Specialists of Almazov Centre and scientists from St. Petersburg Electrotechnical University «LETI» joined their efforts to develop an AI-based diagnostic module that can determine the morphological form of congenital pancreatic disorder in a few minutes.
To this end, they created a new version of the software that allows doctors to quickly and accurately determine the form of hyperinsulinism from CT images. The program is already used by doctors of the clinic at Almazov Centre.
Congenital hyperinsulinism (CHI) is one of the most dangerous diseases, especially for young children. It is a genetic disorder characterized by excessive insulin secretion. The disease can negatively impact the central nervous system and cognitive development of the brain. A key step for quality care is to determine the morphological form of CHI: diffuse, focal or atypical.
“18F-dihydroxyphenylalanine PET is a non-alternative non-invasive method for diagnosing the morphological form of congenital hyperinsulinism. The introduction of a software module for the analysis of images based on artificial intelligence will allow us to quickly and accurately make a differential diagnosis of the morphological form and decide on a treatment strategy,” explains Prof. Daria Ryzhkova, Head of the Department of Nuclear Medicine and Radiation Technologies with Clinic.
The new version of the program has a function to determine the form of CHI. This became possible due to the fact that scientists found a relationship between the form of the disease, the brightness and hue of insulin foci when analyzing each pixel on CT images. Moreover, the program can reliably capture differences in color characteristics that may be invisible to the doctor's eye. In total, the study used 50 images of children with congenital hyperinsulinism.
Scientists have identified three ranges of values for brightness and hue, each corresponding to a specific type of disease. According to Alexander Krasichkov, Professor at the Radio Engineering Department of LETI University, the developed program determines the form of hyperinsulinism if both indicators correspond to the same form of the disease. The reported accuracy is 86%.
“CHI remains difficult to diagnose and treat. Untimely and improper medical care leads to severe intellectual and somatic disability and reduced life expectancy. Almazov Centre is the only medical facility that provides expert-level medical care to children with CHI in Russia and the CIS countries. The introduction of an AI-based software module has made it possible to significantly optimize the diagnostic flow for this complex disease, and as far as atypical forms are concerned, to develop a fundamentally new approach to planning treatment tactics,” comments Prof. Irina Nikitina, Head of Children’s Diseases Department with the Clinic.