LesionScanNet: dual-path convolutional neural network for acute appendicitis diagnosis

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer-Aided Diagnosis

Background:

• Acute appendicitis presents with abdominal pain, vomiting, and fever.
• Computed tomography (CT) aids diagnosis but faces challenges with anatomical variations.
• Accurate and efficient appendicitis detection remains crucial in clinical practice.

Purpose of the Study:

• To introduce LesionScanNet, a novel Convolutional Neural Network (CNN) model.
• To develop a computer-aided detection (CAD) system for acute appendicitis using CT images.
• To evaluate the performance and generalizability of the proposed model.

Main Methods:

• A dataset of 2400 CT images was utilized.
• LesionScanNet, a lightweight CNN with DualKernel blocks (3x3 and 1x1 filters), was developed.
• The model's performance was benchmarked against other deep learning models.

Main Results:

• LesionScanNet achieved a 99% accuracy score on the test dataset.
• The model demonstrated superior performance compared to benchmark deep learning models.
• The model showed generalization capabilities on chest X-ray datasets for pneumonia and COVID-19 detection.

Conclusions:

• LesionScanNet is a lightweight and robust deep learning network for medical image analysis.
• The model achieves superior performance with a reduced number of parameters.
• Its application can be extended to diverse medical imaging domains beyond appendicitis detection.