AI Masters Sleep Study Data to Diagnose Hidden Apnoea

A staggering one billion people globally suffer from obstructive sleep apnoea (OSA), a condition where breathing repeatedly stops and starts during sleep. Shockingly, more than 80% of cases are never diagnosed, largely because the gold-standard test, polysomnography, is so labour-intensive.

Now, artificial intelligence offers a powerful new tool. A meta-analysis, which pools data from multiple studies, scrutinised 19 AI models tested on thousands of participants. The results were impressive: overall, AI achieved high accuracy in identifying OSA. A specific type of AI, known as a neural network, proved to be the top performer.

These neural networks demonstrated a sensitivity of 92.8%—meaning they correctly identified nearly 93 out of 100 true cases. While researchers note the need for further validation before these models are integrated into routine medical practice, this high-quality evidence signals a future where diagnosis could become far more efficient and accessible, easing a significant healthcare burden.