Truace
Subscribe

Identification of adventitious lung sounds in children with respiratory illnesses in Bangladesh using artificial intelligence and digital auscultation

Integrated management of childhood illness (IMCI) guidelines have high sensitivity but low specificity for pneumonia diagnosis. Artificial intelligence (AI)-enabled digital stethoscopes capable of analysing lung sounds may improve IMCI diagnostic performance. We evaluated the performance of an AI algorithm trained to identify normal and adventitious lung sounds in children. Non-physician health workers recorded lung sounds from four chest positions using a digital stethoscope in under-five-year-old children with…

By the Truvace desk2 min read1 sourcePeer-reviewedTRV-2026-1722· cite & verify
Identification of adventitious lung sounds in children with respiratory illnesses in Bangladesh using artificial intelligence and digital auscultation
"Health professional examines a coin while working at a desk during a consultation in a modern office setting" by nenadstojkovicart is licensed under CC BY 2.0. To view a copy of this license, visit https://creativecommons.org/licenses/by/2.0/.

In brief

Researchers tested an AI algorithm that labels recordings from a digital stethoscope as normal or abnormal (crackles and/or wheeze). Non-physician health workers collected the recordings from four chest positions in under-five children with suspected pneumonia at community clinics in Bangladesh, and a paediatrician listening panel provided the reference classifications.

The work matters because standard childhood illness guidelines are sensitive but not specific for pneumonia, so better tools could in principle cut misdiagnosis. The algorithm's moderate performance means it is not yet a reliable fix, and it still needs further training to recognise recordings that cannot be interpreted.

Main points

  1. Of 990 enrolled children, 867 (87.6%) had at least three interpretable chest position recordings and were analysed.
  2. Among all children, the algorithm's sensitivity and specificity for abnormal sounds were 61.8% and 60.7%.
  3. In children meeting IMCI pneumonia criteria, sensitivity was 63.5% and specificity was 66.8%.
  4. The algorithm did not have an uninterpretable category, which the authors say needs further training.

The problem

Against a paediatrician panel, the AI algorithm detected abnormal lung sounds in under-five children with only about 62% sensitivity and 61% specificity, which limits its current usefulness for pneumonia diagnosis.

The rundown

The study responds to a known weakness in integrated management of childhood illness guidelines, which have high sensitivity but low specificity for pneumonia. The authors suggest AI-enabled digital stethoscopes might improve diagnostic performance.

Recordings were gathered by non-physician health workers using a digital stethoscope at community clinics. A trained paediatrician panel labelled each chest position as normal, abnormal, or uninterpretable, and the AI labelled the same recordings except for the uninterpretable category.

Comparisons were made at the child level and covered interpretable, uninterpretable, and high- and low-confidence recordings. The reported accuracy, about 61-67% for sensitivity and specificity, was described as moderate.

Sources

  1. Peer-reviewedScientific Reports2026-10-16

The debate