Artificial Intelligence learns to detect cancer signs at early stages

Oct 22, 2021

SberMedAI has upgraded its AI-powered Chest CT solution, which can now detect cancer at its early stages by analyzing chest CTs and will help doctors with diagnoses.

At its early stages, lung cancer looks like a small tissue change a few millimeters wide. The algorithm can now detect the smallest nodular neoplasms in the lungs and highlight potentially dangerous areas. For example, if the size of at least one of the detected neoplasms exceeds the threshold value of four millimeters, then Chest CT highlights all detected nodular neoplasms, regardless of their size, thereby indicating the areas for the doctor to pay attention to when analyzing and deciding on what to do next to identify the nature of the neoplasms.

The system was trained using anonymized images from open medical data sources. Radiologists labelled more than 12,000 scans manually to improve the Chest CT model; the scans were used to train artificial intelligence algorithms. Currently, the solution is being piloted in one of Russia’s cancer centers.

Yuri Krestinsky, vice president, head of Healthcare Industry, Sberbank:

“The chest CT analyzer has proven its efficiency during the pandemic when the number of scans requiring interpretation increased dramatically. AI would need a fewl minutes to analyze a scan and prioritize them for MDs, putting on top those that were found to have a potential problem. This means that higher-risk group patients could receive treatment earlier. With Chest CT looking for cancer, the quality of medical care should improve, and timely diagnosis can save many lives.”

Lung cancer is the most common cancer in the world and the number one cause of death among oncological pathologies. According to the National Artificial Intelligence Development Strategy 2030, accelerated digitalization is among the priorities when supporting the healthcare industry. The use of the updated Chest CT service can help doctors in identifying neoplasms on CT scans and improve the quality of cancer diagnoses at early stages.

The Chest CT service is currently utilized in 30 clinics and 19 regions of Russia to diagnose viral pneumonia, including COVID-19. Artificial intelligence marks locations and indicates the percentage of affected tissues, assisting doctors in their fight against the coronavirus. Chest CT is part of the systemic solution by SberMedAI, a medical digital diagnostic center (MDDC) uniting 50+ products and solutions by SberMedAI and other companies in the Sber ecosystem and its partners.