Histopathology-Based Diagnosis of Oral Squamous Cell Carcinoma Using Deep Learning

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Oral squamous cell carcinoma (OSCC) presents a global health challenge with a generally poor prognosis.
• Current OSCC diagnosis relies on pathologist expertise in interpreting histopathology images, which can be time-consuming and prone to human error.
• Deep learning (DL) offers potential to enhance the accuracy and efficiency of medical image analysis.

Purpose of the Study:

• To develop and evaluate a custom deep learning model for the automated detection of OSCC in histopathology images.
• To assess the impact of the DL model on the diagnostic performance (accuracy and speed) of pathologists.

Main Methods:

• A custom deep learning model was trained on 1,925 histopathology images of OSCC.
• The model was tested on 100 independent images, evaluating its diagnostic accuracy using sensitivity, specificity, and F1 score.
• The model's assistance was evaluated with junior and senior pathologists on a subset of 100 images.

Main Results:

• The DL model achieved high diagnostic performance with a sensitivity of 0.98, specificity of 0.92, and F1 score of 0.951.
• Pathologists assisted by the model showed improved diagnostic accuracy, with average F1 scores increasing from 0.9221 to 0.9566 (junior) and 0.9361 to 0.9463 (senior).
• Model assistance reduced diagnostic time for junior pathologists by an average of 6.26 minutes per case.

Conclusions:

• Deep learning models can significantly improve the accuracy of OSCC detection from histopathology images.
• AI-assisted diagnosis accelerates the interpretation process and enhances diagnostic performance for pathologists.
• This technology holds promise for reducing diagnostic errors and improving patient outcomes in OSCC management.