Related Experiment Video
Updated: Nov 24, 2025
Author Spotlight: Bridging Gaps in Anatomy and Establishing a Foundation for Algorithmic Studies
Published on: December 15, 2023
Segmentation of Chronic Subdural Hematomas Using 3D Convolutional Neural Networks
Ryan T Kellogg1, Jan Vargas2, Guilherme Barros1
1Department of Neurological Surgery, University of Washington, Seattle, Washington, USA.
Insights
An automated program was developed to calculate chronic subdural hematoma (cSDH) volumes from CT scans. This AI tool accurately measures hematoma size, aiding in treatment evaluation for this common neurological condition.
Area of Science:
- Neurosurgery
- Radiology
- Artificial Intelligence
Background:
- Chronic subdural hematomas (cSDHs) are a growing neurological concern requiring surgical intervention.
- Computed tomography (CT) is crucial for managing cSDHs, with serial imaging guiding treatment.
- Hematoma volume is key for therapy decisions and evaluating new management strategies.
Purpose of the Study:
- To develop an automated program for calculating cSDH volume from CT scans.
- To enable precise pre- and postoperative hematoma volume measurements.
- To create a tool for efficient evaluation of cSDH treatment efficacy.
Main Methods:
- A convolutional neural network (CNN) was trained using 21,710 images from 128 CT scans.
- The dataset included pre- and postoperative coronal head CTs of patients with cSDHs.
- Manual segmentation was performed to train the automated cSDH segmentation model.
Main Results:
- The best CNN model achieved a DICE score of 0.8351 on the test dataset.
- An average DICE score of 0.806 ± 0.06 was obtained on the validation set.
- Model performance was optimized with specific network depth and residual block configurations.
Conclusions:
- A CNN was successfully trained for automated cSDH segmentation on CT scans.
- This automated tool can provide accurate measurements for assessing treatment effectiveness.
- The developed AI assists in managing this prevalent neurological disease.
Objective:
Chronic subdural hematomas (cSDHs) are an increasingly prevalent neurologic disease that often requires surgical intervention to alleviate compression of the brain. Management of cSDHs relies heavily on computed tomography (CT) imaging, and serial imaging is frequently obtained to help direct management. The volume of hematoma provides critical information in guiding therapy and evaluating new methods of management. We set out to develop an automated program to compute the volume of hematoma on CT scans for both pre- and postoperative images.
Methods:
A total of 21,710 images (128 CT scans) were manually segmented and used to train a convolutional neural network to automatically segment cSDHs. We included both pre- and postoperative coronal head CTs from patients undergoing surgical management of cSDHs.
Results:
Our best model achieved a DICE score of 0.8351 on the testing dataset, and an average DICE score of 0.806 ± 0.06 on the validation set. This model was trained on the full dataset with reduced volumes, a network depth of 4, and postactivation residual blocks within the context modules of the encoder pathway. Patch trained models did not perform as well and decreasing the network depth from 5 to 4 did not appear to significantly improve performance.
Conclusions:
We successfully trained a convolutional neural network on a dataset of pre- and postoperative head CTs containing cSDH. This tool could assist with automated, accurate measurements for evaluating treatment efficacy.
