Deep learning based diagnosis of PTSD using 3D-CNN and resting-state fMRI data
Mirza Naveed Shahzad1, Haider Ali1
1Department of Statistics, University of Gujrat, Gujrat, Pakistan.
Psychiatry Research. Neuroimaging
|June 22, 2024
Summary
Accurate detection of Posttraumatic stress disorder (PTSD) is crucial. A 3D-CNN model using resting-state fMRI data achieved high accuracy in classifying PTSD patients, outperforming other machine learning methods.
Area of Science:
- Neuroimaging
- Machine Learning
- Psychiatry
Background:
- Posttraumatic stress disorder (PTSD) incidence is rising globally.
- Accurate PTSD detection is vital for patient treatment.
- This study aims to differentiate individuals with PTSD from healthy controls.
Purpose of the Study:
- To classify individuals with Posttraumatic stress disorder (PTSD) versus healthy controls.
- To evaluate the effectiveness of various machine learning techniques for PTSD classification.
- To identify key brain regions involved in PTSD.
Main Methods:
- Resting-state functional MRI (rs-fMRI) data from 19 PTSD patients and 24 healthy controls were analyzed.
- Group-level independent component analysis (ICA) and t-tests identified brain activation patterns.
- Six machine learning methods, including 3D-CNN, were employed for classification.
Main Results:
- Amygdala and insula showed the highest activation in PTSD subjects.
- Initial machine learning models using ICA components yielded low accuracy.
- The 3D-CNN model achieved high classification accuracies (98.12% training, 98.25% validation, 98.00% testing).
Conclusions:
- The 3D-CNN model significantly outperforms six other machine learning techniques for PTSD classification.
- This deep learning approach offers a promising tool for accurate PTSD patient recognition.
