Accelerated diffusion tensor imaging with self-supervision and fine-tuning.

Area of Science:

• Neuroimaging
• Medical Imaging Analysis
• Artificial Intelligence in Medicine

Background:

• Diffusion Tensor Imaging (DTI) is crucial for evaluating brain microstructure.
• Long DTI acquisition times limit its clinical utility.
• Existing deep learning (DL) methods require large datasets for effective training.

Purpose of the Study:

• To develop a novel framework, Self-Supervised Deep Learning with Fine-Tuning (SSDLFT), to decrease DTI training data requirements.
• To enable high-performance DTI analysis with reduced data demands.

Main Methods:

• Implemented a self-supervised pretraining phase for data denoising without requiring clean labels.
• Utilized a fine-tuning phase with limited high-quality DTI data.
• Validated the SSDLFT framework using data from the Human Connectome Project.

Main Results:

• SSDLFT demonstrated superior performance compared to traditional methods and other DL approaches in Diffusion Weighted Imaging (DWI) reconstructions and tensor metrics.
• The framework maintained high accuracy with fewer training subjects and DWIs.
• Qualitative and quantitative assessments confirmed the effectiveness of SSDLFT.

Conclusions:

• SSDLFT significantly reduces the data requirements for training deep learning models in DTI.
• This advancement enhances the practical applicability of DTI in both clinical and research settings.
• The proposed method offers a more efficient approach to DTI analysis.