Detection of Esophageal Wall via Deep Learning-Based Object Detection with Ultrasound System

Conventional methods of assessing swallowing function are invasive and burdensome to the patient. Although swallowing function assessment can be performed non-invasively using an ultrasound system, one issue must be overcome: the position of the probe is not fixed. Therefore, deep learning-based object detection was used to detect the esophageal wall so that a more stable and accurate examination can be performed. As a result,. Although the detection rate was high enough and false positive rate was low enough for practical use, we believe that the detection rate can be further increased in the future by learning ultrasound images taken from various angles from a larger number of subjects.