Evaluating deep learning-based melanoma classification using immunohistochemistry and routine histology: A three center study

Abstract

Pathologists routinely use immunohistochemical (IHC)-stained tissue slides against MelanA in addition to hematoxylin and eosin (H&E)-stained slides to improve their accuracy in diagnosing melanomas. The use of diagnostic Deep Learning (DL)-based support systems for automated examination of tissue morphology and cellular composition has been well studied in standard H&E-stained tissue slides. In contrast, there are few studies that analyze IHC slides using DL. Therefore, we investigated the separate and joint performance of ResNets trained on MelanA and corresponding H&E-stained slides. The MelanA classifier achieved an area under receiver operating characteristics curve (AUROC) of 0.82 and 0.74 on out of distribution (OOD)-datasets, similar to the H&E-based benchmark classification of 0.81 and 0.75, respectively. A combined classifier using MelanA and H&E achieved AUROCs of 0.85 and 0.81 on the OOD datasets. DL MelanA-based assistance systems show the same performance as the benchmark H&E classification and may be improved by multi stain classification to assist pathologists in their clinical routine.

Fig 1. Schematic diagram of the different models.

The red box shows the pipeline for MelanA-stained WSIs and the purple box the pipeline for H&E-stained WSIs. We tessellated MelanA-stained WSIs corresponding to different magnifications and trained individual models on each tile size. The class probabilities for each tile were predicted and aggregated into a slide score by averaging all tile scores. For the H&E-based model we proceeded in the same way.

Fig 2. ROC plots by data modality with corresponding AUROC values.

The different subplots show results for the individual evolved models: A: MelanA-based performance B: H&E-based performance taking all magnifications into account C: combined model using H&E as well as MelanA by aggregating the individual scores. The different colors of the ROC curves show from which data source site the results come: Red: internal results (Dresden), Blue: external results (Erlangen), Purple: external results (Naples).