A IA consegue identificar cancro do pulmão em fase inicial a partir de biomarcadores no hálito usando narizes eletrónicos portáteis ?

Certos compostos orgânicos voláteis no ar expirado alteram-se na presença de cancro do pulmão em fase inicial, mesmo antes de a imagiologia detetar tumores. Narizes eletrónicos com IA poderiam analisar amostras de ar expirado em clínicas ou farmácias. Isto poderia reduzir a dependência de biópsias invasivas e de TACs. No entanto, fatores ambientais como o tabagismo ou a poluição atmosférica podem confundir os resultados.

Background

Researchers have demonstrated that portable electronic noses (e-noses) can detect volatile organic compounds (VOCs) in exhaled breath with promising sensitivity and specificity for early-stage lung cancer screening. A 2022 meta-analysis reported pooled sensitivity of about 85% and specificity of 87% across multiple studies using machine-learning models trained on breath-chemistry data. Certain volatile organic compounds in exhaled breath change in presence of early lung cancer, even before imaging detects tumors, and AI-powered e-noses could analyze breath samples in clinics or pharmacies, reducing reliance on invasive diagnostics. However, environmental factors like smoking or air pollution may confound results. Furthermore, real-world deployment faces challenges such as sensor drift, environmental confounders like smoking or diet, and the need for larger, multi-center validation cohorts. Regulatory approval remains limited to a few devices with narrow indications, underscoring the gap between promising research and routine clinical use.