Related Experiment Video
Updated: Oct 14, 2025
08:05
Detection and Isolation of Cancer in Prostate Biopsies Using Stimulated Raman Histology and Artificial Intelligence
Published on: June 10, 2025
763
Detecting and grading prostate cancer in radical prostatectomy specimens through deep learning techniques.
Petronio Augusto de Souza Melo1, Carmen Liane Neubarth Estivallet1, Miguel Srougi1
1Laboratorio de Pesquisa Medica - LIM55, Divisao de Urologia, Faculdade de Medicina FMUSP, Universidade de Sao Paulo, Sao Paulo, SP, BR.
Clinics (Sao Paulo, Brazil)
|November 3, 2021
Summary
Deep learning algorithms show promise for prostate cancer (PCa) detection and grading in surgical specimens. Scanning methods using deep learning achieved 89% concordance with pathologists, outperforming simple classification.
Area of Science:
- Digital pathology
- Artificial intelligence in oncology
- Computational pathology
Background:
- Prostate cancer (PCa) diagnosis and grading are crucial for treatment decisions.
- Manual analysis of radical prostatectomy specimens can be time-consuming and subject to inter-observer variability.
Purpose of the Study:
- To evaluate the efficacy of deep learning (DL) algorithms for detecting and grading PCa in radical prostatectomy specimens.
- To compare the performance of DL-based categorical classification versus a DL-based scanning method against expert pathologists.
Main Methods:
- Whole-slide images of radical prostatectomy specimens were analyzed using DL.
- Two DL methods were employed: categorical image classification (Inception v3) and a scanning method (Mask R-CNN).
- Performance was evaluated by comparing DL results with pathologist annotations on a test dataset.
Main Results:
- The DL categorical classification method showed 44% concordance with pathologists on the test set.
- The DL scanning method achieved 89% concordance with pathologists on the test set.
- The scanning method demonstrated a validation accuracy of 91.2%.
Conclusions:
- Deep learning algorithms hold significant potential for PCa diagnosis and grading.
- DL-based scanning methods appear more effective than simple classification for analyzing prostatectomy specimens.
