lesions by analyzing 3D volumetric MRI images. The proposed method allows the extraction and analysis of spatial information, which is essential to accurately diagnose BC, especially for the LA subtype, taking into account their specific morphological characteristics. Our goal is to improve accuracy and efficacy in the diagnosis of the LA phenotype of BC and to contribute to the development of personalized treatment plans for patients. To develop and evaluate the performance of the proposed method, we used a benchmarking public domain MRI-based BC dataset (Duke-Breast-Cancer-MRI). To address the imbalance in the data set, we implemented a class weighting strategy during model training. In experimental settings, we achieved an AUC score of 0.9614 and a F1 score of 0.9328, outperforming state-of-the-art methods, including ResNet-152. These results demonstrate the potential of our work to significantly improve the diagnosis of the luminal A phenotype of breast cancer, paving the way for more accurate and personalized treatment strategies.

Personalized Breast Cancer Prognosis Using a Model Based on MRI and Clinicopathological Variables

This study aimed to develop and internally validate a prognostic prediction model based on MRI, pathological, and clinical findings to predict breast cancer recurrence and death. A retrospective study prediction model was developed using data from 922 breast cancer patients recruited in Duke University Hospital from January 2000 to March 2014. Cox and binary logistic regressions were implemented for hazard score and 2-, 3-, 5-, and 8-year survivals and recurrences. After assessing the collinearity of predictors, both univariable and multivariable analyses were performed. Qualitative and quantitative MRI variables were selected based on clinical expert opinion and literature review. Bootstrap and leave-one-out methods were used for internal validation. Calibration, shrinkage, time-dependent receiver operating characteristic (ROC) curve, and decision-curve analyses were also performed. Finally, a user-friendly calculator was built. Of included participants, 62 (6.72%) died with a mean patient-year follow-up of 8.89 years (CI = 8.74 to 9.04), while 90 (9.76%) experienced recurrence with mean patient-year follow-up of 8.20 years (CI = 7.92 to 8.48). The Akaike information criterion (AIC) value of survival and recurrence models were 752.9 and 1020.7, indicating a good balance between model complexity and fit. Validation model adjusted area under curve (AUC) in 8-, 5-, 3-, and 2-year survivals were 0.740 (CI = 0.711 to 0.768), 0.741 (CI = 0.712 to 0.770), 0.788 (CI = 0.761 to 0.816), and 0.783 (CI = 0.755 to 0.809), while in 8-, 5-, and 3-year recurrences were 0.678 (CI = 0.647 to 0.708), 0.696 (CI = 0.664 to 0.727), and 0.769 (CI = 0.740 to 0.798), respectively. Good calibration and shrinkage parameters were achieved. The internal validation and decision curve analyses highlighted the usefulness of the model across all probability levels. The combined MRI-pathological-clinical model has excellent performance in predicting overall survival and recurrence of breast cancer and may have a role to play in daily personalized breast cancer therapy.

Learning robust medical image segmentation from multi-source annotations

Collecting annotations from multiple independent sources could mitigate the impact of potential noises and biases from a single source, which is a common practice in medical image segmentation. However, learning segmentation networks from multi-source annotations remains a challenge due to the uncertainties brought by the variance of the annotations. In this paper, we proposed an Uncertainty-guided Multi-source Annotation Network (UMA-Net), which guided the training process by uncertainty estimation at both the pixel and the image levels. First, we developed an annotation uncertainty estimation module (AUEM) to estimate the pixel-wise uncertainty of each annotation, which then guided the network to learn from reliable pixels by a weighted segmentation loss. Second, a quality assessment module (QAM) was proposed to assess the image-level quality of the input samples based on the former estimated annotation uncertainties. Furthermore, instead of discarding the low-quality samples, we introduced an auxiliary predictor to learn from them and thus ensured the preservation of their representation knowledge in the backbone without directly accumulating errors within the primary predictor. Extensive experiments demonstrated the effectiveness and feasibility of our proposed UMA-Net on various datasets, including 2D chest X-ray segmentation dataset, 2D fundus image segmentation dataset, 3D breast DCE-MRI segmentation dataset, and the QUBIQ multi-task segmentation dataset. Code will be released at https://github.com/wangjin2945/UMA-Net.

Investigating the Prognostic Role of Telomerase-Related Cellular Senescence Gene Signatures in Breast Cancer Using Machine Learning

Background: Telomeres and cellular senescence are critical biological processes implicated in cancer development and progression, including breast cancer, through their influence on genomic stability and modulation of the tumor microenvironment. Methods: This study integrated bulk RNA sequencing and single-cell RNA sequencing (scRNA-seq) data to establish a gene signature associated with telomere maintenance and cellular senescence for prognostic prediction in breast cancer. Telomere-related genes (TEGs) and senescence-associated genes were curated from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases. A comprehensive machine learning framework incorporating 101 algorithmic combinations across 10 survival modeling approaches, including random survival forests and ridge regression, was employed to develop a robust prognostic model. Results: A set of 19 key telomere- and senescence-related genes was identified as the optimal prognostic signature. The model demonstrated strong predictive accuracy and was successfully validated in multiple independent cohorts. Functional enrichment analyses indicated significant associations with immune responses and aging-related pathways. Single-cell transcriptomic analysis revealed marked cellular heterogeneity, identifying distinct subpopulations (fibroblasts and immune cells) with divergent risk scores and biological pathway activity. Additionally, pseudo-time trajectory analysis and intercellular communication mapping provided insights into the dynamic evolution of the tumor microenvironment. Immunohistochemical (IHC) validation using data from the Human Protein Atlas confirmed differential protein expression between normal and tumor tissues for several of the selected genes, reinforcing their biological relevance and clinical utility. Conclusions: This study presents a novel 19-gene telomere- and senescence-associated signature with strong prognostic value in breast cancer. These findings enhance our understanding of tumor heterogeneity and may inform precision oncology approaches and future therapeutic strategies.

Two-stage augmentation for detecting malignancy of BI-RADS 3 lesions in early breast cancer

OBJECTIVES

In view of inherent attributes of breast BI-RADS 3, benign and malignant lesions are with a subtle difference and the imbalanced ratio (with a very small part of malignancy). The objective of this study is to improve the detection rate of BI-RADS 3 malignant lesions on breast ultrasound (US) images using deep convolution networks.

METHODS

In the study, 1,275 lesions out of 1,096 patients were included from Southwest Hospital (SW) and Tangshan Hospital (TS). In which, 629 lesions, 218 lesions and 428 lesions were utilized for the development dataset, the internal and external testing set. All malignant lesions were biopsy-confirmed, while benign lesions were verified through biopsy or stable (no significant changes) over a three-year follow-up. And each lesion had both B-mode and color Doppler images. We proposed a two-step augmentation method, covering malignancy feature augmentation and data augmentation, and further verified its feasibility on a dual-branches ResNet50 classification model named Dual-ResNet50. We conducted a comparative analysis between our model and four radiologists in breast imaging diagnosis.

RESULTS

After malignancy feature and data augmentations, our model achieved a high area under the receiver operating characteristic curve (AUC) of 0.881 (95% CI: 0.830-0.921), the sensitivity of 77.8% (14/18), in the SW test set, and an AUC of 0.880 (95% CI: 0.847-0.910), a sensitivity of 71.4% (5/7) in the TS test set. Compared to four radiologists with over 10-years of diagnostic experience, our model outperformed their diagnoses.

CONCLUSIONS

Our proposed augmentation method can help the deep learning (DL) classification model to improve the breast cancer detection rate in BI-RADS 3 lesions, demonstrating its potential to enhance diagnostic accuracy in early breast cancer detection. This improvement aids in a timely adjustment of subsequent treatment for these patients in clinical practice.

A large-scale multicenter breast cancer DCE-MRI benchmark dataset with expert segmentations

Artificial Intelligence (AI) research in breast cancer Magnetic Resonance Imaging (MRI) faces challenges due to limited expert-labeled segmentations. To address this, we present a multicenter dataset of 1506 pre-treatment T1-weighted dynamic contrast-enhanced MRI cases, including expert annotations of primary tumors and non-mass-enhanced regions. The dataset integrates imaging data from four collections in The Cancer Imaging Archive (TCIA), where only 163 cases with expert segmentations were initially available. To facilitate the annotation process, a deep learning model was trained to produce preliminary segmentations for the remaining cases. These were subsequently corrected and verified by 16 breast cancer experts (averaging 9 years of experience), creating a fully annotated dataset. Additionally, the dataset includes 49 harmonized clinical and demographic variables, as well as pre-trained weights for a baseline nnU-Net model trained on the annotated data. This resource addresses a critical gap in publicly available breast cancer datasets, enabling the development, validation, and benchmarking of advanced deep learning models, thus driving progress in breast cancer diagnostics, treatment response prediction, and personalized care.

Comparative analysis of nnU-Net and Auto3Dseg for fat and fibroglandular tissue segmentation in MRI

Purpose

Breast cancer, the most common cancer type among women worldwide, requires early detection and accurate diagnosis for improved treatment outcomes. Segmenting fat and fibroglandular tissue (FGT) in magnetic resonance imaging (MRI) is essential for creating volumetric models, enhancing surgical workflow, and improving clinical outcomes. Manual segmentation is time-consuming and subjective, prompting the development of automated deep-learning algorithms to perform this task. However, configuring these algorithms for 3D medical images is challenging due to variations in image features and preprocessing distortions. Automated machine learning (AutoML) frameworks automate model selection, hyperparameter tuning, and architecture optimization, offering a promising solution by reducing reliance on manual intervention and expert knowledge.

Approach

We compare nnU-Net and Auto3Dseg, two AutoML frameworks, in segmenting fat and FGT on T1-weighted MRI images from the Duke breast MRI dataset (100 patients). We used threefold cross-validation, employing the Dice similarity coefficient (DSC) and Hausdorff distance (HD) metrics for evaluation. The F -test and Tukey honestly significant difference analysis were used to assess statistical differences across methods.

Results

nnU-Net achieved DSC scores of 0.946 ± 0.026 (fat) and 0.872 ± 0.070 (FGT), whereas Auto3DSeg achieved 0.940 ± 0.026 (fat) and 0.871 ± 0.074 (FGT). Significant differences in fat HD ( F = 6.3020 , p < 0.001 ) originated from the full resolution and the 3D cascade U-Net. No evidence of significant differences was found in FGT HD or DSC metrics.

Conclusions

Ensemble approaches of Auto3Dseg and nnU-Net demonstrated comparable performance in segmenting fat and FGT on breast MRI. The significant differences in fat HD underscore the importance of boundary-focused metrics in evaluating segmentation methods.

Radiogenomic Landscape of Metastatic Endocrine-Positive Breast Cancer Resistant to Aromatase Inhibitors

Breast cancer poses a significant global health challenge and includes various subtypes, such as endocrine-positive, HER2-positive, and triple-negative. Endocrine-positive breast cancer, characterized by estrogen and progesterone receptors, is commonly treated with aromatase inhibitors. However, resistance to these inhibitors can hinder patient outcomes due to genetic and epigenetic alterations, mutations in the estrogen receptor 1 gene, and changes in signaling pathways. Radiogenomics combines imaging techniques like MRI and CT scans with genomic profiling methods to identify radiographic biomarkers associated with resistance. This approach enhances our understanding of resistance mechanisms and metastasis patterns, linking them to specific genomic profiles and common metastasis sites like the bone and brain. By integrating radiogenomic data, personalized treatment strategies can be developed, improving predictive and prognostic capabilities. Advancements in imaging and genomic technologies offer promising avenues for enhancing radiogenomic research. A thorough understanding of resistance mechanisms is crucial for developing effective treatment strategies, making radiogenomics a valuable integrative approach in personalized medicine that aims to improve clinical outcomes for patients with metastatic endocrine-positive breast cancer.

Comparison of intensity normalization methods in prostate, brain, and breast cancer multi-parametric magnetic resonance imaging

Objectives

Intensity variation in multi-parametric magnetic resonance imaging (MP-MRI) is a confounding factor in MRI analyses. Previous studies have employed several normalization methods, but there is a lack of consensus on which method results in the most comparable images across vendors and acquisitions. This study used MP-MRI collected from patients with confirmed prostate, brain, or breast cancer to examine common intensity normalization methods to identify which best harmonizes intensity values across cofounds.

Materials and methods

Multiple normalization methods were deployed for intensity comparison between three unique sites, MR vendors, and magnetic field strength. Additionally, we calculated radiomic features before and after intensity normalization to determine how downstream analyses may be affected. Specifically, in the prostate cancer cohort, we tested these methods on T2-weighted imaging (T2WI) and additionally looked at a subset of patients who were scanned with and without the use of an endorectal coil (ERC). In a cohort of glioblastoma (GBM) patients, we tested these methods in T1 pre- and post-contrast enhancement (T1, T1C), fluid attenuated inversion recovery (FLAIR), and apparent diffusion coefficient (ADC) maps. Finally, in the breast cancer cohort, we tested methods on T1-weighted nonfat-suppressed images. All methods were compared using a two one-sided test (TOST) to test for equivalence of mean and standard deviation of intensity distributions.

Results

While each organ had unique results, across every tested comparison, using the Z-score of intensity within a mask of the organ consistently provided an equivalent distribution (all p < 0.001).

Conclusions

Our results suggest that intensity normalization using the Z-score of intensity within prostate, breast, and brain MR images produces the most comparable intensities between sites, MR vendors, magnetic field strength, and prostate endorectal coil usage. Likewise, Z-score normalization provided the highest percentage of radiomic features that were statistically equal across the three organs.

Swarm learning with weak supervision enables automatic breast cancer detection in magnetic resonance imaging

BACKGROUND

Over the next 5 years, new breast cancer screening guidelines recommending magnetic resonance imaging (MRI) for certain patients will significantly increase the volume of imaging data to be analyzed. While this increase poses challenges for radiologists, artificial intelligence (AI) offers potential solutions to manage this workload. However, the development of AI models is often hindered by manual annotation requirements and strict data-sharing regulations between institutions.

METHODS

In this study, we present an integrated pipeline combining weakly supervised learning-reducing the need for detailed annotations-with local AI model training via swarm learning (SL), which circumvents centralized data sharing. We utilized three datasets comprising 1372 female bilateral breast MRI exams from institutions in three countries: the United States (US), Switzerland, and the United Kingdom (UK) to train models. These models were then validated on two external datasets consisting of 649 bilateral breast MRI exams from Germany and Greece.

RESULTS

Upon systematically benchmarking various weakly supervised two-dimensional (2D) and three-dimensional (3D) deep learning (DL) methods, we find that the 3D-ResNet-101 demonstrates superior performance. By implementing a real-world SL setup across three international centers, we observe that these collaboratively trained models outperform those trained locally. Even with a smaller dataset, we demonstrate the practical feasibility of deploying SL internationally with on-site data processing, addressing challenges such as data privacy and annotation variability.

CONCLUSIONS

Combining weakly supervised learning with SL enhances inter-institutional collaboration, improving the utility of distributed datasets for medical AI training without requiring detailed annotations or centralized data sharing.

Deep Radiogenomics Sequencing for Breast Tumor Gene-Phenotype Decoding Using Dynamic Contrast Magnetic Resonance Imaging

PURPOSE

We aim to perform radiogenomic profiling of breast cancer tumors using dynamic contrast magnetic resonance imaging (MRI) for the estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2) genes.

METHODS

The dataset used in the current study consists of imaging data of 922 biopsy-confirmed invasive breast cancer patients with ER, PR, and HER2 gene mutation status. Breast MR images, including a T1-weighted pre-contrast sequence and three post-contrast sequences, were enrolled for analysis. All images were corrected using N4 bias correction algorithms. Based on all images and tumor masks, a bounding box of 128 × 128 × 68 was chosen to include all tumor regions. All networks were implemented in 3D fashion with input sizes of 128 × 128 × 68, and four images were input to each network for multi-channel analysis. Data were randomly split into train/validation (80%) and test set (20%) with stratification in class (patient-wise), and all metrics were reported in 20% of the untouched test dataset.

RESULTS

For ER prediction, SEResNet50 achieved an AUC mean of 0.695 (CI95%: 0.610-0.775), a sensitivity of 0.564, and a specificity of 0.787. For PR prediction, ResNet34 achieved an AUC mean of 0.658 (95% CI: 0.573-0.741), a sensitivity of 0.593, and a specificity of 0.734. For HER2 prediction, SEResNext101 achieved an AUC mean of 0.698 (95% CI: 0.560-0.822), a sensitivity of 0.750, and a specificity of 0.625.

CONCLUSION

The current study demonstrated the feasibility of imaging gene-phenotype decoding in breast tumors using MR images and deep learning algorithms with moderate performance.

Multiphysics simulation of liposome release from hydrogels for cavity filling following patient-specific breast tumor surgery

Several studies have recommended the use of hydrogels for localized targeted delivery of chemotherapeutic drugs following tumor removal surgery. This approach aims to both fill the cavity and prevent cancer recurrence. The use of Multiphysics-based simulation emerges as a valuable strategy for minimizing experimental work, providing detailed insights into how drug release occurs in the tissue, and enabling the optimization of the design. In this study, we introduced a mathematical model, utilizing experimental data, to investigate the transport of liposomes carrying MZ1 from a thermosensitive hydrogel and their impact on the viability of breast cancer cells. The proposed comprehensive model considers not just the transport within the interstitial tissue, represented as a porous medium, but also the uptake by cells and its influence on cell viability, along with the potential lymphatic drainage. The six real patient-specific tumor shapes extracted from MRI scans were used to investigate how the size and form of the tumor can modify the transport pattern. The computational results revealed that the concentration of liposomes in the tissue is significantly influenced by their release from the hydrogel, which proved to be the limiting step. Liposome concentrations of approximately 0.1 % weight were found to be sufficient in ensuring minimal cell survival in the vicinity of the tumor.

Multi-physics numerical simulation study on thermo-sensitive gel delivery for a local post-tumor surgery treatment

Numerous studies in the literature have proposed the use of thermo-responsive hydrogels for filling cavities after tumor resection. However, optimizing the injection process is challenging due to the complex interplay of various multi-physics phenomena, such as the coupling of flow and heat transfer, multi-phase interactions, and phase-change dynamics. Therefore, gaining a fundamental understanding of these processes is crucial. In this study, we introduce a thermo-sensitive hydrogel formulated with poloxamer 407 and Gellan gum as a promising filling agent, offering an ideal phase-transition temperature along with suitable elastic and viscous modulus properties. We performed multi-physics simulations to predict the flow and temperature distributions during hydrogel injection. The results suggested that the hydrogel should be kept at 4 °C and injected within 90 s to avoid reaching the transition temperature. Cavity filling simulations indicated a symmetric distribution of the hydrogel, with minimal influence from the syringe's position. The temperature gradient at the cavity edge delays gelation during injection, which is essential to guarantee its administration as a liquid. The hydrogel's viscosity follows a sigmoidal function relative to temperature, taking five minutes to reach its maximum value. In summary, the multi-physics simulations carried out in this study confirm the potential of thermo-responsive hydrogels for use in post-tumor surgery treatment and define the conditions for a proper administration. Furthermore, the proposed model can be widely applied to other thermo-responsive hydrogels or under different conditions.

Breast Tumor Detection and Diagnosis Using an Improved Faster R-CNN in DCE-MRI

AI-based breast cancer detection can improve the sensitivity and specificity of detection, especially for small lesions, which has clinical value in realizing early detection and treatment so as to reduce mortality. The two-stage detection network performs well; however, it adopts an imprecise ROI during classification, which can easily include surrounding tumor tissues. Additionally, fuzzy noise is a significant contributor to false positives. We adopted Faster RCNN as the architecture, introduced ROI aligning to minimize quantization errors and feature pyramid network (FPN) to extract different resolution features, added a bounding box quadratic regression feature map extraction network and three convolutional layers to reduce interference from tumor surrounding information, and extracted more accurate and deeper feature maps. Our approach outperformed Faster R-CNN, Mask R-CNN, and YOLOv9 in breast cancer detection across 485 internal cases. We achieved superior performance in mAP, sensitivity, and false positive rate ((0.752, 0.950, 0.133) vs. (0.711, 0.950, 0.200) vs. (0.718, 0.880, 0.120) vs. (0.658, 0.680, 405)), which represents a 38.5% reduction in false positives compared to manual detection. Additionally, in a public dataset of 220 cases, our model also demonstrated the best performance. It showed improved sensitivity and specificity, effectively assisting doctors in diagnosing cancer.

Radiogenomics Pilot Study: Association Between Radiomics and Single Nucleotide Polymorphism-Based Microarray Copy Number Variation in Diagnosing Renal Oncocytoma and Chromophobe Renal Cell Carcinoma

RO and ChRCC are kidney tumours with overlapping characteristics, making differentiation between them challenging. The objective of this research is to create a radiogenomics map by correlating radiomic features to molecular phenotypes in ChRCC and RO, using resection as the gold standard. Fourteen patients (6 RO and 8 ChRCC) were included in the prospective study. A total of 1,875 radiomic features were extracted from CT scans, alongside 632 cytobands containing 16,303 genes from the genomic data. Feature selection algorithms applied to the radiomic features resulted in 13 key features. From the genomic data, 24 cytobands highly correlated with histology were selected and cross-correlated with the radiomic features. The analysis identified four radiomic features that were strongly associated with seven genomic features. These findings demonstrate the potential of integrating radiomic and genomic data to enhance the differential diagnosis of RO and ChRCC, paving the way for more precise and non-invasive diagnostic tools in clinical practice.

An explainable longitudinal multi-modal fusion model for predicting neoadjuvant therapy response in women with breast cancer

Multi-modal image analysis using deep learning (DL) lays the foundation for neoadjuvant treatment (NAT) response monitoring. However, existing methods prioritize extracting multi-modal features to enhance predictive performance, with limited consideration on real-world clinical applicability, particularly in longitudinal NAT scenarios with multi-modal data. Here, we propose the Multi-modal Response Prediction (MRP) system, designed to mimic real-world physician assessments of NAT responses in breast cancer. To enhance feasibility, MRP integrates cross-modal knowledge mining and temporal information embedding strategy to handle missing modalities and remain less affected by different NAT settings. We validated MRP through multi-center studies and multinational reader studies. MRP exhibited comparable robustness to breast radiologists, outperforming humans in predicting pathological complete response in the Pre-NAT phase (ΔAUROC 14% and 10% on in-house and external datasets, respectively). Furthermore, we assessed MRP's clinical utility impact on treatment decision-making. MRP may have profound implications for enrolment into NAT trials and determining surgery extensiveness.

Advancing personalized oncology: a systematic review on the integration of artificial intelligence in monitoring neoadjuvant treatment for breast cancer patients

PURPOSE

Despite suffering from the same disease, each patient exhibits a distinct microbiological profile and variable reactivity to prescribed treatments. Most doctors typically use a standardized treatment approach for all patients suffering from a specific disease. Consequently, the challenge lies in the effectiveness of this standardized treatment and in adapting it to each individual patient. Personalized medicine is an emerging field in which doctors use diagnostic tests to identify the most effective medical treatments for each patient. Prognosis, disease monitoring, and treatment planning rely on manual, error-prone methods. Artificial intelligence (AI) uses predictive techniques capable of automating prognostic and monitoring processes, thus reducing the error rate associated with conventional methods.

METHODS

This paper conducts an analysis of current literature, encompassing the period from January 2015 to 2023, based on Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA).

RESULTS

In assessing 25 pertinent studies concerning predicting neoadjuvant treatment (NAT) response in breast cancer (BC) patients, the studies explored various imaging modalities (Magnetic Resonance Imaging, Ultrasound, etc.), evaluating results based on accuracy, sensitivity, and area under the curve. Additionally, the technologies employed, such as machine learning (ML), deep learning (DL), statistics, and hybrid models, were scrutinized. The presentation of datasets used for predicting complete pathological response (PCR) was also considered.

CONCLUSION

This paper seeks to unveil crucial insights into the application of AI techniques in personalized oncology, particularly in the monitoring and prediction of responses to NAT for BC patients. Finally, the authors suggest avenues for future research into AI-based monitoring systems.

The potential of federated learning for self-configuring medical object detection in heterogeneous data distributions

Medical Object Detection (MOD) is a clinically relevant image processing method that locates structures of interest in radiological image data at object-level using bounding boxes. High-performing MOD models necessitate large datasets accurately reflecting the feature distribution of the corresponding problem domain. However, strict privacy regulations protecting patient data often hinder data consolidation, negatively affecting the performance and generalization of MOD models. Federated Learning (FL) offers a solution by enabling model training while the data remain at its original source institution. While existing FL solutions for medical image classification and segmentation demonstrate promising performance, FL for MOD remains largely unexplored. Motivated by this lack of technical solutions, we present an open-source, self-configuring and task-agnostic federated MOD framework. It integrates the FL framework Flower with nnDetection, a state-of-the-art MOD framework and provides several FL aggregation strategies. Furthermore, we evaluate model performance by creating simulated Independent Identically Distributed (IID) and non-IID scenarios, utilizing the publicly available datasets. Additionally, a detailed analysis of the distributions and characteristics of these datasets offers insights into how they can impact performance. Our framework's implementation demonstrates the feasibility of federated self-configuring MOD in non-IID scenarios and facilitates the development of MOD models trained on large distributed databases.

Letter to the Editor Regarding Article "Prior to Initiation of Chemotherapy, Can We Predict Breast Tumor Response? Deep Learning Convolutional Neural Networks Approach Using a Breast MRI Tumor Dataset"

The cited article reports on a convolutional neural network trained to predict response to neoadjuvant chemotherapy from pre-treatment breast MRI scans. The proposed algorithm attains impressive performance on the test dataset with a mean Area Under the Receiver-Operating Characteristic curve of 0.98 and a mean accuracy of 88%. In this letter, I raise concerns that the reported results can be explained by inadvertent data leakage between training and test datasets. More precisely, I conjecture that the random split of the full dataset in training and test sets did not occur on a patient level, but rather on the level of 2D MRI slices. This allows the neural network to "memorize" a patient's anatomy and their treatment outcome, as opposed to discovering useful features for treatment response prediction. To provide evidence for these claims, I present results of similar experiments I conducted on a public breast MRI dataset, where I demonstrate that the suspected data leakage mechanism closely reproduces the results reported on in the cited work.

Identifying radiogenomic associations of breast cancer based on DCE-MRI by using Siamese Neural Network with manufacturer bias normalization

BACKGROUND AND PURPOSE

The immunohistochemical test (IHC) for Human Epidermal Growth Factor Receptor 2 (HER2) and hormone receptors (HR) provides prognostic information and guides treatment for patients with invasive breast cancer. The objective of this paper is to establish a non-invasive system for identifying HER2 and HR in breast cancer using dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI).

METHODS

In light of the absence of high-performance algorithms and external validation in previously published methods, this study utilizes 3D deep features and radiomics features to represent the information of the Region of Interest (ROI). A Siamese Neural Network was employed as the classifier, with 3D deep features and radiomics features serving as the network input. To neutralize manufacturer bias, a batch effect normalization method, ComBat, was introduced. To enhance the reliability of the study, two datasets, Predict Your Therapeutic Response with Imaging and moLecular Analysis (I-SPY 1) and I-SPY 2, were incorporated. I-SPY 2 was utilized for model training and validation, while I-SPY 1 was exclusively employed for external validation. Additionally, a breast tumor segmentation network was trained to improve radiomic feature extraction.

RESULTS

The results indicate that our approach achieved an average Area Under the Curve (AUC) of 0.632, with a Standard Error of the Mean (SEM) of 0.042 for HER2 prediction in the I-SPY 2 dataset. For HR prediction, our method attained an AUC of 0.635 (SEM 0.041), surpassing other published methods in the AUC metric. Moreover, the proposed method yielded competitive results in other metrics. In external validation using the I-SPY 1 dataset, our approach achieved an AUC of 0.567 (SEM 0.032) for HR prediction and 0.563 (SEM 0.033) for HER2 prediction.

CONCLUSION

This study proposes a non-invasive system for identifying HER2 and HR in breast cancer. Although the results do not conclusively demonstrate superiority in both tasks, they indicate that the proposed method achieved good performance and is a competitive classifier compared to other reference methods. Ablation studies demonstrate that both radiomics features and deep features for the Siamese Neural Network are beneficial for the model. The introduced manufacturer bias normalization method has been shown to enhance the method's performance. Furthermore, the external validation of the method enhances the reliability of this research. Source code, pre-trained segmentation network, Radiomics and deep features, data for statistical analysis, and Supporting Information of this article are online at: https://github.com/FORRESTHUACHEN/Siamese_Neural_Network_based_Brest_cancer_Radiogenomic.

Multicenter radio-multiomic analysis for predicting breast cancer outcome and unravelling imaging-biological connection

Radiomics offers a noninvasive avenue for predicting clinicopathological factors. However, thorough investigations into a robust breast cancer outcome-predicting model and its biological significance remain limited. This study develops a robust radiomic model for prognosis prediction, and further excavates its biological foundation and transferring prediction performance. We retrospectively collected preoperative dynamic contrast-enhanced MRI data from three distinct breast cancer patient cohorts. In FUSCC cohort (n = 466), Lasso was used to select features correlated with patient prognosis and multivariate Cox regression was utilized to integrate these features and build the radiomic risk model, while multiomic analysis was conducted to investigate the model's biological implications. DUKE cohort (n = 619) and I-SPY1 cohort (n = 128) were used to test the performance of the radiomic signature in outcome prediction. A thirteen-feature radiomic signature was identified in the FUSCC cohort training set and validated in the FUSCC cohort testing set, DUKE cohort and I-SPY1 cohort for predicting relapse-free survival (RFS) and overall survival (OS) (RFS: p = 0.013, p = 0.024 and p = 0.035; OS: p = 0.036, p = 0.005 and p = 0.027 in the three cohorts). Multiomic analysis uncovered metabolic dysregulation underlying the radiomic signature (ATP metabolic process: NES = 1.84, p-adjust = 0.02; cholesterol biosynthesis: NES = 1.79, p-adjust = 0.01). Regarding the therapeutic implications, the radiomic signature exhibited value when combining clinical factors for predicting the pathological complete response to neoadjuvant chemotherapy (DUKE cohort, AUC = 0.72; I-SPY1 cohort, AUC = 0.73). In conclusion, our study identified a breast cancer outcome-predicting radiomic signature in a multicenter radio-multiomic study, along with its correlations with multiomic features in prognostic risk assessment, laying the groundwork for future prospective clinical trials in personalized risk stratification and precision therapy.

Uncertainty Estimation and Out-of-Distribution Detection for Deep Learning-Based Image Reconstruction Using the Local Lipschitz

Accurate image reconstruction is at the heart of diagnostics in medical imaging. Supervised deep learning-based approaches have been investigated for solving inverse problems including image reconstruction. However, these trained models encounter unseen data distributions that are widely shifted from training data during deployment. Therefore, it is essential to assess whether a given input falls within the training data distribution. Current uncertainty estimation approaches focus on providing an uncertainty map to radiologists, rather than assessing the training distribution fit. In this work, we propose a method based on the local Lipschitz metric to distinguish out-of-distribution images from in-distribution with an area under the curve of 99.94% for True Positive Rate versus False Positive Rate. We demonstrate a very strong relationship between the local Lipschitz value and mean absolute error (MAE), supported by a Spearman's rank correlation coefficient of 0.8475, to determine an uncertainty estimation threshold for optimal performance. Through the identification of false positives, we demonstrate the local Lipschitz and MAE relationship can guide data augmentation and reduce uncertainty. Our study was validated using the AUTOMAP architecture for sensor-to-image Magnetic Resonance Imaging (MRI) reconstruction. We demonstrate our approach outperforms baseline techniques of Monte-Carlo dropout and deep ensembles as well as the state-of-the-art Mean Variance Estimation network approach. We expand our application scope to MRI denoising and Computed Tomography sparse-to-full view reconstructions using UNET architectures. We show our approach is applicable to various architectures and applications, especially in medical imaging, where preserving diagnostic accuracy of reconstructed images remains paramount.

2-D Slicewise Waveform Inversion of Sound Speed and Acoustic Attenuation for Ring Array Ultrasound Tomography Based on a Block LU Solver

Ultrasound tomography is an emerging imaging modality that uses the transmission of ultrasound through tissue to reconstruct images of its mechanical properties. Initially, ray-based methods were used to reconstruct these images, but their inability to account for diffraction often resulted in poor resolution. Waveform inversion overcame this limitation, providing high-resolution images of the tissue. Most clinical implementations, often directed at breast cancer imaging, currently rely on a frequency-domain waveform inversion to reduce computation time. For ring arrays, ray tomography was long considered a necessary step prior to waveform inversion in order to avoid cycle skipping. However, in this paper, we demonstrate that frequency-domain waveform inversion can reliably reconstruct high-resolution images of sound speed and attenuation without relying on ray tomography to provide an initial model. We provide a detailed description of our frequency-domain waveform inversion algorithm with open-source code and data that we make publicly available.

Investigating the effects of artificial intelligence on the personalization of breast cancer management: a systematic study

BACKGROUND

Providing appropriate specialized treatment to the right patient at the right time is considered necessary in cancer management. Targeted therapy tailored to the genetic changes of each breast cancer patient is a desirable feature of precision oncology, which can not only reduce disease progression but also potentially increase patient survival. The use of artificial intelligence alongside precision oncology can help physicians by identifying and selecting more effective treatment factors for patients.

METHOD

A systematic review was conducted using the PubMed, Embase, Scopus, and Web of Science databases in September 2023. We performed the search strategy with keywords, namely: Breast Cancer, Artificial intelligence, and precision Oncology along with their synonyms in the article titles. Descriptive, qualitative, review, and non-English studies were excluded. The quality assessment of the articles and evaluation of bias were determined based on the SJR journal and JBI indices, as well as the PRISMA2020 guideline.

RESULTS

Forty-six studies were selected that focused on personalized breast cancer management using artificial intelligence models. Seventeen studies using various deep learning methods achieved a satisfactory outcome in predicting treatment response and prognosis, contributing to personalized breast cancer management. Two studies utilizing neural networks and clustering provided acceptable indicators for predicting patient survival and categorizing breast tumors. One study employed transfer learning to predict treatment response. Twenty-six studies utilizing machine-learning methods demonstrated that these techniques can improve breast cancer classification, screening, diagnosis, and prognosis. The most frequent modeling techniques used were NB, SVM, RF, XGBoost, and Reinforcement Learning. The average area under the curve (AUC) for the models was 0.91. Moreover, the average values for accuracy, sensitivity, specificity, and precision were reported to be in the range of 90-96% for the models.

CONCLUSION

Artificial intelligence has proven to be effective in assisting physicians and researchers in managing breast cancer treatment by uncovering hidden patterns in complex omics and genetic data. Intelligent processing of omics data through protein and gene pattern classification and the utilization of deep neural patterns has the potential to significantly transform the field of complex disease management.

Comparison of radiomics-based machine-learning classifiers for the pretreatment prediction of pathologic complete response to neoadjuvant therapy in breast cancer

Background

Machine learning classifiers are increasingly used to create predictive models for pathological complete response (pCR) in breast cancer after neoadjuvant therapy (NAT). Few studies have compared the effectiveness of different ML classifiers. This study evaluated radiomics models based on pre- and post-contrast first-phase T1 weighted images (T1WI) in predicting breast cancer pCR after NAT and compared the performance of ML classifiers.

Methods

This retrospective study enrolled 281 patients undergoing NAT from the Duke-Breast-Cancer-MRI dataset. Radiomic features were extracted from pre- and post-contrast first-phase T1WI images. The Synthetic Minority Oversampling Technique (SMOTE) was applied, then the dataset was randomly divided into training and validation groups (7:3). The radiomics model was built using selected optimal features. Support vector machine (SVM), random forest (RF), decision tree (DT), k-nearest neighbor (KNN), extreme gradient boosting (XGBoost), and light gradient boosting machine (LightGBM) were classifiers. Receiver operating characteristic curves were used to assess predictive performance.

Results

LightGBM performed best in predicting pCR [area under the curve (AUC): 0.823, 95% confidence interval (CI) [0.743-0.902], accuracy 74.0%, sensitivity 85.0%, specificity 67.2%]. During subgroup analysis, RF was most effective in pCR prediction in luminal breast cancers (AUC: 0.914, 95% CI [0.847-0.981], accuracy 87.0%, sensitivity 85.2%, specificity 88.1%). In triple-negative breast cancers, LightGBM performed best (AUC: 0.836, 95% CI [0.708-0.965], accuracy 78.6%, sensitivity 68.2%, specificity 90.0%).

Conclusion

The LightGBM-based radiomics model performed best in predicting pCR in patients with breast cancer. RF and LightGBM showed promising results for luminal and triple-negative breast cancers, respectively.

Combination of DCE-MRI and NME-DWI via Deep Neural Network for Predicting Breast Cancer Molecular Subtypes

BACKGROUND

To explore whether the combination of dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI) and nonmono-exponential (NME) model-based diffusion-weighted imaging (DWI) via deep neural network (DNN) can improve the prediction of breast cancer molecular subtypes compared to either imaging technique used alone.

PATIENTS AND METHODS

This prospective study examined 480 breast cancers in 475 patients undergoing DCE-MRI and NME-DWI at 3.0 T. Breast cancers were classified as follows: human epidermal growth factor receptor 2 enriched (HER2-enriched), luminal A, luminal B (HER2-), luminal B (HER2+), and triple-negative subtypes. A total of 20% cases were withheld as an independent test dataset, and the remaining cases were used to train DNN with an 80% to 20% training-validation split and 5-fold cross-validation. The diagnostic accuracies of DNN in 5-way subtype classification between the DCE-MRI, NME-DWI, and their combined multiparametric-MRI datasets were compared using analysis of variance with least significant difference posthoc test. Areas under the receiver-operating characteristic curves were calculated to assess the performances of DNN in binary subtype classification between the 3 datasets.

RESULTS

The 5-way classification accuracies of DNN on both DCE-MRI (0.71) and NME-DWI (0.64) were significantly lower (P < .05) than on multiparametric-MRI (0.76), while on DCE-MRI was significantly higher (P < .05) than on NME-DWI. The comparative results of binary classification between the 3 datasets were consistent with the 5-way classification.

CONCLUSION

The combination of DCE-MRI and NME-DWI via DNN achieved a significant improvement in breast cancer molecular subtype prediction compared to either imaging technique used alone. Additionally, DCE-MRI outperformed NME-DWI in differentiating subtypes.

Prediction of Receptor Status in Radiomics: Recent Advances in Breast Cancer Research

Breast cancer is a multifactorial heterogeneous disease and the leading cause of cancer-related deaths in women; its diagnosis and treatment require clinical sensitivity and a comprehensive disciplinary research approach. The expression of different receptors on tumor cells not only provides the basis for molecular typing of breast cancer but also has a decisive role in the diagnosis, treatment, and prognosis of breast cancer. To date, immunohistochemistry (IHC), which uses invasive histological sampling, has been extensively used in clinical practice to analyze the status of receptors and to make an accurate diagnosis of breast cancer. As an invasive assay, IHC can provide important biological information on tumors at a single point in time, but cannot predict future changes (due to treatment or tumor mutations) without additional invasive procedures. These issues highlight the need to develop a non-invasive method for predicting receptor status. The emerging field of radiomics may offer a non-invasive approach to identification of receptor status without requiring biopsy. In this paper, we present a review of the latest research results in radiomics for predicting the status of breast cancer receptors, with potential important clinical applications.

Predicting disease recurrence in breast cancer patients using machine learning models with clinical and radiomic characteristics: a retrospective study

BACKGROUND

The goal is to use three different machine learning models to predict the recurrence of breast cancer across a very heterogeneous sample of patients with varying disease kinds and stages.

METHODS

A heterogeneous group of patients with varying cancer kinds and stages, including both triple-negative breast cancer (TNBC) and non-triple-negative breast cancer (non-TNBC), was examined. Three distinct models were created using the following five machine learning techniques: Adaptive Boosting (AdaBoost), Random Under-sampling Boosting (RUSBoost), Extreme Gradient Boosting (XGBoost), support vector machines (SVM), and Logistic Regression. The clinical model used both clinical and pathology data in conjunction with the machine learning algorithms. The machine learning algorithms were combined with dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) imaging characteristics in the radiomic model, and the merged model combined the two types of data. Each technique was evaluated using several criteria, including the receiver operating characteristic (ROC) curve, precision, recall, and F1 score.

RESULTS

The results suggest that the integration of clinical and radiomic data improves the predictive accuracy in identifying instances of breast cancer recurrence. The XGBoost algorithm is widely recognized as the most effective algorithm in terms of performance.

CONCLUSION

The findings presented in this study offer significant contributions to the field of breast cancer research, particularly in relation to the prediction of cancer recurrence. These insights hold great potential for informing future investigations and clinical interventions that seek to enhance the accuracy and effectiveness of recurrence prediction in breast cancer patients.

Development and validation of peritumoral vascular and intratumoral radiomics to predict pathologic complete responses to neoadjuvant chemotherapy in patients with triple-negative breast cancer

BACKGROUND

To develop and validate a peritumoral vascular and intratumoral radiomics model to improve pretreatment predictions for pathologic complete responses (pCRs) to neoadjuvant chemoradiotherapy (NAC) in patients with triple-negative breast cancer (TNBC).

METHODS

A total of 282 TNBC patients (93 in the primary cohort, 113 in the validation cohort, and 76 in The Cancer Imaging Archive [TCIA] cohort) were retrospectively included. The peritumoral vasculature on the maximum intensity projection (MIP) from pretreatment DCE-MRI was segmented by a Hessian matrix-based filter and then edited by a radiologist. Radiomics features were extracted from the tumor and peritumoral vasculature of the MIP images. The LASSO method was used for feature selection, and the k-nearest neighbor (k-NN) classifier was trained and validated to build a predictive model. The diagnostic performance was assessed using the ROC analysis.

RESULTS

One hundred of the 282 patient (35.5%) with TNBC achieved pCRs after NAC. In predicting pCRs, the combined peritumoral vascular and intratumoral model (fusion model) yields a maximum AUC of 0.82 (95% confidence interval [CI]: 0.75, 0.88) in the primary cohort, a maximum AUC of 0.67 (95% CI: 0.57, 0.76) in the internal validation cohort, and a maximum AUC of 0.65 (95% CI: 0.52, 0.78) in TCIA cohort. The fusion model showed improved performance over the intratumoral model and the peritumoral vascular model, but not significantly (p > 0.05).

CONCLUSION

This study suggested that combined peritumoral vascular and intratumoral radiomics model could provide a non-invasive tool to enable prediction of pCR in TNBC patients treated with NAC.

Comparative Evaluation of Machine Learning Models for Subtyping Triple-Negative Breast Cancer: A Deep Learning-Based Multi-Omics Data Integration Approach

Objective: Triple-negative breast cancer (TNBC) poses significant diagnostic challenges due to its aggressive nature. This research develops an innovative deep learning (DL) model based on the latest multi-omics data to enhance the accuracy of TNBC subtype and prognosis prediction. The study focuses on addressing the constraints of prior studies by showcasing a model with substantial advancements in data integration, statistical performance, and algorithmic optimization. Methods: Breast cancer-related molecular characteristic data, including mRNA, miRNA, gene mutations, DNA methylation, and magnetic resonance imaging (MRI) images, were retrieved from the TCGA and TCIA databases. This study not only compared single-omics with multi-omics machine learning models but also applied Bayesian optimization to innovatively optimize the neural network structure of a DL model for multi-omics data. Results: The DL model for multi-omics data significantly outperformed single-omics models in subtype prediction, achieving a 98.0% accuracy in cross-validation, 97.0% in the validation set, and 91.0% in an external test set. Additionally, the MRI radiomics model showed promising performance, especially with the training set; however, a decrease in performance during transfer testing underscored the advantages of the DL model for multi-omics data in data consistency and digital processing. Conclusion: Our multi-omics DL model presents notable innovations in statistical performance and transfer learning capability, bearing significant clinical relevance for TNBC classification and prognosis prediction. While the MRI radiomics model proved effective, it requires further optimization for cross-dataset application to enhance accuracy and consistency. Our findings offer new insights into improving TNBC classification and prognosis through multi-omics data and DL algorithms.

Enhancing pathological complete response prediction in breast cancer: the role of dynamic characterization of DCE-MRI and its association with tumor heterogeneity

BACKGROUND

Early prediction of pathological complete response (pCR) is important for deciding appropriate treatment strategies for patients. In this study, we aimed to quantify the dynamic characteristics of dynamic contrast-enhanced magnetic resonance images (DCE-MRI) and investigate its value to improve pCR prediction as well as its association with tumor heterogeneity in breast cancer patients.

METHODS

The DCE-MRI, clinicopathologic record, and full transcriptomic data of 785 breast cancer patients receiving neoadjuvant chemotherapy were retrospectively included from a public dataset. Dynamic features of DCE-MRI were computed from extracted phase-varying radiomic feature series using 22 CAnonical Time-sereis CHaracteristics. Dynamic model and radiomic model were developed by logistic regression using dynamic features and traditional radiomic features respectively. Various combined models with clinical factors were also developed to find the optimal combination and the significance of each components was evaluated. All the models were evaluated in independent test set in terms of area under receiver operating characteristic curve (AUC). To explore the potential underlying biological mechanisms, radiogenomic analysis was implemented on patient subgroups stratified by dynamic model to identify differentially expressed genes (DEGs) and enriched pathways.

RESULTS

A 10-feature dynamic model and a 4-feature radiomic model were developed (AUC = 0.688, 95%CI: 0.635-0.741 and AUC = 0.650, 95%CI: 0.595-0.705) and tested (AUC = 0.686, 95%CI: 0.594-0.778 and AUC = 0.626, 95%CI: 0.529-0.722), with the dynamic model showing slightly higher AUC (train p = 0.181, test p = 0.222). The combined model of clinical, radiomic, and dynamic achieved the highest AUC in pCR prediction (train: 0.769, 95%CI: 0.722-0.816 and test: 0.762, 95%CI: 0.679-0.845). Compared with clinical-radiomic combined model (train AUC = 0.716, 95%CI: 0.665-0.767 and test AUC = 0.695, 95%CI: 0.656-0.714), adding the dynamic component brought significant improvement in model performance (train p < 0.001 and test p = 0.005). Radiogenomic analysis identified 297 DEGs, including CXCL9, CCL18, and HLA-DPB1 which are known to be associated with breast cancer prognosis or angiogenesis. Gene set enrichment analysis further revealed enrichment of gene ontology terms and pathways related to immune system.

CONCLUSION

Dynamic characteristics of DCE-MRI were quantified and used to develop dynamic model for improving pCR prediction in breast cancer patients. The dynamic model was associated with tumor heterogeniety in prognostic-related gene expression and immune-related pathways.

Image Quality Assessment Tool for Conventional and Dynamic Magnetic Resonance Imaging Acquisitions

Image quality assessment of magnetic resonance imaging (MRI) data is an important factor not only for conventional diagnosis and protocol optimization but also for fairness, trustworthiness, and robustness of artificial intelligence (AI) applications, especially on large heterogeneous datasets. Information on image quality in multi-centric studies is important to complement the contribution profile from each data node along with quantity information, especially when large variability is expected, and certain acceptance criteria apply. The main goal of this work was to present a tool enabling users to assess image quality based on both subjective criteria as well as objective image quality metrics used to support the decision on image quality based on evidence. The evaluation can be performed on both conventional and dynamic MRI acquisition protocols, while the latter is also checked longitudinally across dynamic series. The assessment provides an overall image quality score and information on the types of artifacts and degrading factors as well as a number of objective metrics for automated evaluation across series (BRISQUE score, Total Variation, PSNR, SSIM, FSIM, MS-SSIM). Moreover, the user can define specific regions of interest (ROIs) to calculate the regional signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR), thus individualizing the quality output to specific use cases, such as tissue-specific contrast or regional noise quantification.

Deep Learning in Breast Cancer Imaging: State of the Art and Recent Advancements in Early 2024

The rapid advancement of artificial intelligence (AI) has significantly impacted various aspects of healthcare, particularly in the medical imaging field. This review focuses on recent developments in the application of deep learning (DL) techniques to breast cancer imaging. DL models, a subset of AI algorithms inspired by human brain architecture, have demonstrated remarkable success in analyzing complex medical images, enhancing diagnostic precision, and streamlining workflows. DL models have been applied to breast cancer diagnosis via mammography, ultrasonography, and magnetic resonance imaging. Furthermore, DL-based radiomic approaches may play a role in breast cancer risk assessment, prognosis prediction, and therapeutic response monitoring. Nevertheless, several challenges have limited the widespread adoption of AI techniques in clinical practice, emphasizing the importance of rigorous validation, interpretability, and technical considerations when implementing DL solutions. By examining fundamental concepts in DL techniques applied to medical imaging and synthesizing the latest advancements and trends, this narrative review aims to provide valuable and up-to-date insights for radiologists seeking to harness the power of AI in breast cancer care.

Machine learning and new insights for breast cancer diagnosis

Breast cancer (BC) is the most prominent form of cancer among females all over the world. The current methods of BC detection include X-ray mammography, ultrasound, computed tomography, magnetic resonance imaging, positron emission tomography and breast thermographic techniques. More recently, machine learning (ML) tools have been increasingly employed in diagnostic medicine for its high efficiency in detection and intervention. The subsequent imaging features and mathematical analyses can then be used to generate ML models, which stratify, differentiate and detect benign and malignant breast lesions. Given its marked advantages, radiomics is a frequently used tool in recent research and clinics. Artificial neural networks and deep learning (DL) are novel forms of ML that evaluate data using computer simulation of the human brain. DL directly processes unstructured information, such as images, sounds and language, and performs precise clinical image stratification, medical record analyses and tumour diagnosis. Herein, this review thoroughly summarizes prior investigations on the application of medical images for the detection and intervention of BC using radiomics, namely DL and ML. The aim was to provide guidance to scientists regarding the use of artificial intelligence and ML in research and the clinic.

A Channel-Dimensional Feature-Reconstructed Deep Learning Model for Predicting Breast Cancer Molecular Subtypes on Overall b-Value Diffusion-Weighted MRI

BACKGROUND

Dynamic contrast-enhanced (DCE) MRI commonly outperforms diffusion-weighted (DW) MRI in breast cancer discrimination. However, the side effects of contrast agents limit the use of DCE-MRI, particularly in patients with chronic kidney disease.

PURPOSE

To develop a novel deep learning model to fully exploit the potential of overall b-value DW-MRI without the need for a contrast agent in predicting breast cancer molecular subtypes and to evaluate its performance in comparison with DCE-MRI.

STUDY TYPE

Prospective.

SUBJECTS

486 female breast cancer patients (training/validation/test: 64%/16%/20%).

FIELD STRENGTH/SEQUENCE

3.0 T/DW-MRI (13 b-values) and DCE-MRI (one precontrast and five postcontrast phases).

ASSESSMENT

The breast cancers were divided into four subtypes: luminal A, luminal B, HER2+, and triple negative. A channel-dimensional feature-reconstructed (CDFR) deep neural network (DNN) was proposed to predict these subtypes using pathological diagnosis as the reference standard. Additionally, a non-CDFR DNN (NCDFR-DNN) was built for comparative purposes. A mixture ensemble DNN (ME-DNN) integrating two CDFR-DNNs was constructed to identify subtypes on multiparametric MRI (MP-MRI) combing DW-MRI and DCE-MRI.

STATISTICAL TESTS

Model performance was evaluated using accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC). Model comparisons were performed using the one-way analysis of variance with least significant difference post hoc test and the DeLong test. P < 0.05 was considered significant.

RESULTS

The CDFR-DNN (accuracies, 0.79 ~ 0.80; AUCs, 0.93 ~ 0.94) demonstrated significantly improved predictive performance than the NCDFR-DNN (accuracies, 0.76 ~ 0.78; AUCs, 0.92 ~ 0.93) on DW-MRI. Utilizing the CDFR-DNN, DW-MRI attained the predictive performance equal (P = 0.065 ~ 1.000) to DCE-MRI (accuracies, 0.79 ~ 0.80; AUCs, 0.93 ~ 0.95). The predictive performance of the ME-DNN on MP-MRI (accuracies, 0.85 ~ 0.87; AUCs, 0.96 ~ 0.97) was superior to those of both the CDFR-DNN and NCDFR-DNN on either DW-MRI or DCE-MRI.

DATA CONCLUSION

The CDFR-DNN enabled overall b-value DW-MRI to achieve the predictive performance comparable to DCE-MRI. MP-MRI outperformed DW-MRI and DCE-MRI in subtype prediction.

LEVEL OF EVIDENCE

2 TECHNICAL EFFICACY STAGE: 1.

Differentiation of invasive ductal and lobular carcinoma of the breast using MRI radiomic features: a pilot study

Background

Breast cancer (BC) is one of the main causes of cancer-related mortality among women. For clinical management to help patients survive longer and spend less time on treatment, early and precise cancer identification and differentiation of breast lesions are crucial. To investigate the accuracy of radiomic features (RF) extracted from dynamic contrast-enhanced Magnetic Resonance Imaging (DCE MRI) for differentiating invasive ductal carcinoma (IDC) from invasive lobular carcinoma (ILC).

Methods

This is a retrospective study. The IDC of 30 and ILC of 28 patients from Dukes breast cancer MRI data set of The Cancer Imaging Archive (TCIA), were included. The RF categories such as shape based, Gray level dependence matrix (GLDM), Gray level co-occurrence matrix (GLCM), First order, Gray level run length matrix (GLRLM), Gray level size zone matrix (GLSZM), NGTDM (Neighbouring gray tone difference matrix) were extracted from the DCE-MRI sequence using a 3D slicer. The maximum relevance and minimum redundancy (mRMR) was applied using Google Colab for identifying the top fifteen relevant radiomic features. The Mann-Whitney U test was performed to identify significant RF for differentiating IDC and ILC. Receiver Operating Characteristic (ROC) curve analysis was performed to ascertain the accuracy of RF in distinguishing between IDC and ILC.

Results

Ten DCE MRI-based RFs used in our study showed a significant difference (p <0.001) between IDC and ILC. We noticed that DCE RF, such as Gray level run length matrix (GLRLM) gray level variance (sensitivity (SN) 97.21%, specificity (SP) 96.2%, area under curve (AUC) 0.998), Gray level co-occurrence matrix (GLCM) difference average (SN 95.72%, SP 96.34%, AUC 0.983), GLCM interquartile range (SN 95.24%, SP 97.31%, AUC 0.968), had the strongest ability to differentiate IDC and ILC.

Conclusions

MRI-based RF derived from DCE sequences can be used in clinical settings to differentiate malignant lesions of the breast, such as IDC and ILC, without requiring intrusive procedures.

A publicly available deep learning model and dataset for segmentation of breast, fibroglandular tissue, and vessels in breast MRI

Breast density, or the amount of fibroglandular tissue (FGT) relative to the overall breast volume, increases the risk of developing breast cancer. Although previous studies have utilized deep learning to assess breast density, the limited public availability of data and quantitative tools hinders the development of better assessment tools. Our objective was to (1) create and share a large dataset of pixel-wise annotations according to well-defined criteria, and (2) develop, evaluate, and share an automated segmentation method for breast, FGT, and blood vessels using convolutional neural networks. We used the Duke Breast Cancer MRI dataset to randomly select 100 MRI studies and manually annotated the breast, FGT, and blood vessels for each study. Model performance was evaluated using the Dice similarity coefficient (DSC). The model achieved DSC values of 0.92 for breast, 0.86 for FGT, and 0.65 for blood vessels on the test set. The correlation between our model's predicted breast density and the manually generated masks was 0.95. The correlation between the predicted breast density and qualitative radiologist assessment was 0.75. Our automated models can accurately segment breast, FGT, and blood vessels using pre-contrast breast MRI data. The data and the models were made publicly available.

MRI-Based Radiomics Methods for Predicting Ki-67 Expression in Breast Cancer: A Systematic Review and Meta-analysis

RATIONALE AND OBJECTIVES

The purpose of this systematic review and meta-analysis was to assess the quality and diagnostic accuracy of MRI-based radiomics for predicting Ki-67 expression in breast cancer.

MATERIALS AND METHODS

A systematic literature search was performed to find relevant studies published in different databases, including PubMed, Web of Science, and Embase up until March 10, 2023. All papers were independently evaluated for eligibility by two reviewers. Studies that matched research questions and provided sufficient data for quantitative synthesis were included in the systematic review and meta-analysis, respectively. The quality of the articles was assessed using Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) and Radiomics Quality Score (RQS) tools. The predictive value of MRI-based radiomics for Ki-67 antigen in patients with breast cancer was assessed using pooled sensitivity (SEN), specificity, and area under the curve (AUC). Meta-regression was performed to explore the cause of heterogeneity. Different covariates were used for subgroup analysis.

RESULTS

31 studies were included in the systematic review; among them, 21 reported sufficient data for meta-analysis. 20 training cohorts and five validation cohorts were pooled separately. The pooled sensitivity, specificity, and AUC of MRI-based radiomics for predicting Ki-67 expression in training cohorts were 0.80 [95% CI, 0.73-0.86], 0.82 [95% CI, 0.78-0.86], and 0.88 [95%CI, 0.85-0.91], respectively. The corresponding values for validation cohorts were 0.81 [95% CI, 0.72-0.87], 0.73 [95% CI, 0.62-0.82], and 0.84 [95%CI, 0.80-0.87], respectively. Based on QUADAS-2, some risks of bias were detected for reference standard and flow and timing domains. However, the quality of the included article was acceptable. The mean RQS score of the included articles was close to 6, corresponding to 16.6% of the maximum possible score. Significant heterogeneity was observed in pooled sensitivity and specificity of training cohorts (I2 > 75%). We found that using deep learning radiomic methods, magnetic field strength (3 T vs. 1.5 T), scanner manufacturer, region of interest structure (2D vs. 3D), route of tissue sampling, Ki-67 cut-off, logistic regression for model construction, and LASSO for feature reduction as well as PyRadiomics software for feature extraction had a great impact on heterogeneity according to our joint model analysis. Diagnostic performance in studies that used deep learning-based radiomics and multiple MRI sequences (e.g., DWI+DCE) was slightly higher. In addition, radiomic features derived from DWI sequences performed better than contrast-enhanced sequences in terms of specificity and sensitivity. No publication bias was found based on Deeks' funnel plot. Sensitivity analysis showed that eliminating every study one by one does not impact overall results.

CONCLUSION

This meta-analysis showed that MRI-based radiomics has a good diagnostic accuracy in differentiating breast cancer patients with high Ki-67 expression from low-expressing groups. However, the sensitivity and specificity of these methods still do not surpass 90%, restricting them from being used as a supplement to current pathological assessments (e.g., biopsy or surgery) to predict Ki-67 expression accurately.

HRadNet: A Hierarchical Radiomics-Based Network for Multicenter Breast Cancer Molecular Subtypes Prediction

Breast cancer is a heterogeneous disease, where molecular subtypes of breast cancer are closely related to the treatment and prognosis. Therefore, the goal of this work is to differentiate between luminal and non-luminal subtypes of breast cancer. The hierarchical radiomics network (HRadNet) is proposed for breast cancer molecular subtypes prediction based on dynamic contrast-enhanced magnetic resonance imaging. HRadNet fuses multilayer features with the metadata of images to take advantage of conventional radiomics methods and general convolutional neural networks. A two-stage training mechanism is adopted to improve the generalization capability of the network for multicenter breast cancer data. The ablation study shows the effectiveness of each component of HRadNet. Furthermore, the influence of features from different layers and metadata fusion are also analyzed. It reveals that selecting certain layers of features for a specified domain can make further performance improvements. Experimental results on three data sets from different devices demonstrate the effectiveness of the proposed network. HRadNet also has good performance when transferring to other domains without fine-tuning.

The effect of data resampling methods in radiomics

Radiomic datasets can be class-imbalanced, for instance, when the prevalence of diseases varies notably, meaning that the number of positive samples is much smaller than that of negative samples. In these cases, the majority class may dominate the model's training and thus negatively affect the model's predictive performance, leading to bias. Therefore, resampling methods are often utilized to class-balance the data. However, several resampling methods exist, and neither their relative predictive performance nor their impact on feature selection has been systematically analyzed. In this study, we aimed to measure the impact of nine resampling methods on radiomic models utilizing a set of fifteen publicly available datasets regarding their predictive performance. Furthermore, we evaluated the agreement and similarity of the set of selected features. Our results show that applying resampling methods did not improve the predictive performance on average. On specific datasets, slight improvements in predictive performance (+ 0.015 in AUC) could be seen. A considerable disagreement on the set of selected features was seen (only 28.7% of features agreed), which strongly impedes feature interpretability. However, selected features are similar when considering their correlation (82.9% of features correlated on average).

Double Transformer Super-Resolution for Breast Cancer ADC Images

Diffusion-weighted imaging (DWI) has been extensively explored in guiding the clinic management of patients with breast cancer. However, due to the limited resolution, accurately characterizing tumors using DWI and the corresponding apparent diffusion coefficient (ADC) is still a challenging problem. In this paper, we aim to address the issue of super-resolution (SR) of ADC images and evaluate the clinical utility of SR-ADC images through radiomics analysis. To this end, we propose a novel double transformer-based network (DTformer) to enhance the resolution of ADC images. More specifically, we propose a symmetric U-shaped encoder-decoder network with two different types of transformer blocks, named as UTNet, to extract deep features for super-resolution. The basic backbone of UTNet is composed of a locally-enhanced Swin transformer block (LeSwin-T) and a convolutional transformer block (Conv-T), which are responsible for capturing long-range dependencies and local spatial information, respectively. Additionally, we introduce a residual upsampling network (RUpNet) to expand image resolution by leveraging initial residual information from the original low-resolution (LR) images. Extensive experiments show that DTformer achieves superior SR performance. Moreover, radiomics analysis reveals that improving the resolution of ADC images is beneficial for tumor characteristic prediction, such as histological grade and human epidermal growth factor receptor 2 (HER2) status.

Integrating Multi-Omics Data With EHR for Precision Medicine Using Advanced Artificial Intelligence

With the recent advancement of novel biomedical technologies such as high-throughput sequencing and wearable devices, multi-modal biomedical data ranging from multi-omics molecular data to real-time continuous bio-signals are generated at an unprecedented speed and scale every day. For the first time, these multi-modal biomedical data are able to make precision medicine close to a reality. However, due to data volume and the complexity, making good use of these multi-modal biomedical data requires major effort. Researchers and clinicians are actively developing artificial intelligence (AI) approaches for data-driven knowledge discovery and causal inference using a variety of biomedical data modalities. These AI-based approaches have demonstrated promising results in various biomedical and healthcare applications. In this review paper, we summarize the state-of-the-art AI models for integrating multi-omics data and electronic health records (EHRs) for precision medicine. We discuss the challenges and opportunities in integrating multi-omics data with EHRs and future directions. We hope this review can inspire future research and developing in integrating multi-omics data with EHRs for precision medicine.

Radiomic Machine Learning in Invasive Ductal Breast Cancer: Prediction of Ki-67 Expression Level Based on Radiomics of DCE-MRI

PURPOSE

Our study aimed to investigate the potential of radiomics with DCE-MRI for predicting Ki-67 expression in invasive ductal breast cancer.

METHOD

We conducted a retrospective study including 223 patients diagnosed with invasive ductal breast cancer. Radiomics features were extracted from DCE-MRI using 3D-Slicer software. Two Ki-67 expression cutoff values (20% and 29%) were examined. Patients were divided into training (70%) and test (30%) sets. The Elastic Net method selected relevant features, and five machine-learning models were established. Radiomics models were created from intratumoral, peritumoral, and combined regions. Performance was assessed using ROC curves, accuracy, sensitivity, and specificity.

RESULT

For a Ki-67 cutoff value of 20%, the combined model exhibited the highest performance, with area under the curve (AUC) values of 0.838 (95% confidence interval (CI): 0.774-0.897) for the training set and 0.863 (95% CI: 0.764-0.949) for the test set. The AUC values for the tumor model were 0.816 (95% CI: 0.745-0.880) and 0.830 (95% CI: 0.724-0.916), and for the peritumor model were 0.790 (95% CI: 0.711-0.857) and 0.808 (95% CI: 0.682-0.910). When the Ki-67 cutoff value was set at 29%, the combined model also demonstrated superior predictive ability in both training set (AUC: 0.796; 95% CI: 0.724-0.862) and the test set (AUC: 0.823; 95% CI: 0.723-0.911). The AUC values for the tumor model were 0.785 (95% CI: 0.708-0.861) and 0.784 (95% CI: 0.663-0.882), and for the peritumor model were 0.773 (95% CI: 0.690-0.844) and 0.729 (95% CI: 0.603-0.847).

CONCLUSION

Radiomics with DCE-MRI can predict Ki-67 expression in invasive ductal breast cancer. Integrating radiomics features from intratumoral and peritumoral regions yields a dependable prognostic model, facilitating pre-surgical detection and treatment decisions. This holds potential for commercial diagnostic tools.

Radiologic imaging biomarkers in triple-negative breast cancer: a literature review about the role of artificial intelligence and the way forward

Breast cancer is one of the most common and deadly cancers in women. Triple-negative breast cancer (TNBC) accounts for approximately 10%-15% of breast cancer diagnoses and is an aggressive molecular breast cancer subtype associated with important challenges in its diagnosis, treatment, and prognostication. This poses an urgent need for developing more effective and personalized imaging biomarkers for TNBC. Towards this direction, artificial intelligence (AI) for radiologic imaging holds a prominent role, leveraging unique advantages of radiologic breast images, being used routinely for TNBC diagnosis, staging, and treatment planning, and offering high-resolution whole-tumour visualization, combined with the immense potential of AI to elucidate anatomical and functional properties of tumours that may not be easily perceived by the human eye. In this review, we synthesize the current state-of-the-art radiologic imaging applications of AI in assisting TNBC diagnosis, treatment, and prognostication. Our goal is to provide a comprehensive overview of radiomic and deep learning-based AI developments and their impact on advancing TNBC management over the last decade (2013-2024). For completeness of the review, we start with a brief introduction of AI, radiomics, and deep learning. Next, we focus on clinically relevant AI-based diagnostic, predictive, and prognostic models for radiologic breast images evaluated in TNBC. We conclude with opportunities and future directions for AI towards advancing diagnosis, treatment response predictions, and prognostic evaluations for TNBC.

The use of radiomics in magnetic resonance imaging for the pre-treatment characterisation of breast cancers: A scoping review

Radiomics is an emerging field that aims to extract and analyse a comprehensive set of quantitative features from medical images. This scoping review is focused on MRI-based radiomic features for the molecular profiling of breast tumours and the implications of this work for predicting patient outcomes. A thorough systematic literature search and outcome extraction were performed to identify relevant studies published in MEDLINE/PubMed (National Centre for Biotechnology Information), EMBASE and Scopus from 2015 onwards. The following information was retrieved from each article: study purpose, study design, extracted radiomic features, machine learning technique(s), sample size/characteristics, statistical result(s) and implications on patient outcomes. Based on the study purpose, four key themes were identified in the included 63 studies: tumour subtype classification (n = 35), pathologically complete response (pCR) prediction (n = 15), lymph node metastasis (LNM) detection (n = 7) and recurrence rate prediction (n = 6). In all four themes, reported accuracies widely varied among the studies, for example, area under receiver characteristics curve (AUC) for detecting LNM ranged from 0.72 to 0.91 and the AUC for predicting pCR ranged from 0.71 to 0.99. In all four themes, combining radiomic features with clinical data improved the predictive models. Preliminary results of this study showed radiomics potential to characterise the whole tumour heterogeneity, with clear implications for individual-targeted treatment. However, radiomics is still in the pre-clinical phase, currently with an insufficient number of large multicentre studies and those existing studies are often limited by insufficient methodological transparency and standardised workflow. Consequently, the clinical translation of existing studies is currently limited.

Breast Fibroglandular Tissue Segmentation for Automated BPE Quantification With Iterative Cycle-Consistent Semi-Supervised Learning

Background Parenchymal Enhancement (BPE) quantification in Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) plays a pivotal role in clinical breast cancer diagnosis and prognosis. However, the emerging deep learning-based breast fibroglandular tissue segmentation, a crucial step in automated BPE quantification, often suffers from limited training samples with accurate annotations. To address this challenge, we propose a novel iterative cycle-consistent semi-supervised framework to leverage segmentation performance by using a large amount of paired pre-/post-contrast images without annotations. Specifically, we design the reconstruction network, cascaded with the segmentation network, to learn a mapping from the pre-contrast images and segmentation predictions to the post-contrast images. Thus, we can implicitly use the reconstruction task to explore the inter-relationship between these two-phase images, which in return guides the segmentation task. Moreover, the reconstructed post-contrast images across multiple auto-context modeling-based iterations can be viewed as new augmentations, facilitating cycle-consistent constraints across each segmentation output. Extensive experiments on two datasets with various data distributions show great segmentation and BPE quantification accuracy compared with other state-of-the-art semi-supervised methods. Importantly, our method achieves 11.80 times of quantification accuracy improvement along with 10 times faster, compared with clinical physicians, demonstrating its potential for automated BPE quantification. The code is available at https://github.com/ZhangJD-ong/Iterative-Cycle-consistent-Semi-supervised-Learning-for-fibroglandular-tissue-segmentation.

Generalizable attention U-Net for segmentation of fibroglandular tissue and background parenchymal enhancement in breast DCE-MRI

OBJECTIVES

Development of automated segmentation models enabling standardized volumetric quantification of fibroglandular tissue (FGT) from native volumes and background parenchymal enhancement (BPE) from subtraction volumes of dynamic contrast-enhanced breast MRI. Subsequent assessment of the developed models in the context of FGT and BPE Breast Imaging Reporting and Data System (BI-RADS)-compliant classification.

METHODS

For the training and validation of attention U-Net models, data coming from a single 3.0-T scanner was used. For testing, additional data from 1.5-T scanner and data acquired in a different institution with a 3.0-T scanner was utilized. The developed models were used to quantify the amount of FGT and BPE in 80 DCE-MRI examinations, and a correlation between these volumetric measures and the classes assigned by radiologists was performed.

RESULTS

To assess the model performance using application-relevant metrics, the correlation between the volumes of breast, FGT, and BPE calculated from ground truth masks and predicted masks was checked. Pearson correlation coefficients ranging from 0.963 ± 0.004 to 0.999 ± 0.001 were achieved. The Spearman correlation coefficient for the quantitative and qualitative assessment, i.e., classification by radiologist, of FGT amounted to 0.70 (p < 0.0001), whereas BPE amounted to 0.37 (p = 0.0006).

CONCLUSIONS

Generalizable algorithms for FGT and BPE segmentation were developed and tested. Our results suggest that when assessing FGT, it is sufficient to use volumetric measures alone. However, for the evaluation of BPE, additional models considering voxels' intensity distribution and morphology are required.

CRITICAL RELEVANCE STATEMENT

A standardized assessment of FGT density can rely on volumetric measures, whereas in the case of BPE, the volumetric measures constitute, along with voxels' intensity distribution and morphology, an important factor.

KEY POINTS

• Our work contributes to the standardization of FGT and BPE assessment. • Attention U-Net can reliably segment intricately shaped FGT and BPE structures. • The developed models were robust to domain shift.