Penza State University (PSU) presented a development for medical universities — an interactive simulator that teaches how to recognize pathologies on X-rays even with low image quality. Certificates of registration of computer programs have been obtained for the invention.
The quality of X-ray images plays a key role in the accurate diagnosis of a wide range of diseases, including tuberculosis, pneumonia, oncological pathologies and traumatic injuries. The clarity and contrast of the visualization of anatomical structures directly determine how timely the pathological changes will be detected and treatment initiated. However, in practice, the quality of the images is not always perfect. It is affected by the condition of the equipment, exposure parameters, radiation dose, quantum noise level, patient's position, immobility, the presence of foreign objects in the frame, as well as the features of image post-processing.
"We need competent X—ray technicians and radiologists who are able to evaluate the quality of images and understand which defects can distort the clinical picture. And to do this, they must have enough visibility," said Leonid Krivonogov, one of the developers, professor of the Department of Medical Cybernetics and Computer Science at PSU, Doctor of Technical Sciences.
The simulator allows future doctors not only to study typical defects, but also to master objective quantitative criteria for evaluating image quality, visually linking visual manifestations of distortions with their physical causes.
The system is built as a modular architecture and consists of three interrelated components: a database, a subsystem for calculating metrics, and a subsystem for training and testing. The database provides centralized storage of X-ray images, calculated quality metrics, and related information. The metrics calculation subsystem is responsible for an objective assessment of the quality of images: it uploads images, calculates a set of quantitative indicators, builds graphs, compares images with benchmarks, and generates reports.
"The subsystem evaluates the ratio of the useful signal to the noise level, analyzes image gradients, the difference between maximum and minimum brightness, and many others. All metrics are calculated directly from pixel data using rigorous mathematical formulas. This guarantees reproducibility and objectivity of the results," Leonid Krivonogov explained.
The training and testing subsystem generates training tasks, providing students with distorted versions of reference images. The process of training and testing skills is implemented through a web interface. A browser and Internet access are sufficient for operation, without installing additional software. Students see the final scores and details for each assignment, including correct and erroneous answers. This allows you to identify typical errors and adjust the training program.
Using the developed simulator, scientists evaluated 200 test cases. Based on five reference images — chest, limbs, abdominal cavity — 10 distorted copies were generated for each of the four types of defects. According to the developers, the system correctly classified the type of defect in 100% of cases due to precisely selected threshold values of metrics.
The researchers are confident that with the help of the simulator, students will develop the skills of visual recognition of defects and understanding the causes of their occurrence. The test mode, where a reference and a distorted image are shown side by side, simulates a real clinical situation. In practice, the doctor often compares the current snapshot with previous studies or with examples of the norm. Disabling prompts and automatic scoring will also be a significant advantage, which will allow the teacher to objectively assess the level of training and identify topics that cause the greatest difficulties.
Students of the PSU Medical Institute have already started studying using an interactive training simulator for recognizing defects in X-ray images.
Please note that this press release is based on materials provided by the company. AK&M Information Agency shall not be held liable for its contents, nor for the legal and other consequences of its publication.
