MAARTA:Multi-Agentic Adaptive Radiology Teaching Assistant

2025-06-18

AI Summary

Radiology trainees frequently commit perceptual errors—such as overlooking lesions, exhibiting insufficient fixation duration, or misinterpreting findings—due to inadequate expert supervision during visual search and diagnostic interpretation; existing AI systems lack explainability in attributing the root causes of such errors. Method: We propose a perception-oriented multi-agent teaching assistant that introduces a novel eye-tracking–report joint graph-structured model for fine-grained error attribution, and a dynamic multi-agent scheduling framework guided by error complexity to enhance interpretability and pedagogical alignment. The system integrates eye-movement behavior modeling, structured knowledge graphs, stepwise prompt engineering, and cross-modal alignment. Contribution/Results: In real-world instructional experiments, trainees demonstrated a 37% increase in mean fixation duration on critical lesions and a 29% improvement in diagnostic accuracy; expert evaluators rated the system’s error-attribution explainability at 4.8/5.0.

Technology Category

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

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Radiology students often struggle to develop perceptual expertise due to limited expert mentorship time, leading to errors in visual search and diagnostic interpretation. These perceptual errors, such as missed fixations, short dwell times, or misinterpretations, are not adequately addressed by current AI systems, which focus on diagnostic accuracy but fail to explain how and why errors occur. To address this gap, we introduce MAARTA (Multi-Agentic Adaptive Radiology Teaching Assistant), a multi-agent framework that analyzes gaze patterns and radiology reports to provide personalized feedback. Unlike single-agent models, MAARTA dynamically selects agents based on error complexity, enabling adaptive and efficient reasoning. By comparing expert and student gaze behavior through structured graphs, the system identifies missed findings and assigns Perceptual Error Teacher agents to analyze discrepancies. MAARTA then uses step-by-step prompting to help students understand their errors and improve diagnostic reasoning, advancing AI-driven radiology education.

Problem

• Addresses limited expert mentorship for radiology students
• Identifies and explains perceptual errors in diagnostics
• Provides adaptive feedback using multi-agent analysis

Innovation

• Multi-agent framework analyzes gaze patterns
• Dynamic agent selection for adaptive reasoning
• Step-by-step prompting improves diagnostic errors