23 November 2023 in Biomedical Physics & Engineering Express
Anatomical segmentations generated using artificial intelligence (AI) have the potential to significantly improve video fluoroscopic swallow study (VFS) analysis. AI segments allow for various metrics to be determined without additional time constraints streamlining and creating new opportunities for analysis. While the opportunity is vast, it is important to understand the challenges and limitations of the underlying AI task. This work evaluates a bolus segmentation network. The first swallow of thin or liquid bolus from 80 unique patients were manually contoured from bolus first seen in the oral cavity to end of swallow motion. The data was split into a 75/25 training and validation set and a 4-fold cross validation was done. A U—Net architecture along with variations were tested with the dice coefficient as the loss function and overall performance metric. The average validation set resulted in a dice coefficient of 0.67. Additional analysis to characterize the variability of images and performance on sub intervals was conducted indicating high variability among the processes required for training the network. It was found that bolus in the oral cavity consistently degrades performance due to misclassification of teeth and unimportant residue. The dice coefficients dependence on structure size can have substantial effects on the reported value. This work shows the efficacy of bolus segmentation and identifies key areas that are detriments to the performance of the network.
29 January 2025 in Current Physical Medicine and Rehabilitation Reports
23 October 2025 in Dysphagia
The Dynamic Imaging Grade of Swallowing Toxicity (DIGEST) is a psychometric tool used during modified barium swallow studies (MBSSs) to grade swallowing safety, efficiency, and overall pharyngeal swallow function ranging from 0 (no impairment) to 4 (life-threatening). DIGEST was originally validated on a protocol without routine inclusion of thickened liquids. To ensure stability of DIGEST in different settings, we aimed to assess whether DIGEST grades remained stable when derived from a bolus protocol with or without standard inclusion of thickened liquids. MBSSs from 118 unique participants were retrospectively analyzed by 4 reliable raters using a master protocol of thin, thickened, pudding, and solid boluses. DIGEST grades were derived from four bolus protocol conditions: (1) DIGEST core protocol (without thickened liquids); (2) core plus mildly thick liquid; (3) core plus moderately thick; and (4) core plus mildly and moderately thick. Prevalence and Bias-Adjusted Kappa (PABAK) values were calculated to determine agreement of DIGEST grades between the core protocol and variations. PABAK ranged from 0.83 to 1.00 (near-perfect to perfect agreement). In rare instances where thickened liquids affected grades, they invariably worsened by one grade. There was no clear change-driving bolus type for this effect. Safety was more susceptible to change than efficiency. Inclusion of thickened liquids does not appear necessary in a minimum bolus protocol for DIGEST. Clinicians and researchers who wish to routinely include thickened liquids in their protocol should be aware that DIGEST grades may be worsened by one in a minority of cases and that safety grades appear more likely to be affected than efficiency grades.