INTERACT-CMIL: Multi-Task Shared Learning and Inter-Task Consistency for Conjunctival Melanocytic Intraepithelial Lesion Grading

TL;DR

CMIL grading is clinically challenging due to subtle morphology and interrelated criteria. The authors propose INTERACT-CMIL, a multi-head architecture built on a CHIEF foundation encoder that jointly predicts five CMIL axes, with combinatorial partial supervision and an inter-task dependency loss to enforce cross-criterion consistency. The approach achieves superior macro F1 across all axes on a 486-sample, multi-center CMIL dataset, with notable gains over baselines and ablations highlighting the value of pretrained features, dependency regularization, calibration, and selective supervision. This work provides a reproducible, interpretable benchmark for CMIL diagnosis and moves toward standardized digital ocular pathology. The dataset and method hold potential for improved diagnostic coherence and downstream prognostic applications in conjunctival melanoma screening.

Abstract

Accurate grading of Conjunctival Melanocytic Intraepithelial Lesions (CMIL) is essential for treatment and melanoma prediction but remains difficult due to subtle morphological cues and interrelated diagnostic criteria. We introduce INTERACT-CMIL, a multi-head deep learning framework that jointly predicts five histopathological axes; WHO4, WHO5, horizontal spread, vertical spread, and cytologic atypia, through Shared Feature Learning with Combinatorial Partial Supervision and an Inter-Dependence Loss enforcing cross-task consistency. Trained and evaluated on a newly curated, multi-center dataset of 486 expert-annotated conjunctival biopsy patches from three university hospitals, INTERACT-CMIL achieves consistent improvements over CNN and foundation-model (FM) baselines, with relative macro F1 gains up to 55.1% (WHO4) and 25.0% (vertical spread). The framework provides coherent, interpretable multi-criteria predictions aligned with expert grading, offering a reproducible computational benchmark for CMIL diagnosis and a step toward standardized digital ocular pathology.

Figures (3)

• Figure 1: Representative conjunctival biopsy patches illustrating the diagnostic complexity of CMIL grading. Each image is annotated with the five grading criteria: WHO4 (W4), WHO5 (W5), horizontal spread (HS), vertical spread (VS), and cytologic atypia (AP), showing the expert ground truth (GT) and model prediction (Pred.). Left: a high-grade lesion exhibiting nests of melanocytes and marked melanocytic atypia, accurately predicted by our method across all criteria. Right: a benign lesion with limited melanocytic proliferation and orderly epithelium, likewise correctly classified.
• Figure 2: Pipeline overview. Multi-center data (left) produce whole-slide images (WSI). Patches are extracted and fed to a frozen CHIEF feature extractor yielding a 768-D vector, transformed by shared layers into a 256-D shared representation. Five heads (WHO4, WHO5, HS, VS, AP) produce task probabilities; at each iteration, three heads are updated (selective supervision). Training minimizes $\mathcal{L}_{\text{total}}=\mathcal{L}_{\text{cls}}+\lambda\,\mathcal{L}_{\text{dep}}$, where $\mathcal{L}_{\text{dep}}$ aligns the predicted joint distribution $\hat{P}$ with the empirical prior $P$ (schematized as 2D Gaussians).
• Figure 3: ROC Curves of our method for different tasks; WHO4, WHO5, Horizontal Spread, Vertical Spread, and Atypia, respectively.