Ultrasonography Training System

Abstract

---

Ultrasonography is widely used for screening of traumas or disorders of internal organs noninvasively, while its diagnostic accuracy depends on skills of physicians. Therefore, trainings of inexperienced physicians are important. The purpose of this study is to propose a novel training system for ultrasonography by integrating simulators (ultrasound phantom) and image processing techniques.

This system can provide feedback to the trainee based on recognition of target organs in ultrasound images after role-plays of ultrasound examinations on the simulator. An experiment using simulator for FAST (Focused Assessment with Sonography for Trauma) was conducted and this algorithm could recognize the target organs with the 96-100% recognition rates.

We conclude that proposed method could be useful not only for evaluations after the training but also for real-time feedback during the each training trails.

Algorythm

---

The organ recognition algorithm uses some registered images, the sample images of the target organs which should be scanned in the simulation trainings. This algorithm recognizes the target organs in ultrasound images matching the input images and the registered images.

This algorithm detects some contours (we call these keypoints) in the input images and the registered images, then this algorithm matches similar keypoints between the input and registered image. When the target organ is scanned correctly, the matched keypoints in the input image are densely. Therefore, the target organs can be recognized according to the density of the matched keypoints.

Demonstration

---