Computer-Aided Detection System for the Classification of Non-Small Cell Lung Lesions using SVM

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Oncology

Background:

• Lung carcinoma is a leading cause of cancer deaths globally, primarily linked to smoking.
• Non-small cell lung cancer (NSCLC) encompasses major subtypes like Adenocarcinomas, Squamous cell carcinomas, and large cell carcinomas.
• Accurate detection and classification of lung cancer are crucial for effective treatment.

Purpose of the Study:

• To develop and evaluate a Computer Aided Detection (CADe) system for classifying three subtypes of NSCLC.
• To enhance diagnostic convenience for radiologists using ultrasonic images.
• To achieve precise and superior classification results for lung cancer.

Main Methods:

• The study focused on a three-class classification of Adenocarcinomas, Squamous cell carcinomas, and large cell carcinomas within NSCLC.
• A CADe system was designed to analyze image quality, select regions of interest, preprocess data, extract features, and classify cancer.
• Laws' mask features and a Support Vector Machine (SVM) classifier with Gaussian RBF kernels were employed.

Main Results:

• Experimentation was conducted on 92 images with a 50%-50% training and testing split.
• The developed system achieved an average accuracy of 95.65% for classifying the three NSCLC subtypes.
• This accuracy was found to be the highest compared to existing methods using the same dataset.

Conclusions:

• The proposed CADe system demonstrates high efficacy in differentiating between Adenocarcinomas, Squamous cell carcinomas, and large cell carcinomas.
• The combination of Laws' mask features and SVM classifier provides a robust approach for lung cancer classification.
• The system offers a significant advancement in computer-aided diagnosis for NSCLC, aiding radiologists in accurate and efficient diagnosis.