In this paper, we propose a structured and interpretable assessment method, decomposing image-based diagnosis into components following the Toulmin model of argumentation. This model includes a claim, grounds, warrant, qualifier, rebuttal, and backing.
Taking a claim generated by a machine learning (ML) model for retinal diagnosis as an example, we argue that this claim should not be accepted at face value; instead, we can either apply explainable AI (XAI) methods or adopt an argumentation-based approach.
In our framework, a model specialized in biomarker extraction from images provides the grounds. The warrant, linking the grounds to the claim, is analyzed by an agent equipped with medical knowledge; in our architecture, this role is fulfilled by the MedGemma agent.
The qualifier is determined based on the overall quantitative evaluation of both the warrant and grounds models. Finally, a rebuttal is constructed using image similarity measures computed with MedSigLip. All these components are presented to the human expert, enabling a more informed and critical assessment of the ML-generated diagnosis.
Blogger's Review: This paper offers an innovative framework using the Toulmin model to enhance the application of machine learning in medical diagnostics, emphasizing the importance of interpretability. In the medical field, transparency and explainability are crucial for improving physicians' trust and decision-making quality.