Externally Validated Multi-Task Learning via Consistency Regularization Using Differentiable BI-RADS Features for Breast Ultrasound Tumor Segmentation

2025-11-19

AI Summary

Multi-task learning for breast ultrasound tumor segmentation often suffers from destructive interference between segmentation and classification tasks, leading to degraded performance and limited generalizability. To address this, we propose a differentiable BI-RADS morphological feature–based consistency regularization method that explicitly enforces semantic alignment between the two tasks within a multi-task framework, thereby mitigating task conflict. Our model is trained on the BrEaST dataset and evaluated on three external datasets—UDIAT, BUSI, and BUS-UCLM—achieving Dice scores of 0.81, 0.66, and 0.69, respectively. These results significantly outperform both single-task baselines and state-of-the-art multi-task approaches, with the UDIAT score representing the current best performance. Notably, this work introduces the first integration of differentiable BI-RADS morphological priors into consistency regularization for medical multi-task learning, enhancing both interpretability and cross-dataset generalizability.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics

Abstract

Multi-task learning can suffer from destructive task interference, where jointly trained models underperform single-task baselines and limit generalization. To improve generalization performance in breast ultrasound-based tumor segmentation via multi-task learning, we propose a novel consistency regularization approach that mitigates destructive interference between segmentation and classification. The consistency regularization approach is composed of differentiable BI-RADS-inspired morphological features. We validated this approach by training all models on the BrEaST dataset (Poland) and evaluating them on three external datasets: UDIAT (Spain), BUSI (Egypt), and BUS-UCLM (Spain). Our comprehensive analysis demonstrates statistically significant (p<0.001) improvements in generalization for segmentation task of the proposed multi-task approach vs. the baseline one: UDIAT, BUSI, BUS-UCLM (Dice coefficient=0.81 vs 0.59, 0.66 vs 0.56, 0.69 vs 0.49, resp.). The proposed approach also achieves state-of-the-art segmentation performance under rigorous external validation on the UDIAT dataset.

Problem

• Improving breast ultrasound tumor segmentation generalization via multi-task learning
• Mitigating destructive task interference between segmentation and classification tasks
• Validating segmentation performance across multiple external ultrasound datasets

Innovation

• Consistency regularization mitigates multi-task interference
• Differentiable BI-RADS features enable morphological consistency
• External validation across three international datasets demonstrates generalization