A Basic Study for Predicting Dysphagia in Panoramic X-ray Images Using Artificial Intelligence (AI)—Part 1: Determining Evaluation Factors and Cutoff Levels

Abstract

Background: Dysphagia relates to quality of life; this disorder is related to the difficulties of dental treatment. Purpose: To detect radiographic signs of dysphagia by using panoramic radiograph with an AI system. Methods: Seventy-seven patients who underwent a panoramic radiograph and a videofluorographic swallowing study were analyzed. Age, gender, the number of remaining teeth, the distance between the tongue and the palate, the vertical and horizontal hyoid bone position, and the width of the tongue were analyzed. Logistic regression analysis was used. For the statistically significant factors, the cutoff level was determined. The cutoff level was determined by using analysis of the receiver operations characteristic (ROC) curve and the Youden Index. Results: A significant relationship with presence of dysphagia was only observed for the vertical hyoid bone position. The area under the curve (AUC) was 0.72. The cutoff level decided for the hyoid bone was observed to be lower than the mandibular border line. Conclusions: In cases where the hyoid bone is lower than the mandibular border line on a panoramic radiograph, it suggests the risk of dysphagia would be high. We will create an AI model for the detection of the risk of dysphagia by using the assessment of vertical hyoid bone position.