An intensity-based self-supervised domain adaptation method for intervertebral disc segmentation in magnetic...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Spinal Diagnostics

Background:

• Accurate intervertebral disc (IVD) segmentation is vital for diagnosing spinal conditions.
• Traditional deep learning requires extensive annotated data, which is difficult to obtain.
• This limits the application of deep learning in clinical settings.

Purpose of the Study:

• To develop an intensity-based self-supervised domain adaptation method for IVD segmentation.
• To reduce the reliance on large annotated datasets for IVD segmentation.
• To improve the generalizability of IVD segmentation models across different data domains.

Main Methods:

• An intensity-based self-supervised learning approach was developed for IVD segmentation in MRI scans.
• A dual-task model simultaneously segments IVDs and predicts intensity transformations.
• The model was trained on unlabeled multi-domain data to learn domain-invariant features.

Main Results:

• The proposed model outperformed baseline models on three public datasets.
• The method demonstrated superior performance in handling domain shifts.
• Higher accuracy in IVD segmentation was achieved compared to single-domain trained models.

Conclusions:

• Intensity-based self-supervised domain adaptation shows significant potential for IVD segmentation.
• The approach enhances model generalizability across datasets with domain shifts.
• This methodology can be extended to other medical imaging segmentation tasks.