Automatic Identification of Brain Injury Mechanism Based on Deep Learning.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Forensic Science

Background:

• Head injuries from acceleration and deceleration forces are common in forensic investigations.
• Accurate identification of injury mechanisms is crucial for determining the cause of death or injury.
• Current methods for analyzing brain injuries may benefit from advanced computational techniques.

Purpose of the Study:

• To evaluate the effectiveness of the Inception_v3 convolutional neural network (CNN) model for automatically identifying acceleration and deceleration brain injuries using CT images.
• To explore the potential of deep learning in forensic inference of head injury mechanisms.

Main Methods:

• A dataset of 320 brain CT scans (190 injury cases, 130 normal controls) was utilized.
• The data was randomly divided into training, validation, and testing sets.
• Model performance was assessed using accuracy, precision, recall, F1-score, and AUC.

Main Results:

• The Inception_v3 model achieved high accuracy rates during training (99.00%) and validation (87.21%).
• In the testing set, the model demonstrated an overall accuracy of 87.18%.
• Specific performance metrics for recognizing acceleration injury, deceleration injury, and normal brains were reported, with AUC values ranging from 0.92 to 0.98.

Conclusions:

• The Inception_v3 model shows significant potential for distinguishing between acceleration and deceleration brain injuries on CT scans.
• This deep learning model could serve as a valuable auxiliary tool in forensic investigations for inferring head injury mechanisms.