A neural network to help conduct a gentle diagnosis of ischemic heart disease

Igor Skirnevsky, a research fellow from TPU Medical Device Design Laboratory is developing software for automated analysis of myocardial perfusion scintigraphy results - a non-invasive method for diagnosing heart pathologies. The project is implemented jointly with experts from the Cardiology Research Institute of TNRMC. The developer expects to improve diagnostic accuracy by up to 90 %. 

‘The implementation of the proposed solution in clinical practice will reduce the number of misdiagnosed cases of coronary heart disease, as well as the number of wrong indications for surgical invasion, help evaluate risks of complications and treatment effectiveness,’

says the developer.

At present, invasive coronary angiography is one of the common methods of diagnosis of heart pathologies, which implies endovascular intervention which entails both additional costs for medical institutions and certain risks for patients. However, there is a non-invasive method of diagnosing the cardiovascular system, perfusion scintigraphy. The study is carried out using hi-tech equipment, a single-photon emission computed tomography (gamma camera), and the introduction of a radiopharmaceutical into a patient’s body. The amount of radiopharmaceuticals is carefully calculated to a safe dosage. Patients are examined at rest and by stress-test. A specialist can conclude if there is a pathology by the results of scintigraphy, as well as identify risks for developing complications and treatment options. Scintigraphy has become widespread in the United States, Europe and a number of other countries but in Russia, the method is not widely used yet. 

Photo: Perfusion scintigraphy results

‘Yes, invasive coronary angiography is a gold standard for the diagnosis of coronary artery disease. There are other diagnostic methods that do not require invasive intervention, hospitalization, expenditures for preparation and so on. Perfusion scintigraphy is similar to X-ray test: patients are placed in a special machine to carry out diagnostics and then sent home. The method allows identifying a myocardial pathology in a quite accurate way.

Foreign studies showed that in the countries where perfusion scintigraphy is widespread the most results of non-invasive tests are proved after conducting additional invasive examinations. Besides, perfusion scintigraphy is expected to become a standard, too. In Russia, it is included in the standards of care for patients with coronary artery disease, but not all clinics are provided with the required equipment. In Tomsk, at Cardiology Research Institute such studies are conducted on a regular basis,’ says Igor Skirnevskiy.

The developed software belongs to decision support systems and is based on computer image analysis using machine learning. The research will use depersonalized information about real patients.

The project of Software development based on machine learning algorithms for the creation of a decision support system for diagnosing coronary heart disease by perfusion scintigraphy was supported by a two-year grant from the UMNIK-2018 Foundation for Assistance to Small Innovation Enterprises.

During 2019, Igor Skirnevskiy plans to develop a model of a neural network. By the end of the year, he expects to develop a model with an accuracy of about 90 %. The second year will be devoted to the improvement of the model, the study of its effectiveness, obtaining feedback from Cardiology Research Institute and production of the finished software.