Cardiology Diagnosis Automation Zhurek AI

Abstract

The leading cause of deaths in Kazakhstan is Cardiovascular disease(CDV). Some revisions were made in international heart-failure guidelines since 2016, but new diagnostic standards are adopted slowly in primary care, especially outside main cities [1]. This gap is caused by a noticeable shortage of specialists capable of interpreting echocardiographic (echo) studies. Based on this, the team identified three key objectives: 1. Lessen the reliance on specialists by automating echocardiographic analysis. 2. Lower the amount of misinterpretation of important cardiac measurements through AI-driven report generation. 3. Ensure clinical transparency through explainable and verifiable AI outputs that clinicians can trust. To address these objectives the team built Zhurek AI, an AI-powered web platform for echocardiography analysis. The platform creates structured clinical reports covering ejection fraction, valvular evaluation, wall motion, and pericardial findings. It is powered by EchoPrime [2], a multi-video view-informed vision-language model trained on 12.1 million echocardiographic clips, which achieves a mean AUC of 0.85–0.92 and exceeds human cardiologists on several tasks.