Related Experiment Video
Updated: Aug 2, 2025
Detection and Isolation of Cancer in Prostate Biopsies Using Stimulated Raman Histology and Artificial Intelligence
Published on: June 10, 2025
A comparative study of the inter-observer variability on Gleason grading against Deep Learning-based approaches for
José M Marrón-Esquivel1, L Duran-Lopez2, A Linares-Barranco2
1Robotics and Tech. of Computers Lab., Universidad de Sevilla, 41012 Seville, Spain; Escuela Técnica Superior de Ingeniería Informática (ETSII), Avenida de Reina Mercedes s/n, Universidad de Sevilla, 41012 Seville, Spain; Escuela Politécnica Superior (EPS), Universidad de Sevilla, 41011 Seville, Spain.
Artificial intelligence can reduce pathologist variability in prostate cancer diagnosis. Deep learning models achieved high accuracy, offering a potential second opinion tool for improved diagnostic consistency.
Area of Science:
- Digital pathology
- Computational pathology
- Artificial intelligence in oncology
Background:
- Prostate cancer is a common diagnosis in men, with mortality reduced by modern medicine but still a leading cause of cancer death.
- Diagnosis relies on biopsy and Whole Slide Images (WSIs) analyzed by pathologists using the Gleason scale, where grade 3+ indicates malignancy.
- Significant inter-observer variability exists among pathologists when assigning Gleason scale values, impacting diagnostic consistency.
Purpose of the Study:
- To analyze inter-observer variability in prostate cancer diagnosis using a local dataset of WSIs.
- To evaluate the performance of Convolutional Neural Network (CNN) architectures in assisting prostate cancer diagnosis.
- To explore the potential of AI as a second opinion tool to reduce diagnostic discrepancies.
Main Methods:
- Analysis of inter-observer variability on 80 WSIs annotated by 5 pathologists at both area and label levels.
- Training six different CNN architectures using four distinct approaches.
- Evaluation of trained CNN models on the same dataset used for variability analysis.
Main Results:
- An inter-observer variability of 0.6946 kappa (κ) was observed, with a 46% discrepancy in annotation area size.
- The best-performing CNN models achieved an accuracy of 0.826±0.014 kappa (κ) on the test set.
- Models trained on data from the same source demonstrated superior performance.
Conclusions:
- Deep learning-based systems show promise in reducing inter-observer variability in prostate cancer diagnosis.
- AI can serve as a valuable second opinion or triage tool for pathologists in medical centers.
- Automated diagnosis systems can enhance diagnostic consistency and support clinical decision-making.
