Journal DOI 10.68956/journal.human-performance-futures

Creative Commons source reading · Explainable decision support

Development and Evaluation of a Deep Learning Model for Ovarian Cancer Histotype Classification Using Whole-Slide Imaging

Dagoberto Pulido; Nathalia Arias-Mendoza

Originally published in J Imaging · 2026-03-25

From the article

Selected source text. JournalOS removes markup and shortens excerpts where indicated. The full article provides the complete context.

Abstract

The histopathological classification of ovarian carcinoma is fundamental for patient management. While microscopic evaluation by pathologists is the current diagnostic standard, it is known to be subject to interobserver variability, which can affect consistency in treatment decisions. This study addresses this clinical need by developing and validating a deep learning-based diagnostic support tool designed to enhance the objectivity and reproducibility of this classification. In this work, we address a key challenge in computational pathology—the tendency of attention mechanisms to overfit by concentrating on limited features—by systematically evaluating a direct regularization method within multiple instance learning (MIL) models. The models were trained and validated using 10-fold cross-validation on a public training set of 538…

4. Results

To develop a tool that could reliably address diagnostic variability, we systematically evaluated our models, first on the primary task of histological classification on 538 WSIs. As detailed in Table 2, the stochastic attention model with k=4 (SA-4) demonstrated the highest overall performance. A key finding was that all attention-based models consistently yielded higher performance metrics than the non-attention-based dynamic instance pruning (DIP) model, underscoring the importance of the attention mechanism for this diagnostic task. In direct contrast with the low inter-rater agreement (Kappa coefficients as low as 0.34) reported in manual assessment [6], our best model achieved a Cohen’s Kappa of 0.791, alongside a balanced accuracy of 0.854. This highlights the model’s potential to provide a highly reproducible classification…

5. Discussion

This study presents a high-performing AI tool that directly addresses a key unmet need in diagnostic pathology: the significant interobserver variability in the classification of ovarian carcinoma histotypes. Where traditional microscopic assessment can yield diagnostic agreement as low as 60% [5], our tool provides a highly consistent and accurate classification, achieving a balanced accuracy of 0.854 and a strong Kappa score of 0.791. The robustness of this approach was further validated by its high performance on the distinct and more difficult task of molecular subtype classification from histology, confirming its generalization capabilities across problems of varying complexity. The performance of our model is comparable to other state-of-the-art methods evaluated on the UBC-OCEAN dataset. For instance, the top-performing…

Editorial reading guide

  • What data created and tested the model, and was evaluation independent?
  • What is the cost of a false positive, false negative or out-of-distribution decision?