Evaluating reproducibility of AI algorithms in digital pathology with DAPPER

TLDR

Artificial Intelligence is rapidly transforming healthcare, and deep learning applied to digital pathology promises to improve routine reporting, standardize results, and uncover novel histological patterns. The study aims to assess how accurately deep learning–derived predictive models perform in digital pathology, addressing reproducibility concerns. The authors developed the DAPPER framework, a rigorous validation pipeline inspired by FDA MAQC, and applied it to 53,000 GTEx tiles using VGG, ResNet, and Inception feature extractors with three classifiers across multiple class settings, also testing the framework on the KIMIA24 dataset. The DAPPER software, along with the Histological Imaging‑Newsy Tiles benchmark dataset, is released to support standardization and validation in AI for digital pathology.

Abstract

Artificial Intelligence is exponentially increasing its impact on healthcare. As deep learning is mastering computer vision tasks, its application to digital pathology is natural, with the promise of aiding in routine reporting and standardizing results across trials. Deep learning features inferred from digital pathology scans can improve validity and robustness of current clinico-pathological features, up to identifying novel histological patterns, e.g., from tumor infiltrating lymphocytes. In this study, we examine the issue of evaluating accuracy of predictive models from deep learning features in digital pathology, as an hallmark of reproducibility. We introduce the DAPPER framework for validation based on a rigorous Data Analysis Plan derived from the FDA's MAQC project, designed to analyze causes of variability in predictive biomarkers. We apply the framework on models that identify tissue of origin on 787 Whole Slide Images from the Genotype-Tissue Expression (GTEx) project. We test three different deep learning architectures (VGG, ResNet, Inception) as feature extractors and three classifiers (a fully connected multilayer, Support Vector Machine and Random Forests) and work with four datasets (5, 10, 20 or 30 classes), for a total of 53, 000 tiles at 512 × 512 resolution. We analyze accuracy and feature stability of the machine learning classifiers, also demonstrating the need for diagnostic tests (e.g., random labels) to identify selection bias and risks for reproducibility. Further, we use the deep features from the VGG model from GTEx on the KIMIA24 dataset for identification of slide of origin (24 classes) to train a classifier on 1, 060 annotated tiles and validated on 265 unseen ones. The DAPPER software, including its deep learning pipeline and the Histological Imaging-Newsy Tiles (HINT) benchmark dataset derived from GTEx, is released as a basis for standardization and validation initiatives in AI for digital pathology.