This article provides layman’s explanations for terms involved:

• Standard Plane——For the ultrasound imaging of a certain organ, it needs to include specified structures and parts, such as a standard cardiac plane should include the left atrium, left ventricle, right atrium, right ventricle, and aorta.
• Image Segmentation——Also called semantic segmentation. An image is given to the neural network (we use BiSeNet, originally used for remote sensing images, but we have modified it), and it will mark the position of each target in the image.

Overview

The core functionalities of the entire quality control system are two-fold:

1. Standard Plane Recognition (to be completed soon)
2. Segmentation of images and measuring distance (completed)

System Interface

The large image in the middle is the input original image. The upper left shows the results of the plane recognition, displaying the identified sections of the plane. For example, in a cardiac standard plane, 5 boxes should be used to mark the corresponding structures. The middle left shows the results of image segmentation and distance measurement. The large colored areas are the results of semantic segmentation, through which we can obtain complete morphological information about the cardiac structure. The yellow boxes are based on the semantic segmentation for distance measurement. The upper right shows the recognized plane information, based on whether the structure is complete, it can be judged whether it is a standard plane. The lower right is the distance measurement result based on semantic segmentation, which can be compared and scored against the input distance measurement information. System Demonstration Video

System Usage

• Assist in training ultrasound examination doctors
• Help outpatient doctors screen positive cases
• Embedded in ultrasound examination equipment, identify standard planes in real-time
• ….

Future Prospects

• Batch import pictures and doctors’ distance measurement information
• Expand the types of planes, including thoracic cavity, abdominal cavity, sternum, blood vessels…
• OCR recognition of information marked by doctors and generate inspection reports
• …