Rapid and Robust Automated MRI Dixon Water and Fat Image Classification

Motivation: In clinical MRI, the Dixon techniques is widely used for simultaneously obtaining water-fat images. However, it often encounters water-fat misclassification, where water is incorrectly identified as fat and vice versa. Goal(s): This paper proposes a fast and robust automatic Dixon water-fat image classification method. Approach: Background noise is suppressed using non-local means filter, followed by extraction of tissue edge information. Adipose-rich ROIs are then segmented through morphological operations. Within these ROIs, the maximum inter-class variance between water and fat is computed. Results: The results from datasets of foot and hand datasets demonstrate that proposed method achieves precise and efficient classification of water-fat images. Impact: Misclassification of water-fat in clinical MRI examinations can lead to diagnostic ambiguity, potentially impacting clinical decision-making and increasing the risk of medical disputes. This paper proposes a rapid and robust automated water-fat image classification method to address this issue.