# Protocol for detecting oral squamous cell carcinoma in histopathology images using the momentum contrast framework

## TL;DR

This paper presents a protocol using machine learning to detect oral cancer in tissue images, aiming to improve diagnosis accuracy.

## Contribution

A novel protocol for OSCC detection using transfer learning and the MoCo framework with histopathology images.

## Key findings

- Pretraining with diverse datasets improves model performance for OSCC detection.
- Fine-tuning with OSCC-specific data enhances classification accuracy.
- Augmentation techniques increase model robustness.

## Abstract

The detection of oral squamous cell carcinoma (OSCC) in histopathology images is crucial for improving diagnostic accuracy and patient outcomes. Here, we present a protocol for detecting OSCC in histopathology images using transfer learning. We describe steps for installing software and prerequisites, preparing datasets, and pretraining a model on images from various tissue types using the momentum contrast (MoCo) framework. We then detail procedures for evaluating the fine-tuned HistoMOCO model’s performance on a test dataset.

•Pretrain the model with diverse histopathology datasets using MoCo for OSCC detection•Fine-tune the model with OSCC-specific datasets to improve image classification•Use rotation, scaling, and augmentation techniques to enhance model robustness

Pretrain the model with diverse histopathology datasets using MoCo for OSCC detection

Fine-tune the model with OSCC-specific datasets to improve image classification

Use rotation, scaling, and augmentation techniques to enhance model robustness

Publisher’s note: Undertaking any experimental protocol requires adherence to local institutional guidelines for laboratory safety and ethics.

The detection of oral squamous cell carcinoma (OSCC) in histopathology images is crucial for improving diagnostic accuracy and patient outcomes. Here, we present a protocol for detecting OSCC in histopathology images using transfer learning. We describe steps for installing software and prerequisites, preparing datasets, and pretraining a model on images from various tissue types using the momentum contrast (MoCo) framework. We then detail procedures for evaluating the fine-tuned HistoMOCO model’s performance on a test dataset.