Intracerebral hemorrhage detection on computed tomography images using a residual neural network

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neurology

Background:

• Intracerebral hemorrhage (ICH) is a severe condition with high mortality, necessitating rapid diagnosis.
• Deep learning offers advanced capabilities for analyzing medical images to detect diseases like ICH.

Purpose of the Study:

• To develop and evaluate a deep learning model (ResNet-18) for differentiating CT scans of healthy brains from those with ICH.
• To enhance the interpretability of the model's decisions using Grad-CAM visualization.

Main Methods:

• Utilized a ResNet-18 deep residual convolutional neural network for image classification.
• Employed Gradient-weighted Class Activation Mapping (Grad-CAM) for visual explanations.
• Performed 100-iteration Monte Carlo cross-validation with an 80/20 train/test split on 200 CT brain images.

Main Results:

• Achieved an average accuracy of 95.93%, specificity of 96.20%, sensitivity of 95.65%, precision of 96.40%, and F1-score of 95.91%.
• Demonstrated efficient performance with an average training time of 165.90 seconds and testing time of 1.17 seconds.
• Results are comparable to state-of-the-art methods with a simpler, less computationally intensive approach.

Conclusions:

• The developed deep learning detector shows high efficacy in identifying ICH from CT images.
• The model can serve as a valuable tool for physicians, improving diagnostic speed, accuracy, and resource allocation.
• This approach offers a computationally efficient method for ICH detection, supporting clinical decision-making.