Precise segmentation technology for digital radiography

AI-based image processing supports X-ray photography reduces retakes

Helping to improve the radiographic efficiency and reduce the burden on the patient

Diagnostic lateral X-ray images of joints like the knees, ankles, etc. require accurate imaging straight from the side of the joint without displacement of the joint. To obtain such accurate X-ray images, however, is complicated, and since the joint may easily be displaced, it is sometimes difficult to judge whether an image is suitable to reach a diagnosis. To solve such problems, Konica Minolta has developed a positioning judgment function that measures the displacement of joints immediately after X-ray imaging using precise segmentation technology, and classifies the extent of the measured displacement into the relevant level of acceptability, and displays the level. If the displacement is judged to exceed the acceptable limit, an alert can be displayed to indicate the need for re-imaging. Confusion between the left and right side of an X-rayed site may be a simple mistake but it carries a huge medical safety risk. To prevent such mistakes, Konica Minolta has developed a left/right judgment function that infers the left and right side of the joint immediately after taking the X-ray image using subject prediction technology. If it is judged that the image was not taken from the side instructed by the physician, the system can display an alert. These functions allow the radiological technologist to promptly decide whether to take the X-ray image again based on objective information provided immediately after radiography. This can avoid unnecessary repeat exposure and having patients return for re-examination, improve radiographic efficiency, and shorten the time of radiographic testing, thereby contributing to reduce the burden on the patient and the radiation dose.

Technology Overview

The positioning judgment function identifies the area and extent of displacement of the joint on the obtained image. The function (1) identifies the local area of the joint on the image based on deep learning, (2) performs segmentation processing of the identified local area based on deep learning to identify the displaced area, and then (3) measures the distance of the segmented area from the center of the joint to determine the maximally displaced width. The left/right judgment function firstly (1) performs pretreatment to change the image size and adjust the gradation and density of the captured image, then (2) estimates the probability that the image is that of the left or the right side of the joint based on deep learning, and then (3) determines the side with the higher probability as the side of the joint of which the image was actually taken. These functions involve high-speed processing because the results have to be displayed immediately after the X-rays have been taken.