Automatic measurement of pharyngeal contraction ratio during deglutition using 2D cine MRI with deep learning: A pilot study.
Authors
Affiliations (3)
Affiliations (3)
- Department of Radiological Sciences, Ibaraki Prefectural University of Health Sciences, 4669-2 Ami, Amimachi, Ibaraki, 300-0394, Japan. [email protected].
- Department of Radiological Sciences, Ibaraki Prefectural University of Health Sciences, 4669-2 Ami, Amimachi, Ibaraki, 300-0394, Japan.
- Department of Radiological Technology, Faculty of Medical Technology, Niigata University of Health and Welfare, Niigata, Japan.
Abstract
This study aimed to develop a deep learning-based method for automatic segmentation of the pharyngeal area (PA) and measurement of the pharyngeal contraction ratio (PCR) during deglutition using cine magnetic resonance imaging (MRI). The proposed algorithm combines PA region extraction by a 2D U-Net with automatic calculation of PA and PCR. Segmentation performance was evaluated using the Dice coefficient (DC), and the PCR measured by the model ([Formula: see text]) was compared with that obtained manually ([Formula: see text]) using correlation and Bland-Altman analyses. Cine MRI data of 20 healthy adults (10 men, 10 women; age 22-29 years) were analyzed. The average DC in the test cases was 0.890 ± 0.025, and the PA of the model correlated well with the manual reference (r = 0.70-0.97). The mean [Formula: see text] was 0.105 ± 0.035, consistent with values reported in videofluoroscopic swallowing studies. These results demonstrate the technical feasibility of automatic PCR measurement from cine MRI using deep learning.