ScaFE: Data-Efficient Scar Classification with LLM-Generated Clinical Feature Programs

2026-07-30

AI Summary

This study addresses the challenges in pathological scar classification posed by scarce annotated data, heterogeneous multi-center imaging, and the data dependency and limited auditability of end-to-end models. The authors propose a privacy-preserving, data-efficient framework that leverages a large language model (LLM) to translate clinical knowledge into deterministic, executable feature programs deployable locally. Only aggregated statistics and SHAP-based feature importance are shared for model iteration, ensuring raw images remain within their originating institutions. A lightweight random forest classifier is then applied for final prediction. This approach achieves the first successful conversion of LLM-derived knowledge into auditable, executable medical feature programs, attaining an 81.0% site-wise macro-averaged accuracy across 600 images from three hospitals—surpassing the strongest baseline by 10 percentage points. Remarkably, it maintains 72.0% accuracy with only 10% of the data, achieves a 95.0% program executability rate, and ensures 91.7% of features are clinically grounded.

Abstract

Classifying pathological scars from clinical photographs requires distinguishing keloids from hypertrophic scars despite limited expert-labeled data and substantial acquisition variation across hospitals. End-to-end image models remain data-dependent, whereas sending photographs to a hosted vision-language model (VLM) may conflict with local data-governance requirements and yields decisions that are difficult to reproduce and audit. We introduce ScaFE (Scar Feature Engineering), which transfers clinical knowledge from a large language model (LLM) into deterministic, executable feature programs instead of asking the model to diagnose images. A web-enabled LLM retrieves clinical evidence and synthesizes programs that measure visually assessable scar attributes. Candidate programs execute in a restricted local environment, and only aggregate validation statistics and feature-level SHAP summaries are returned for iterative repair and refinement; raw images and patient-level outputs remain local. A lightweight Random Forest then operates on the resulting structured representation. On 600 photographs from three hospitals under leave-one-site-out evaluation, ScaFE achieves 81.0% site-macro balanced accuracy, exceeding the strongest baseline, BiomedCLIP, by 10.0 percentage points. With only 10% of the development data, ScaFE retains 72.0% balanced accuracy and an 11.8-point lead. Iterative refinement also raises the executable-program rate from 66.7% to 95.0%, with verified evidence for 91.7% of the final features. These results show that LLM knowledge can support data-efficient, cross-site medical image classification through local and auditable feature programs rather than direct VLM decisions.

Problem

• scar classification
• data efficiency
• clinical photographs
• cross-site variation
• medical image analysis

Innovation

• LLM-generated feature programs
• data-efficient medical classification
• local executable programs
• cross-site generalization
• auditable AI