A deep learning approach to predict temporal changes of subdural hemorrhage on computed tomography

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Subdural hemorrhage (SDH) requires timely assessment of its progression via computed tomography (CT).
• Estimating hemorrhage age is crucial for managing SDH, but can be challenging.
• Current methods may not fully leverage CT data for precise age estimation.

Purpose of the Study:

• To develop a deep learning model for predicting temporal changes in subdural hemorrhage.
• To utilize Hounsfield Units (HU) to estimate hemorrhage age across acute, subacute, and chronic stages.
• To enhance diagnostic efficiency and support clinical workflows in SDH management.

Main Methods:

• A convolutional neural network (CNN) was developed using Python on Google Colab.
• 825 pre-processed CT slices from the RSNA dataset were analyzed.
• The model was trained and validated using metrics including accuracy, sensitivity, specificity, precision, F1-score, and AUC-ROC.

Main Results:

• The model achieved 85.33% prediction accuracy.
• High sensitivity and specificity were reported across acute (86.67%, 94%), subacute (84%, 88%), and chronic (85.33%, 96%) SDH stages.
• AUC-ROC values ranged from 0.9394 to 0.9731, indicating robust classification performance.

Conclusions:

• The deep learning model shows significant potential for predicting subdural hemorrhage age.
• The model can serve as a valuable second-reader tool for radiologists.
• Implementation can streamline emergency triage and improve diagnostic efficiency in clinical settings.