Related Experiment Video
Updated: Oct 29, 2025
05:56
Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application
Published on: April 14, 2023
2.8K
Deep Learning for Differentiating Benign From Malignant Parotid Lesions on MR Images
Xianwu Xia1,2,3, Bin Feng1,2, Jiazhou Wang1,2
1Department of Radiation Oncology, Fudan University Shanghai Cancer Center, Shanghai, China.
Frontiers in Oncology
|July 12, 2021
Summary
This study developed a deep learning network to diagnose parotid gland tumors using MRI scans. The model shows promise in assisting clinicians with accurate tumor diagnosis.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Salivary gland tumors are rare and histologically diverse.
- Accurate distinction between benign and malignant parotid gland tumors is crucial for patient management.
Purpose of the Study:
- To develop and evaluate a deep learning (DL) network for diagnosing parotid gland tumors.
- To leverage deep learning on MR images for improved diagnostic accuracy.
Main Methods:
- A modified ResNet model was trained on 3791 segmented MR images from 233 patients.
- Data included T1-weighted, CE-T1-weighted, and T2-weighted MRI series.
- The model was trained using PyTorch, employing data augmentation and the Adam optimizer.
Main Results:
- The DL model achieved 92.94% accuracy on the training dataset with a micro-AUC of 0.98.
- The final algorithm demonstrated 82.18% accuracy in diagnosing and staging parotid cancer (micro-AUC 0.93).
Conclusions:
- The proposed deep learning model can aid clinicians in diagnosing parotid tumors.
- Further large-scale, multicenter studies are necessary for comprehensive validation.
