Artificial intelligence to improve the diagnosis of pulmonary hypertension: promises and pitfalls

Pulmonary hypertension (PH) is a clinical–physiological syndrome thought to affect 1% of the global population.1 PH is defined haemodynamically by mean pulmonary artery pressure >20 mm Hg, resulting in right ventricular (RV) overload and often RV failure. Patients with PH experience symptoms including dyspnoea, fatigue and oedema, often associated with physical and psychosocial disability. In some forms of untreated PH, mean survival is less than 3 years, related in part to delays in diagnosis of 1–2 years after clinical onset of disease. Diagnostic delays are due to non-specific symptoms, subtle clinical examination findings and low sensitivity of detection of PH features on imaging, ECG and pulmonary function testing. Indeed, patients are still most often diagnosed at an advanced stage, typically in New York Heart Association (NYHA) functional class 3. Guidelines recommend screening high-risk patients, such as those with scleroderma, with transthoracic echocardiogram looking for elevated pulmonary pressures or secondary features of PH, such as right-sided structural or functional abnormalities.

PH is most commonly due to underlying diseases of the left side of the heart (WHO group 2 PH) and respiratory conditions characterised by hypoxaemia (WHO group 3 PH), cumulatively comprising about 90% of PH cases globally. WHO group 1 pulmonary arterial hypertension (PAH) and group 4 PH due to pulmonary arterial obstruction are uncommon, each being diagnosed in less than 5% of all patients with PH. However, they are a major focus of clinical PH diagnostic assessment as there are specific treatment approaches for patients in both groups. In WHO group 4, the most common and important condition chronic thromboembolic PH (CTEPH) is often curable; thus, all patients with significant PH should be rigorously assessed for possible CTEPH through pulmonary vascular imaging. WHO group 1 PAH is a disease specifically of the pulmonary circulation, sometimes related to underlying conditions and exposures (eg, connective …

留言 (0)

沒有登入
gif