Related Experiment Video
Updated: Dec 28, 2025
05:33
Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System
Published on: July 11, 2025
696
Tailored for Real-World: A Whole Slide Image Classification System Validated on Uncurated Multi-Site Data Emulating
Julianna D Ianni1, Rajath E Soans2, Sivaramakrishnan Sankarapandian3
1Proscia Inc., Philadelphia, Pennsylvania, USA. julianna@proscia.com.
Scientific Reports
|February 22, 2020
Summary
A new deep learning system accurately classifies digitized skin cancer slides, achieving up to 98% accuracy with confidence scoring. This AI tool promises faster diagnoses and improved efficiency in dermatopathology.
Area of Science:
- Dermatopathology
- Artificial Intelligence
- Medical Diagnostics
Background:
- Current skin cancer diagnosis relies on pathologist examination of H&E stained slides.
- High inter-pathologist variability and increasing biopsy rates highlight the need for improved diagnostic efficiency and reproducibility.
Purpose of the Study:
- To develop and validate a deep learning system for classifying digitized dermatopathology slides.
- To assess the system's accuracy and real-world applicability in skin cancer diagnosis.
Main Methods:
- A deep learning system was trained on 5,070 images from one laboratory.
- The system was tested on 13,537 uncurated images from three different laboratories using various whole slide scanners.
- Confidence scoring was implemented to evaluate classification accuracy.
Main Results:
- The deep learning system achieved up to 98% accuracy when utilizing deep-learning-based confidence scoring.
- Without confidence scoring, the system's accuracy was 78%.
- The system demonstrated adaptability across different labs and scanner vendors.
Conclusions:
- The validated deep learning system offers a promising foundation for accelerating skin cancer diagnosis.
- The AI tool can enhance diagnostic reproducibility and efficiency in dermatopathology.
- It can aid in identifying cases for specialist review and support targeted diagnostic classifications.
