DL for Prediction of Malignant Transformation in Oral Epithelial Dysplasia in Slide Images

Deep Learning Prediction of Malignant Transformation in Oral Epithelial Dysplasia Using Whole Slide Images

Authors:

Raja Muhammad Saad Bashir; Adam J. Shephard; Hanya Mahmood; Neda Azarmehr; Shan E Ahmed Raza; Syed Ali Khurram; Nasir M. Rajpoot

Summary:

Study investigating weakly supervised deep learning methods for predicting malignant transformation in oral epithelial dysplasia (OED) using whole-slide histology images (WSIs). A cohort of 163 WSIs (137 OED cases, 50 with malignant transformation) was analysed. A weakly supervised pipeline using IDaRS (Iterative Draw-and-Rank Sampling) with ResNet-34 achieved the highest performance (AUROC = 0.78; F1 = 0.69). Hotspot analysis identified peri-epithelial lymphocyte (PEL) count, epithelium layer nuclei count, and basal layer nuclei count as significant predictors. Survival analysis showed that PELs and epithelial/basal layer nuclear features improve prognostic stratification. The study demonstrates that deep learning can predict malignant transformation and progression-free survival, offering potential support for clinical risk assessment.

Check the original publication:

Key Words:

OED; Oral Dysplasia; Histology; WSI; Deep Learning; Weak Supervision; IDaRS; Prognostic Model

Extracted Data:

Year: 2023

Modality: Histology Whole Slide Images (WSI)

Dataset: 163 WSIs (137 OED cases; 50 with malignant transformation)

Dataset Split: Stratified 5-fold cross-validation (3 runs, different seeds)

Network Architecture: ResNet-34, ResNet-50, DenseNet

Metrics: AUROC 0.78 (±0.07); F1 Score 0.69 (±0.05)

AP - Professional Qty: 2 certified pathologists + 1 additional blinded expert

AP - Supervisor Presence: No information

AP - Experience Level: Not explicitly quantified (consultant pathologists)

AP - Expertise Area: Oral & Maxillofacial Pathology

AP - Tool or System: No information

ML Task: Classification and Segmentation

Project Objective: To predict malignant transformation in oral epithelial dysplasia using deep learning on WSIs and to identify nuclear features associated with progression-free survival.

Clinical Relevance:

Provides early prognostic insights for OED, enabling improved patient stratification and potentially better clinical decision-making for preventing malignant transformation.

Watch full video on YouTube