Oral squamous cell carcinoma detection using EfficientNet on histopathological images.

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Oral Squamous Cell Carcinoma (OSCC) diagnosis is challenged by imprecise tools, leading to delayed detection.
• Current diagnostic methods for OSCC lack optimal accuracy and efficiency.
• There is a critical need for advanced, reliable diagnostic approaches for OSCC.

Purpose of the Study:

• To explore the discriminative potential of histopathological images for OSCC detection.
• To develop and evaluate a deep learning model for differentiating normal oral epithelium from OSCC tissues.
• To assess the efficacy of a customized EfficientNetB3 model in OSCC diagnosis.

Main Methods:

• Utilized a database of 1224 histopathological images from 230 patients with varying magnifications.
• Developed a customized deep learning model based on the EfficientNetB3 architecture.
• Employed data augmentation, regularization, and optimization techniques during model training.

Main Results:

• The deep learning model achieved 99% accuracy in differentiating normal epithelium from OSCC tissues.
• The model demonstrated high precision, recall, and F1-score metrics.
• The results indicate the model's robustness as a potential diagnostic tool for OSCC.

Conclusions:

• Deep learning models show significant promise for overcoming OSCC diagnostic challenges.
• The developed model represents a substantial advancement in early and accurate OSCC detection.
• Machine learning techniques can improve patient outcomes through timely and precise OSCC identification.