China Digital Health: Telemedicine, AI Diagnosis, and Healthcare Technology

China's AI medical diagnosis systems have demonstrated impressive accuracy in specific clinical domains, though they remain decision-support tools rather than replacements for doctors: in lung cancer screening from CT scans, systems from Infervision, SenseTime Medical, and Baidu achieved sensitivities of 95-97% and specificities of 85-92% in clinical trials, comparable to experienced radiologists and in some cases exceeding average radiologist performance; in diabetic retinopathy screening from fundus photography, systems from Airdoc and Tencent Miying achieved over 95% sensitivity at detecting referable diabetic retinopathy, meeting WHO screening guidelines; in stroke detection from CT scans, systems from VoxelCloud and BioMind achieved detection times of under 1 minute compared to 5-10 minutes for human radiologists, critical for time-sensitive treatment decisions; in ECG interpretation, systems from Lan-bridge Technology and DMK achieved diagnostic accuracy of 98% for common arrhythmias, approved for use as a Class II medical device; in pathology, AI systems from Yitu Healthcare achieved 95% accuracy in cervical cell screening, reducing the workload of pathologists by 70%; however, important limitations remain including these systems are trained on specific Chinese patient populations and may have reduced accuracy in other populations, they work best as screening tools rather than definitive diagnostic tools, regulatory approval requires ongoing post-market surveillance, and complex multi-system diseases remain beyond current AI capabilities. China's NMPA (National Medical Products Administration) classifies most AI diagnostic devices as Class II or Class III medical devices, requiring rigorous clinical validation before approval.