Advancing Outpatient Diagnosis of Obstructive Sleep Apnoea through Artificial Intelligence: Promise and Pitfalls

Dear Editor, Obstructive Sleep Apnoea (OSA) is a prevalent and clinically significant disorder of sleep characterized by recurrent episodes of sleep-related airflow obstruction despite continued respiratory effort. These events are predominantly caused by obstruction of the upper airway. These events are frequently accompanied by daytime somnolence, sleep disturbance, and recurrent hypoxia. A meta-analysis of 17 studies approximates that about 425 million people worldwide (95% CI: 399–450 million) have moderate-to-severe OSA aged 30 to 69 years (1). OSA is among the most underdiagnosed and undertreated of noncommunicable diseases worldwide, reflecting its high burden. The pathophysiologic consequences of OSA extend far beyond mere disturbance of sleep. Multiple studies have found robust links between OSA and systemic hypertension, coronary heart disease, stroke, atrial fibrillation, type 2 diabetes mellitus, cognitive impairment, and even oncogenic potential (2). Therefore, early diagnosis and management are required to avoid long-term neurological, metabolic, and cardiovascular morbidity and improve quality of life. Polysomnography has many advantages but with pros comes the cons like expenses, time taking methods, need for high maintenance instruments and skilled officials. These setbacks have pivoted the way of interest in direction of time saving, more easily approachable and measurable screening opportunities that are favorable to working environment and accelerate the best outcome with limited resources. With Artificial Intelligence being the owner of modern technology, procedures like advanced and deep Machine Learning have become core of diagnostics in recent developments, supported by proof in scanning for obstructive sleep apnea (OSA), in which Al-controlled systems are being constructed to overtake the manual interpretations for electrocardiography (ECG) and oxygen saturations (SpO₂) in running time, minimizing the need for laboratory services. An appreciable approach in this field of AI is Type IV Artificial Intelligence Sleep Monitoring (AISM) device. With minimum expenditure, this detects OSA biomarkers with barely any invasion and costs are lowered down. These developments have impacted in neglecting unnecessary lab tests that are costly, prioritizing the patient comfort and short listing for PSG (3), and efficient to detect early triage equity. Based on these advantages, Al screening tools are being designated as significant for early diagnosis of OSA.