Experience-Guided Self-Adaptive Cascaded Agents for Breast Cancer Screening and Diagnosis with Reduced Biopsy Referrals

2026-02-27

AI Summary

This study addresses the problem of excessive biopsy referrals in breast ultrasound screening, which leads to unnecessary resource consumption and patient burden. To mitigate this, the authors propose BUSD-Agent, an experience-guided cascaded multi-agent framework that models the screening process as a two-stage decision pipeline: a lightweight screening agent first filters out low-risk cases, while a diagnostically sophisticated agent with enhanced perceptual capabilities performs secondary evaluation on high-risk cases. A novel memory mechanism based on structured decision trajectories is introduced, enabling retrieval-based contextual adaptation without parameter updates and allowing dynamic adjustment of model confidence and referral thresholds. Evaluated across ten datasets, the approach reduces biopsy referral rates from 59.50% to 37.08% and diagnostic upgrade rates from 84.95% to 58.72%, while improving screening and diagnostic specificity by 68.48% and 6.33%, respectively.

Technology Category

• Multiagent Systems: Modeling other Agents
• Search and Optimization: Distributed Search
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

We propose an experience-guided cascaded multi-agent framework for Breast Ultrasound Screening and Diagnosis, called BUSD-Agent, that aims to reduce diagnostic escalation and unnecessary biopsy referrals. Our framework models screening and diagnosis as a two-stage, selective decision-making process. A lightweight `screening clinic'agent, restricted to classification models as tools, selectively filters out benign and normal cases from further diagnostic escalation when malignancy risk and uncertainty are estimated as low. Cases that have higher risks are escalated to the `diagnostic clinic'agent, which integrates richer perception and radiological description tools to make a secondary decision on biopsy referral. To improve agent performance, past records of pathology-confirmed outcomes along with image embeddings, model predictions, and historical agent actions are stored in a memory bank as structured decision trajectories. For each new case, BUSD-Agent retrieves similar past cases based on image, model response and confidence similarity to condition the agent's current decision policy. This enables retrieval-conditioned in-context adaptation that dynamically adjusts model trust and escalation thresholds from prior experiences without parameter updates. Evaluation across 10 breast ultrasound datasets shows that the proposed experience-guided workflow reduces diagnostic escalation in BUSD-Agent from 84.95% to 58.72% and overall biopsy referrals from 59.50% to 37.08%, compared to the same architecture without trajectory conditioning, while improving average screening specificity by 68.48% and diagnostic specificity by 6.33%.

Problem

• breast cancer screening
• biopsy referral reduction
• diagnostic escalation
• ultrasound diagnosis
• unnecessary biopsy

Innovation

• experience-guided agents
• cascaded decision-making
• retrieval-based adaptation
• biopsy referral reduction
• memory-augmented diagnosis