1. A privacy-preserving federated breast cancer prediction system (100), the system (100) comprising: a user device (102) configured to collect real-time medical imaging data including mammography and ultrasound images, patient risk factors from a user through a user interface (104); a communication network (106) configured to establish a communication link for data transmission within the system (100); a processing unit (108) connected to the user device (102) via the communication network (106), the processing unit (108) configured to process real-time data for the prediction of breast cancer, wherein the processing unit (108) further comprises: a data input module (110) configured to receive real-time data from the user device (102); a pre-processing module (112) configured to clean, normalize and reduce noise from the received data; a feature extraction module (114) configured to extract the most relevant features from the pre-processed data including local lesion features, global contextual features and multi-view features; a breast lesion detection module (116) configured to detect and localize abnormal breast lesions based on the extracted features; a breast lesion prediction module (118) configured to predict a malignancy risk score based on the detected breast lesions; a breast lesion classification module (120) configured to classify the predicted malignancy risk scores into predefined categories; a risk-factor fusion module (122) configured to combine the classified breast lesions with patient risk factors to generate risk-aware breast cancer prediction outputs; an explainability module (124) configured to allow the clinician to interpret and understand the breast cancer predictions; a report generation module (126) configured to generate a structured clinical report based on the breast cancer predictions and explainability outputs for use in clinical decision-making and record-keeping; an output module (134) configured to transmit the detected breast lesion, predicted breast lesion, classified breast lesion, fused risk-factor, explainability and generated report.

2. The system (100) as claimed in claim 1, wherein the system (100) further comprises a cloud database (136) configured to store and managed clinical reports and analytics for centralized access, backup, and remote review.

3. The system (100) as claimed in claim 1, wherein the feature extraction module (114) further comprises a hybrid backbone combining convolutional neural networks and vision transformer layers to extract both local and global image features.

4. The system (100) as claimed in claim 1, wherein the breast lesion classification module (120) classify the predicted malignancy risk scores into predefined categories including benign, malignant, normal and assigns breast imaging-reporting and data system scores.

5. The system (100) as claimed in claim 1, wherein the risk-factor fusion module (122) employs a cross-attention mechanism to integrate patient risk factors including age, family history, prior biopsies and prior screening history with imaging-based features.

6. The system (100) as claimed in claim 1, wherein the explainability module (124) configured to generate visualizations including lesion heatmaps, token attributions, and risk-factor contribution scores to support clinician interpretation.

7. The system (100) as claimed in claim 1, wherein the processing unit (108) further comprises an alert generation module (128) configured to generate alerts to notify a clinician based on the predicted malignancy risk exceeding a predefined threshold.

8. The system (100) as claimed in claim 1, wherein the processing unit (108) further comprises a feedback module (130) configured to receive feedback from clinicians regarding the breast cancer prediction results and to use the received feedback to improve the system (100) performance for more accurate predictions in future analyses.

9. The system (100) as claimed in claim 1, wherein the processing unit (108) further comprises a federated learning module (132) configured to train and update the system (100) across multiple sites without sharing raw patient data, and employing privacy-preserving mechanisms including secure aggregation, differential privacy and anomaly detection.

10. A method (200) for privacy-preserving federated breast cancer prediction system (100), the method (200) comprising: collecting real-time medical imaging data including mammography and ultrasound images, patient risk factors from a user through a user interface (104) via the user device (102); establishing a communication link for data transmission within the system (100) via the communication network (106); processing real-time data for the prediction of breast cancer via the processing unit (108), comprising several modules; receiving real-time data from the user device (102) via the data input module (110); cleaning, normalize and reduce noise from the received data via the pre-processing module (112); extracting the most relevant features from the pre-processed data including local lesion features, global contextual features and multi-view features via the feature extraction module (114); detecting and localize abnormal breast lesions based on the extracted features via the breast lesion detection module (116); predicting a malignancy risk score based on the detected breast lesions via the breast lesion prediction module (118); classifying the predicted malignancy risk scores into predefined categories via the breast lesion classification module (120); combining the classified breast lesions with patient risk factors to generate risk-aware breast cancer prediction outputs via the risk-factor fusion module (122); allowing the clinician to interpret and understand the breast cancer predictions via the explainability module (124); generating a structured clinical report based on the breast cancer predictions and explainability outputs for use in clinical decision-making and record-keeping via the report generation module (126); generating alerts to notify a clinician based on the predicted malignancy risk exceeding a predefined threshold via the alert generation module (128); receiving feedback from clinicians regarding the breast cancer prediction results and to use the received feedback to improve the system (100) performance for more accurate predictions in future analyses via the feedback module (130); training and update the system (100) across multiple sites without sharing raw patient data, and employing privacy-preserving mechanisms including secure aggregation, differential privacy and anomaly detection via the federated learning module (132); transmitting the detected breast lesion, predicted breast lesion, classified breast lesion, fused risk-factor, explainability and generated report via the output module (134).

Description:FIELD OF DISCLOSURE
[0001] The present disclosure generally relates to artificial intelligence-based medical imaging and healthcare data analysis, more specifically, relates to a privacy-preserving federated breast cancer prediction system.
BACKGROUND OF THE DISCLOSURE
[0002] A privacy-preserving federated breast cancer prediction system for collaborative analysis of medical imaging and patient risk factors across multiple healthcare institutions. The system performs distributed model training on local hospital data without transferring raw patient information, thereby maintaining data confidentiality while improving predictive accuracy. It further generates interpretable diagnostic outputs and clinical decision support insights to assist healthcare professionals in early detection, risk assessment, and informed management of breast cancer cases.
[0003] Traditional breast cancer detection and prediction systems typically rely on centralized data processing approaches, in which medical imaging data from multiple healthcare institutions is collected and stored at a single location for training artificial intelligence models such systems raise significant concerns related to patient data privacy, regulatory compliance, and secure sharing of sensitive medical information. Additionally, many existing solutions depend primarily on imaging data without incorporating relevant patient risk factors and provide limited interpretability of model predictions, thereby reducing clinical trust and limiting their effective integration into routine diagnostic workflows. Furthermore, these systems often exhibit poor generalization across different hospitals, imaging devices, and patient populations, leading to inconsistent performance. They lack real-time, on-site inference capabilities, requiring high bandwidth and long processing times when large datasets are transferred to centralized servers. Existing solutions also offer minimal support for multi-task outputs, such as concurrent binary, multi-class, and Breast Imaging-Reporting and Data System scoring, and unable to provide integrated monitoring to detect model drift, bias, and fairness issues, which reduces their reliability and safety in clinical deployment. Conventional system lack incremental learning capabilities, making it difficult to update models with new imaging modalities, protocols, and patient populations without retraining from scratch. Many systems fail to provide cross-institutional collaboration, limiting exposure to diverse datasets and reducing robustness against underrepresented demographics. Existing system integrate with hospital information systems for automated reporting, auditing and often lack fail-safe mechanisms, redundancy, and monitoring for real-time system performance, which compromise reliability in clinical deployment. Finally, conventional systems rarely provide standardized, clinician-aligned explainability, making it difficult for radiologists to validate Artificial Intelligence predictions and incorporate them into diagnostic decision-making workflows.
[0004] The present invention solves the problem of the prior art by providing an privacy-preserving federated breast cancer prediction system that enables multiple healthcare institutions to collaboratively train artificial intelligence models without sharing raw patient data. The invention integrates multi-view medical imaging with structured patient risk factors, improving prediction accuracy and enabling risk-aware classification. It provides real-time, on-site inference capabilities through edge processing, reducing dependency on high-bandwidth connections and long processing times. The invention supports multi-task outputs, including binary classification, multi-class lesion prediction, and Breast Imaging-Reporting and Data System scoring, while offering clinician-aligned explainable outputs for improved interpretability and trust. Additionally, the invention includes integrated monitoring for model drift, bias, and fairness, secure aggregation, and privacy-preserving mechanisms, ensuring reliability, safety, and regulatory compliance across diverse institutions. By combining federated learning, hybrid Convolutional Neural Network–Vision Transformer architectures, risk-factor fusion, and standardized explainability, the invention provides a robust, scalable, and clinically deployable solution that overcomes the limitations of traditional centralized Artificial Intelligence systems.
SUMMARY OF THE DISCLOSURE
[0005] The following is a summary description of illustrative embodiments of the invention. It is provided as a preface to assist those skilled in the art to more rapidly assimilate the detailed design discussion which ensues and is not intended in any way to limit the scope of the claims which are appended hereto in order to particularly point out the invention.
[0006] According to illustrative embodiments, the present disclosure focuses on a privacy-preserving federated breast cancer prediction system which overcomes the above-mentioned disadvantages or provide the users with a useful or commercial choice.
[0007] An objective of the present disclosure is to enable collaborative training of artificial intelligence models without sharing raw patient imaging.
[0008] Another objective of the present disclosure is to integrate multi-view medical imaging data for improved lesion detection and classification.
[0009] Another objective of the present disclosure is to incorporate structured patient risk factors such as age, family history, Breast Imaging-Reporting and Data System history, and prior biopsies into the predictive model.
[0010] Another objective of the present disclosure is to use a hybrid Convolutional Neural Network–Vision Transformer architecture to combine local lesion features with global contextual imaging information.
[0011] Another objective of the present disclosure is to provide real-time, on-site inference capabilities at hospital edge devices, reducing dependency on high-bandwidth connections.
[0012] Another objective of the present disclosure is to offer clinician-aligned explainable Artificial Intelligence outputs, including heatmaps, token attributions, and risk-factor contributions.
[0013] Another objective of the present disclosure is to facilitate clinical decision support by providing actionable insights for early detection, triage, and reporting.
[0014] Another objective of the present disclosure is to implement secure aggregation protocols for model updates to maintain patient data privacy and regulatory compliance.
[0015] Another objective of the present disclosure is to provide anomaly detection and differential privacy mechanisms to protect against gradient inversion and model poisoning attacks.
[0016] Another objective of the present disclosure is to monitor model drift, bias, and fairness across institutions and populations, ensuring robust and equitable predictions.
[0017] Another objective of the present disclosure is to support incremental and adaptive learning, allowing updates to models without retraining from scratch.
[0018] Another objective of the present disclosure is to reduce false positives and false negatives in breast cancer detection by combining imaging and risk factors.
[0019] Another objective of the present disclosure is to ensure scalability across hospitals, imaging devices, and patient populations.
[0020] Another objective of the present disclosure is to enable continuous performance evaluation with dashboards for Receiver Operating Curves, confusion matrices, and site-specific monitoring.
[0021] Another objective of the present disclosure is to provide a deployment-ready pipeline with modular components for pre-processing, training, inference, and reporting.
[0022] Yet another objective of the present disclosure is to allow multi-institution collaboration, enabling the use of diverse datasets for improved generalization.
[0023] In light of the above, in one aspect of the present disclosure, a privacy-preserving federated breast cancer prediction system is disclosed herein. The system comprises a user device configured to collect real-time medical imaging data including mammography and ultrasound images, patient risk factors from a user through a user interface. The system includes a communication network configured to establish a communication link for data transmission within the system. The system also includes a processing unit connected to the user device via the communication network, the processing unit configured to process real-time data for the prediction of breast cancer, wherein the processing unit further comprises a data input module configured to receive real-time data from the user device, a pre-processing module configured to clean, normalize and reduce noise from the received data, a feature extraction module configured to extract the most relevant features from the pre-processed data including local lesion features, global contextual features and multi-view features, a breast lesion detection module configured to detect and localize abnormal breast lesions based on the extracted features, a breast lesion prediction module configured to predict a malignancy risk score based on the detected breast lesions, a breast lesion classification module configured to classify the predicted malignancy risk scores into predefined categories, a risk-factor fusion module configured to combine the classified breast lesions with patient risk factors to generate risk-aware breast cancer prediction outputs, an explainability module configured to allow the clinician to interpret and understand the breast cancer predictions, a report generation module configured to generate a structured clinical report based on the breast cancer predictions and explainability outputs for use in clinical decision-making and record-keeping, an output module configured to transmit the detected breast lesion, predicted breast lesion, classified breast lesion, fused risk-factor, explainability and generated report.
[0024] In one embodiment, the cloud database configured to store and manage clinical reports and analytics for centralized access, backup, and remote review.
[0025] In one embodiment, the feature extraction module further comprises a hybrid backbone combining convolutional neural networks and vision transformer layers to extract both local and global image features.
[0026] In one embodiment, the breast lesion classification module classify the predicted malignancy risk scores into predefined categories including benign, malignant, normal and assigns breast imaging-reporting and data system scores.
[0027] In one embodiment, the risk-factor fusion module employs a cross-attention mechanism to integrate patient risk factors including age, family history, prior biopsies, and prior screening history with imaging-based features.
[0028] In one embodiment, the explainability module configured to generate visualizations including lesion heatmaps, token attributions, and risk-factor contribution scores to support clinician interpretation.
[0029] In one embodiment, the alert generation module configured to generate alerts to notify a clinician based on the predicted malignancy risk exceeding a predefined threshold.
[0030] In one embodiment, the feedback module configured to receive feedback from clinicians regarding the breast cancer prediction results and to use the received feedback to improve the system performance for more accurate predictions in future analyses.
[0031] In one embodiment, the federated learning module configured to train and update the system across multiple sites without sharing raw patient data, and employing privacy-preserving mechanisms including secure aggregation, differential privacy and anomaly detection.
[0032] In light of the above, in one aspect of recent disclosure a method of making a privacy-preserving federated breast cancer prediction system is disclosed herein. The method comprises collecting real-time medical imaging data including mammography and ultrasound images, patient risk factors from a user through a user interface via the user device. The method includes establishing a communication link for data transmission within the system via the communication network. The method also includes processing real-time data for the prediction of breast cancer via the processing unit. The method further includes receiving real-time data from the user device via the data input module. The method also includes cleaning, normalize and reduce noise from the received data via the pre-processing module. The method includes extracting the most relevant features from the pre-processed data including local lesion features, global contextual features and multi-view features via the feature extraction module. The method also includes detecting and localize abnormal breast lesions based on the extracted features via the breast lesion detection module. The method includes predicting a malignancy risk score based on the detected breast lesions via the breast lesion prediction module. The method also includes classifying the predicted malignancy risk scores into predefined categories via the breast lesion classification module. The method includes combining the classified breast lesions with patient risk factors to generate risk-aware breast cancer prediction outputs via the risk-factor fusion module. The method also includes allowing the clinician to interpret and understand the breast cancer predictions via the explainability module. The method includes generating a structured clinical report based on the breast cancer predictions and explainability outputs for use in clinical decision-making and record-keeping via the report generation module. The method also includes generating alerts to notify a clinician based on the predicted malignancy risk exceeding a predefined threshold via the alert generation module. Moreover, the method includes receiving feedback from clinicians regarding the breast cancer prediction results and to use the received feedback to improve the system performance for more accurate predictions in future analyses via the feedback module. The method also includes training and update the system across multiple sites without sharing raw patient data, and employing privacy-preserving mechanisms including secure aggregation, differential privacy and anomaly detection via the federated learning module. Furthermore, the method also includes transmitting the detected breast lesion, predicted breast lesion, classified breast lesion, fused risk-factor, explainability and generated report via the output module.
[0033] These and other advantages will be apparent from the present application of the embodiments described herein.
[0034] The preceding is a simplified summary to provide an understanding of some embodiments of the present invention. This summary is neither an extensive nor exhaustive overview of the present invention and its various embodiments. The summary presents selected concepts of the embodiments of the present invention in a simplified form as an introduction to the more detailed description presented below. As will be appreciated, other embodiments of the present invention are possible utilizing, alone or in combination, one or more of the features set forth above or described in detail below.
[0035] These elements, together with the other aspects of the present disclosure and various features are pointed out with particularity in the claims annexed hereto and form a part of the present disclosure. For a better understanding of the present disclosure, its operating advantages, and the specified object attained by its uses, reference should be made to the accompanying drawings and descriptive matter in which there are illustrated exemplary embodiments of the present disclosure.
BRIEF DESCRIPTION OF THE DRAWINGS
[0036] To describe the technical solutions in the embodiments of the present disclosure or in the prior art more clearly, the following briefly describes the accompanying drawings required for describing the embodiments or the prior art. Apparently, the accompanying drawings in the following description merely show some embodiments of the present disclosure, and a person of ordinary skill in the art can derive other implementations from these accompanying drawings without creative efforts. All of the embodiments or the implementations shall fall within the protection scope of the present disclosure.
[0037] The advantages and features of the present disclosure will become better understood with reference to the following detailed description taken in conjunction with the accompanying drawing, in which:
[0038] FIG. 1 illustrates a block diagram of a privacy-preserving federated breast cancer prediction system, in accordance with an embodiment of the present disclosure;
[0039] FIG. 2 illustrates a flowchart of a method, outlining the sequential steps for the privacy-preserving federated breast cancer prediction system, in accordance with an embodiment of the present disclosure;
[0040] Like reference, numerals refer to like parts throughout the description of several views of the drawing.
[0041] The privacy-preserving federated breast cancer prediction system is illustrated in the accompanying drawings, which like reference letters indicate corresponding parts in the various figures. It should be noted that the accompanying figure is intended to present illustrations of exemplary embodiments of the present disclosure. This figure is not intended to limit the scope of the present disclosure. It should also be noted that the accompanying figure is not necessarily drawn to scale.
DETAILED DESCRIPTION OF THE DISCLOSURE
[0042] The following is a detailed description of embodiments of the disclosure depicted in the accompanying drawings. The embodiments are in such detail as to communicate the disclosure. However, the amount of detail offered is not intended to limit the anticipated variations of embodiments; on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the scope of the present disclosure.
[0043] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the present disclosure. It may be apparent to one skilled in the art that embodiments of the present disclosure may be practiced without some of these specific details.
[0044] Various terms as used herein are shown below. To the extent a term is used, it should be given the broadest definition persons in the pertinent art have given that term as reflected in printed publications and issued patents at the time of filing.
[0045] The terms “a” and “an” herein do not denote a limitation of quantity, but rather denote the presence of at least one of the referenced items.
[0046] The terms “having”, “comprising”, “including”, and variations thereof signify the presence of a component.
[0047] Referring now to FIG. 1 to FIG. 2 to describe various exemplary embodiments of the present disclosure. FIG. 1 illustrates a block diagram of a privacy-preserving federated breast cancer prediction system 100, in accordance with an embodiment of the present disclosure.
[0048] The system 100 may include a user device 102. The system 100 may include a user interface 104. The system 100 may include a communication network 106. The system 100 may include a processing unit 108, further comprising a data input module 110, a pre-processing module 112, a feature extraction module 114, a breast lesion detection module 116, a breast lesion prediction module 118, a breast lesion classification module 120, a risk-factor fusion module 122, an explainability module 124, a report generation module 126, an alert generation module 128, a feedback module 130, a federated learning module 132, an output module 134. The system 100 may include a cloud database 136.
[0049] The user device 102 configured to collect real-time medical imaging data including mammography and ultrasound images, patient risk factors from a user through a user interface 104. The user device 102 may include a workstation, desktop computer, tablet, mobile device, and imaging console that allows clinicians to enter patient information, upload imaging data, and provide feedback.
[0050] In one embodiment of the present invention, the user interface 104 further includes graphical elements such as forms, buttons, and display panels that allow users to confirm patient demographics, annotate regions of interest in medical images, and provide feedback on artificial intelligence outputs generated by the system 100.
[0051] In one embodiment of the present invention, the user interface 104 supports real-time visualization of imaging data, enabling clinicians to perform immediate quality checks, verify image clarity, and make necessary adjustments prior to processing by the system 100.
[0052] In one embodiment of the present invention, the user interface 104 includes dashboards for receiver operating characteristic curves, confusion matrices, site-stratified performance, fairness, and drift monitoring. These tools provide accountability, continuous performance assessment, and deployment safety across multiple institutions.
[0053] The communication network 106 configured to establish a communication link for data transmission within the system 100. The communication network 106 enabling secure transmission of real-time clinical parameters, physiological data and health information.
[0054] In one embodiment of the present invention, the communication network 106 may include but not limited to wired and wireless networks.
[0055] In one embodiment of the present invention, the communication network 106 includes wireless networks including but not limited to Bluetooth and wireless fidelity.
[0056] The processing unit 108 connected to the user device 102 via the communication network 106, the processing unit 108 configured to process real-time data for the prediction of breast cancer. The processing unit 108 performs computational operations for analysing the received data and generating prediction outputs that assist clinicians in identifying and assessing potential breast lesions.
[0057] In one embodiment of the present invention, the processing unit 108 may comprise more than one processors, memory units, and associated software components configured to execute algorithms for data analysis, prediction, and decision support within the system 100.
[0058] The data input module 110 configured to receive real-time data from the user device 102. The data input module 110 may include medical imaging data such as mammography images and ultrasound images, along with associated patient information and risk factors.
[0059] The pre-processing module 112 configured to clean, normalize and reduce noise from the received data. The pre-processing module 112 may perform operations including image normalization, contrast enhancement, noise filtering, resolution adjustment, and alignment of multi-view medical images.
[0060] In one embodiment of the present invention, the pre-processing module 112 may further align and standardize multi-view medical images to ensure consistency in image orientation, resolution, and scale. The pre-processing module 112 performs operations including removal of imaging artifacts, noise filtering, contrast enhancement, intensity normalization, image resizing, and pixel intensity scaling of the received mammography and ultrasound images.
[0061] In one embodiment of the present invention, the pre-processing module 112 validates the received patient risk factors, including handling incomplete and inconsistent information, to ensure compatibility with subsequent analytical modules.
[0062] The feature extraction module 114 configured to extract the most relevant features from the pre-processed data including local lesion features, global contextual features and multi-view features. The feature extraction module 114 analyses the pre-processed mammography and ultrasound images to identify discriminative patterns associated with potential breast abnormalities.
[0063] In one embodiment of the present invention, the feature extraction module 114 further comprises a hybrid backbone combining convolutional neural networks and vision transformer layers to extract both local and global image features.
[0064] In one embodiment of the present invention, the feature extraction module 114 further includes local lesion features may include characteristics such as lesion shape, texture, edges, and intensity variations, while global contextual features capture broader anatomical structures and spatial relationships within the breast region.
[0065] In one embodiment of the present invention, the feature extraction module 114 further includes convolutional representations are paired with a Vision Transformer that models global contextual relationships across multi-view mammography images, including cranio-caudal and medio-lateral oblique views.
[0066] In one embodiment of the present invention, the feature extraction module 114 further includes global contextual features capture broader anatomical information of the breast region surrounding the detected lesions, including tissue distribution, structural patterns, spatial relationships between tissues, and background parenchymal characteristics that provide additional context for accurate prediction.
[0067] In one embodiment of the present invention, the feature extraction module 114 further includes multi-view features are derived by combining information from multiple imaging views, such as different mammography projections and complementary ultrasound images, enabling the system 100 to correlate lesion appearance across views and improve detection robustness and prediction accuracy.
[0068] In one embodiment of the present invention, the feature extraction module 114 further includes cross-modal features integrate complementary information from different imaging modalities, such as mammography and ultrasound, as well as structured patient risk factors, to generate a fused representation that enhances the system 100 ability to predict malignancy risk with higher accuracy.
[0069] The breast lesion detection module 116 configured to detect and localize abnormal breast lesions based on the extracted features. The breast lesion detection module 116 may employ deep learning techniques, including convolutional neural networks and vision transformer-based architectures, to generate bounding boxes, segmentation masks, and region proposals corresponding to suspected lesions.
[0070] The breast lesion prediction module 118 configured to predict a malignancy risk score based on the detected breast lesions. The breast lesion prediction module 118 analyses local lesion features, such as shape, texture, margin irregularity, and intensity patterns, together with global contextual features that represent the surrounding breast tissue and spatial relationships, multi-view features obtained from different mammography projections and ultrasound scans, and cross-modal features that integrate information across imaging modalities and structured patient risk factors.
[0071] In one embodiment of the present invention, the breast lesion prediction module 118 applies calibration techniques to adjust the risk scores for site-specific population variations, ensuring that the predictions are consistent and clinically interpretable across multiple institutions.
[0072] In one embodiment of the present invention, the breast lesion prediction module 118 enhances accuracy, reduces false positives and negatives, provides a foundation for explainable and trustworthy Artificial Intelligence-assisted breast cancer prediction.
[0073] The breast lesion classification module 120 configured to classify the predicted malignancy risk scores into predefined categories. The breast lesion classification module 120 utilizes the quantitative risk scores, multi-view and cross-modal features, and patient risk factors to enhance classification accuracy, taking into account both imaging-based and clinical context.
[0074] In one embodiment of the present invention, the breast lesion classification module 120 classify the predicted malignancy risk scores into predefined categories including benign, malignant, normal and assigns breast imaging-reporting and data system scores.
[0075] In one embodiment of the present invention, the breast lesion classification module 120 categorizing lesions into clinically meaningful classes and linking them to standardized Breast Imaging-Reporting and Data System descriptors. The breast lesion classification module 120 facilitates triage, second-read prioritization, and integration with downstream explainability and reporting modules, thereby supporting interpretable and actionable clinical recommendations within the system 100.
[0076] In one embodiment of the present invention, the breast lesion classification module 120 may employ deep learning-based classifiers capable of mapping continuous risk scores to discrete categories while maintaining high sensitivity and specificity.
[0077] The risk-factor fusion module 122 configured to combine the classified breast lesions with patient risk factors to generate risk-aware breast cancer prediction outputs. The risk-factor fusion module 122 enhances predictive accuracy, reduces false positives and negatives, produces outputs that are clinically meaningful and personalized for each patient.
[0078] In one embodiment of the present invention, the risk-factor fusion module 122 employs a cross-attention mechanism to integrate patient risk factors including age, family history, prior biopsies, and prior screening history with imaging-based features.
[0079] The explainability module 124 configured to allow the clinician to interpret and understand the breast cancer predictions. The explainability module 124 enabling clinicians to identify cases requiring for further review. The explainability module 124 enhances clinician trust in the Artificial Intelligence system, facilitates audit and quality assurance processes, and ensures transparency in both federated learning and multi-site deployment scenarios.
[0080] In one embodiment of the present invention, the explainability module 124 generates visualizations including lesion heatmaps, token attributions, and risk-factor contribution scores to support clinician interpretation.
[0081] The report generation module 126 configured to generate a structured clinical report based on the breast cancer predictions and explainability outputs for use in clinical decision-making and record-keeping. The report generation module 126 presenting the results in a clear and structured format and supports clinical decision-making, facilitates second-read reviews, enables audit and quality assurance, and ensures that medical records are complete and compliant with regulatory standards.
[0082] In one embodiment of the present invention, the report generation module 126 further includes Breast Imaging-Reporting and Data System scores, quantitative uncertainty measures, and clinician feedback.
[0083] The alert generation module 128 configured to generate alerts to notify a clinician based on the predicted malignancy risk exceeding a predefined threshold. The alert generation module 128 may support configurable thresholds based on institutional guidelines, patient-specific risk profiles, and regulatory requirements.
[0084] In one embodiment of the present invention, the alert generation module 128 configured to generate alert in the form of short message services, emails and many more.
[0085] The feedback module 130 configured to receive feedback from clinicians regarding the breast cancer prediction results and to use the received feedback to improve the system 100 performance for more accurate predictions in future analyses. The feedback module 130 may include annotations, corrections, confirmations, and suggestions provided via the user interface 104 and stored securely to update the federated learning process across multiple sites.
[0086] The federated learning module 132 configured to train and update the system 100 across multiple sites without sharing raw patient data, and employing privacy-preserving mechanisms including secure aggregation, differential privacy and anomaly detection. The federated learning module 132 supports uncertainty-driven active learning, where high-uncertainty cases are flagged for clinician review and subsequently used to improve model performance across the network.
[0087] In one embodiment of the present invention, the federated learning module 132 further includes coordination server securely aggregates encrypted model updates from multiple sites to form a global model while ensuring differential privacy, anomaly detection, and secure aggregation, mitigating risks of data leakage. The federated learning module 132 supports multi-institution diversity to improve model generalization across scanners, protocols, and patient populations.
[0088] The output module 134 configured to transmit the detected breast lesion, predicted breast lesion, classified breast lesion, fused risk-factor, explainability and generated report. The output module 134 ensures that these results are delivered to the clinician, hospital information systems, and the cloud database 136 for real-time review, decision support, record-keeping, and downstream analytics, while maintaining privacy and security of patient data.
[0089] The cloud database 136 configured to store and manage clinical reports, and analytics for centralized access, backup, and remote review. The cloud database 136 ensures that all stored data remains encrypted and privacy-compliant, consistent with federated learning principles, while supporting collaborative use across multiple institutions.
[0090] In one embodiment of the present invention, the cloud database 136 provides centralized access for authorized clinicians and administrators, supports remote review, enables backup and disaster recovery, and facilitates long-term data analytics for quality assurance, performance monitoring, and regulatory compliance.
[0091] FIG. 2 illustrates a flowchart of a method, outlining the sequential steps for the privacy-preserving federated breast cancer prediction system 100, in accordance with an embodiment of the present disclosure;
[0092] At step 202, the real-time medical imaging data is collected including mammography and ultrasound images, patient risk factors from a user through a user interface 104 via the user device 102.
[0093] At step 204, the communication link is established for data transmission within the system 100 via the communication network 106.
[0094] At step 206, the real-time data is processed for the prediction of breast cancer via the processing unit 108, comprising several modules.
[0095] At step 208, the real-time data is received from the user device 102 via the data input module 110.
[0096] At step 210, the received data is cleaned, normalize and reduce noise via the pre-processing module 112.
[0097] At step 212, the most relevant features is extracted from the pre-processed data including local lesion features, global contextual features and multi-view features via the feature extraction module 114.
[0098] At step 214, the abnormal breast lesions is detected and localize based on the extracted features via the breast lesion detection module 116.
[0099] At step 216, the malignancy risk score is predicted based on the detected breast lesions via the breast lesion prediction module 118.
[0100] At step 218, the predicted malignancy risk scores is classified into predefined categories via the breast lesion classification module 120.
[0101] At step 220, the classified breast lesions is combined with patient risk factors to generate risk-aware breast cancer prediction outputs via the risk-factor fusion module 122.
[0102] At step 222, the clinician is allowed to interpret and understand the breast cancer predictions via the explainability module 124.
[0103] At step 224, the structured clinical report is generated based on the breast cancer predictions and explainability outputs for use in clinical decision-making and record-keeping via the report generation module 126.
[0104] At step 226, the alerts is generated to notify a clinician based on the predicted malignancy risk exceeding a predefined threshold via the alert generation module 128.
[0105] At step 228, the clinicians is received feedback regarding the breast cancer prediction results and to use the received feedback to improve the system 100 performance for more accurate predictions in future analyses 130.
[0106] At step 230, the system 100 is trained and update across multiple sites without sharing raw patient data, and employing privacy-preserving mechanisms including secure aggregation, differential privacy and anomaly detection via the federated learning module 132.
[0107] At step 232, the detected breast lesion, predicted breast lesion, classified breast lesion, fused risk-factor, explainability is transmitted and generated report via the output module 134.
[0108] In the best mode of operation of the present invention, the system 100 operates by collecting real-time medical imaging data, including mammography and ultrasound images, along with patient risk factors, through the user device 102 and the user interface 104. The collected data is transmitted via the communication network 106 to the processing unit 108, where it is received by the data input module 110 and processed by the pre-processing module 112 to clean, normalize, and reduce noise, ensuring uniform quality across devices and imaging protocols. The feature extraction module 114 extracts the most relevant features, including local lesion features, global contextual features, multi-view features, and cross-modal features using a hybrid convolutional neural network and vision transformer backbone. The breast lesion detection module 116 identifies and localizes abnormal lesions based on the extracted features, followed by the breast lesion prediction module 118, which computes a malignancy risk score for each detected lesion. The breast lesion classification module 120 classifies the predicted malignancy risk scores into predefined categories, such as benign, malignant, and normal, while optionally assigning Breast Imaging-Reporting and Data System scores. The risk-factor fusion module 122 integrates patient risk factors, including age, family history, prior biopsies, and prior screening history, with imaging features using a cross-attention mechanism to produce risk-aware predictions. The explainability module 124 generates visualizations, including lesion heatmaps, token attributions, and risk-factor contribution scores, allowing clinicians to interpret and validate the Artificial Intelligence predictions. The report generation module 126 produces structured clinical reports based on the predictions and explainability outputs, suitable for clinical decision-making, triage, audit, and record-keeping. The alert generation module 128 may notify clinicians when predicted malignancy risks exceed predefined thresholds, and the feedback module 130 allows clinicians to provide feedback to continuously improve the system 100 predictions. The federated learning module 132 enables the system 100 to train and update across multiple sites without sharing raw patient data, employing secure aggregation, differential privacy, and anomaly detection to maintain privacy while improving generalization and robustness. The output module 134 transmits the results, including detected lesions, predicted risk scores, classified lesions, fused risk-aware outputs, explainability visualizations, and reports, to clinicians and the cloud database 136 for centralized storage, backup, analytics, and remote access. The system 100 is designed to operate with modularity, allowing incremental addition of imaging modalities, multi-task prediction heads, and deployment-ready integration with hospital Graphics Processing Unit infrastructure, ensuring accurate, interpretable, and privacy-preserving breast cancer prediction across diverse clinical settings.
[0109] While the invention has been described in connection with what is presently considered to be the most practical and various embodiments, it will be understood that the invention is not to be limited to the disclosed embodiments, but on the contrary, is intended to cover various modifications and equivalent arrangements included within the scope of the appended claims.
[0110] A person of ordinary skill in the art may be aware that, in combination with the examples described in the embodiments disclosed in this specification, units and algorithm steps may be implemented by electronic hardware, computer software, or a combination thereof.
[0111] The foregoing descriptions of specific embodiments of the present disclosure have been presented for purposes of illustration and description. They are not intended to be exhaustive or to limit the present disclosure to the precise forms disclosed, and many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described to best explain the principles of the present disclosure and its practical application, and to thereby enable others skilled in the art to best utilize the present disclosure and various embodiments with various modifications as are suited to the particular use contemplated. It is understood that various omissions and substitutions of equivalents are contemplated as circumstances may suggest or render expedient, but such omissions and substitutions are intended to cover the application or implementation without departing from the scope of the present disclosure.
[0112] Disjunctive language such as the phrase “at least one of X, Y, Z,” unless specifically stated otherwise, is otherwise understood with the context as used in general to present that an item, term, etc., may be either X, Y, or Z, or any combination thereof (e.g., X, Y, and/or Z). Thus, such disjunctive language is not generally intended to, and should not, imply that certain embodiments require at least one of X, at least one of Y, or at least one of Z to each be present.
[0113] In a case that no conflict occurs, the embodiments in the present disclosure and the features in the embodiments may be mutually combined. The foregoing descriptions are merely specific implementations of the present disclosure but are not intended to limit the protection scope of the present disclosure. Any variation or replacement readily figured out by a person skilled in the art within the technical scope disclosed in the present disclosure shall fall within the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure shall be subject to the protection scope of the claims.
, Claims:I/We Claim:
1. A privacy-preserving federated breast cancer prediction system (100), the system (100) comprising:
a user device (102) configured to collect real-time medical imaging data including mammography and ultrasound images, patient risk factors from a user through a user interface (104);
a communication network (106) configured to establish a communication link for data transmission within the system (100);
a processing unit (108) connected to the user device (102) via the communication network (106), the processing unit (108) configured to process real-time data for the prediction of breast cancer, wherein the processing unit (108) further comprises:
a data input module (110) configured to receive real-time data from the user device (102);
a pre-processing module (112) configured to clean, normalize and reduce noise from the received data;
a feature extraction module (114) configured to extract the most relevant features from the pre-processed data including local lesion features, global contextual features and multi-view features;
a breast lesion detection module (116) configured to detect and localize abnormal breast lesions based on the extracted features;
a breast lesion prediction module (118) configured to predict a malignancy risk score based on the detected breast lesions;
a breast lesion classification module (120) configured to classify the predicted malignancy risk scores into predefined categories;
a risk-factor fusion module (122) configured to combine the classified breast lesions with patient risk factors to generate risk-aware breast cancer prediction outputs;
an explainability module (124) configured to allow the clinician to interpret and understand the breast cancer predictions;
a report generation module (126) configured to generate a structured clinical report based on the breast cancer predictions and explainability outputs for use in clinical decision-making and record-keeping;
an output module (134) configured to transmit the detected breast lesion, predicted breast lesion, classified breast lesion, fused risk-factor, explainability and generated report.
2. The system (100) as claimed in claim 1, wherein the system (100) further comprises a cloud database (136) configured to store and managed clinical reports and analytics for centralized access, backup, and remote review.
3. The system (100) as claimed in claim 1, wherein the feature extraction module (114) further comprises a hybrid backbone combining convolutional neural networks and vision transformer layers to extract both local and global image features.
4. The system (100) as claimed in claim 1, wherein the breast lesion classification module (120) classify the predicted malignancy risk scores into predefined categories including benign, malignant, normal and assigns breast imaging-reporting and data system scores.
5. The system (100) as claimed in claim 1, wherein the risk-factor fusion module (122) employs a cross-attention mechanism to integrate patient risk factors including age, family history, prior biopsies and prior screening history with imaging-based features.
6. The system (100) as claimed in claim 1, wherein the explainability module (124) configured to generate visualizations including lesion heatmaps, token attributions, and risk-factor contribution scores to support clinician interpretation.
7. The system (100) as claimed in claim 1, wherein the processing unit (108) further comprises an alert generation module (128) configured to generate alerts to notify a clinician based on the predicted malignancy risk exceeding a predefined threshold.
8. The system (100) as claimed in claim 1, wherein the processing unit (108) further comprises a feedback module (130) configured to receive feedback from clinicians regarding the breast cancer prediction results and to use the received feedback to improve the system (100) performance for more accurate predictions in future analyses.
9. The system (100) as claimed in claim 1, wherein the processing unit (108) further comprises a federated learning module (132) configured to train and update the system (100) across multiple sites without sharing raw patient data, and employing privacy-preserving mechanisms including secure aggregation, differential privacy and anomaly detection.
10. A method (200) for privacy-preserving federated breast cancer prediction system (100), the method (200) comprising:
collecting real-time medical imaging data including mammography and ultrasound images, patient risk factors from a user through a user interface (104) via the user device (102);
establishing a communication link for data transmission within the system (100) via the communication network (106);
processing real-time data for the prediction of breast cancer via the processing unit (108), comprising several modules;
receiving real-time data from the user device (102) via the data input module (110);
cleaning, normalize and reduce noise from the received data via the pre-processing module (112);
extracting the most relevant features from the pre-processed data including local lesion features, global contextual features and multi-view features via the feature extraction module (114);
detecting and localize abnormal breast lesions based on the extracted features via the breast lesion detection module (116);
predicting a malignancy risk score based on the detected breast lesions via the breast lesion prediction module (118);
classifying the predicted malignancy risk scores into predefined categories via the breast lesion classification module (120);
combining the classified breast lesions with patient risk factors to generate risk-aware breast cancer prediction outputs via the risk-factor fusion module (122);
allowing the clinician to interpret and understand the breast cancer predictions via the explainability module (124);
generating a structured clinical report based on the breast cancer predictions and explainability outputs for use in clinical decision-making and record-keeping via the report generation module (126);
generating alerts to notify a clinician based on the predicted malignancy risk exceeding a predefined threshold via the alert generation module (128);
receiving feedback from clinicians regarding the breast cancer prediction results and to use the received feedback to improve the system (100) performance for more accurate predictions in future analyses via the feedback module (130);
training and update the system (100) across multiple sites without sharing raw patient data, and employing privacy-preserving mechanisms including secure aggregation, differential privacy and anomaly detection via the federated learning module (132);
transmitting the detected breast lesion, predicted breast lesion, classified breast lesion, fused risk-factor, explainability and generated report via the output module (134).