Related Experiment Video
Updated: May 23, 2025
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
1.7K
Binary Classification of Laryngeal Images Utilising ResNet-50 CNN Architecture.
Rakesh Srivastava1,2, Nitish Kumar2, Tushar Sandhan2
1Sushrut Institute of Plastic Surgery & Super-speciality Hospital, Lucknow/Raj ENT Centre, 3/387, Vishal Khand-3, Gomtinagar, Lucknow, India.
Summary
This study developed a deep learning model using endoscopy images for early laryngeal cancer detection in India. The advanced convolutional neural network approach shows high accuracy in classifying cancerous and non-cancerous lesions.
Area of Science:
- Oncology
- Medical Imaging
- Artificial Intelligence
Background:
- Laryngeal cancer presents a significant health challenge in India, necessitating improved early detection strategies.
- Current diagnostic methods may lack accessibility or require specialized expertise, particularly in regions with high head and neck cancer incidence.
Purpose of the Study:
- To develop and evaluate a novel deep learning model for classifying laryngeal lesions from endoscopy images.
- To differentiate between cancerous (Squamous Cell Carcinoma) and non-cancerous laryngeal lesions using advanced image analysis techniques.
Main Methods:
- A dataset of 1978 endoscopy images from 960 patients was utilized, employing convolutional neural networks and image processing.
- ResNet50 was used for feature extraction, with the dataset split into 90% for training/validation and 10% for testing.
- The model's performance was assessed using Receiver Operating Characteristic (ROC) curves and accuracy metrics.
Main Results:
- The deep learning model demonstrated high effectiveness in classifying laryngeal lesions.
- Areas under the ROC curve were 0.95 (combined), 0.98 (NBI-only), and 0.93 (WL-only), indicating strong diagnostic potential.
- High accuracy rates across all tested datasets suggest the model's utility for early cancer detection.
Conclusions:
- The developed model shows significant promise for aiding in the early detection and classification of laryngeal cancer, especially in resource-limited settings.
- This approach could be instrumental in improving diagnostic capabilities for laryngopharyngeal cancer in India, addressing a critical unmet need.
