Search

Background: Manual auscultation to detect abnormal breath sounds has poor inter-observer reliability. Digital stethoscopes (DS) with artificial intelligence (Al) could improve reliable detection of these sounds. Objective(s): We aimed to independently test the abilities of Al developed for the purpose of detecting wheezes/rhonchi and crackles in children. Method(s): 192 auscultation recordings collected from children using two different DS (ClinicloudTM and LittmanTM) were each tagged as containing wheezes/rhonchi, crackles or neither by a paediatric respiratory physician, based on audio playback and careful spectrogram and waveform analysis. Untagged versions of the recordings were submitted for analysis by a blinded Al algorithm (StethoMeTM Al) trained to detect pathologic paediatric breath sounds, which generated a probability score of the likelihood of presence of crackles orwheeze/rhonchi. Al outcome was compared with tagged outcomes on a per-recording basis, with receiver operating characteristic curves used to identify optimal cutoffs representing best Al performance. Result(s): With optimised Al thresholds, crackle detection positive percent agreement (PPA) was 0.95 and negative percent agreement (NPA) was 0.99 for Clinicloud recordings; for Littman-collected sounds PPA was 0.82 and NPA was 0.96. Wheeze detection PPA and NPA were 0.90 and 0.97 respectively (Clinicloud auscultation), with PPA 0.80 and NPA 0.95 for Littman recordings. Conclusion(s): Al can detect crackles and wheeze from breath sounds obtained using different DS devices with a degree of accuracy that approaches (or exceeds) that of clinicians. Careful integration into clinical practice may improve standards of care., A. Kevat, Monash Children's Hospital, Melbourne, Australia. E-mail: ajaykevat@gmail.com, CONFERENCE ABSTRACT

Background: Manual auscultation to detect abnormal breath sounds has poor inter-observer reliability. Digital stethoscopes (DS) with artificial intelligence (Al) could improve reliable detection of these sounds. Objective(s): We aimed to independently test the abilities of Al developed for the purpose of detecting wheezes/rhonchi and crackles in children. Method(s): 192 auscultation recordings collected from children using two different DS (ClinicloudTM and LittmanTM) were each tagged as containing wheezes/rhonchi, crackles or neither by a paediatric respiratory physician, based on audio playback and careful spectrogram and waveform analysis. Untagged versions of the recordings were submitted for analysis by a blinded Al algorithm (StethoMeTM Al) trained to detect pathologic paediatric breath sounds, which generated a probability score of the likelihood of presence of crackles orwheeze/rhonchi. Al outcome was compared with tagged outcomes on a per-recording basis, with receiver operating characteristic curves used to identify optimal cutoffs representing best Al performance. Result(s): With optimised Al thresholds, crackle detection positive percent agreement (PPA) was 0.95 and negative percent agreement (NPA) was 0.99 for Clinicloud recordings; for Littman-collected sounds PPA was 0.82 and NPA was 0.96. Wheeze detection PPA and NPA were 0.90 and 0.97 respectively (Clinicloud auscultation), with PPA 0.80 and NPA 0.95 for Littman recordings. Conclusion(s): Al can detect crackles and wheeze from breath sounds obtained using different DS devices with a degree of accuracy that approaches (or exceeds) that of clinicians. Careful integration into clinical practice may improve standards of care.

Background: Manual auscultation to detect abnormal breath sounds has poor inter-observer reliability. Digital stethoscopes (DS) with artificial intelligence (Al) could improve reliable detection of these sounds. Objective(s): We aimed to independently test the abilities of Al developed for the purpose of detecting wheezes/rhonchi and crackles in children. Method(s): 192 auscultation recordings collected from children using two different DS (ClinicloudTM and LittmanTM) were each tagged as containing wheezes/rhonchi, crackles or neither by a paediatric respiratory physician, based on audio playback and careful spectrogram and waveform analysis. Untagged versions of the recordings were submitted for analysis by a blinded Al algorithm (StethoMeTM Al) trained to detect pathologic paediatric breath sounds, which generated a probability score of the likelihood of presence of crackles orwheeze/rhonchi. Al outcome was compared with tagged outcomes on a per-recording basis, with receiver operating characteristic curves used to identify optimal cutoffs representing best Al performance. Result(s): With optimised Al thresholds, crackle detection positive percent agreement (PPA) was 0.95 and negative percent agreement (NPA) was 0.99 for Clinicloud recordings; for Littman-collected sounds PPA was 0.82 and NPA was 0.96. Wheeze detection PPA and NPA were 0.90 and 0.97 respectively (Clinicloud auscultation), with PPA 0.80 and NPA 0.95 for Littman recordings. Conclusion(s): Al can detect crackles and wheeze from breath sounds obtained using different DS devices with a degree of accuracy that approaches (or exceeds) that of clinicians. Careful integration into clinical practice may improve standards of care.

Background: Digital stethoscope technology has been used to record and study normal and abnormal breath sounds in the paediatric population. There may be scope to utilise this technology to describe breath sound characteristics in the transitioning newborn. Method(s): A commercially available digital stethoscope and acquisition software was used to record breath sounds of term infants born via normal vaginal delivery (NVD) and elective caesarean section (CS) at one minute and two hours of life. Recordings had a bandpass filter applied and analysed using MATLAB. Power spectrum profiles, total power (TP), peak width (PW), spectrum slope (SL) and area under the regression line (PLN) were extracted and compared between delivery modes. Result(s): Sixty-one newborns were studied. Mean (SD) gestation and weight of 32 NVD infants was 38 (0.9) weeks and 3165.6 (447.1) grams, and 29 elective CS infants was 38 (0.9) weeks and 3222.1 (297.1) grams. 83.6% of 1 min and 100% of 2-h recordings were analysable. Mean difference in spectral features between CS and NVD groups at 1-min: TP 1.56 dB/Hz (p < 0.001), PW 4.14 Hz (p < 0.001), PLN 412.82 dB (p = 0.002), SL 0.63 dB/octave (p = 0.003). Mean difference at 2 h: TP 0.99 dB/Hz (p = 0.004), PW 1.54 Hz (p = 0.01), PLN 200.56 dB (p = 0.03), SL 0.34 dB/octave (p = 0.01). Conclusion(s): It is feasible to use digital stethoscope technology to assess breath sounds in the transitioning newborn. We were able to depict a difference in breath sound characteristics between infants delivered via elective CS and NVD. This may be associated with differential lung fluid clearance.

Background: Digital stethoscope technology has been used to record and study normal and abnormal breath sounds in the paediatric population. There may be scope to utilise this technology to describe breath sound characteristics in the transitioning newborn. Method(s): A commercially available digital stethoscope and acquisition software was used to record breath sounds of term infants born via normal vaginal delivery (NVD) and elective caesarean section (CS) at one minute and two hours of life. Recordings had a bandpass filter applied and analysed using MATLAB. Power spectrum profiles, total power (TP), peak width (PW), spectrum slope (SL) and area under the regression line (PLN) were extracted and compared between delivery modes. Result(s): Sixty-one newborns were studied. Mean (SD) gestation and weight of 32 NVD infants was 38 (0.9) weeks and 3165.6 (447.1) grams, and 29 elective CS infants was 38 (0.9) weeks and 3222.1 (297.1) grams. 83.6% of 1 min and 100% of 2-h recordings were analysable. Mean difference in spectral features between CS and NVD groups at 1-min: TP 1.56 dB/Hz (p < 0.001), PW 4.14 Hz (p < 0.001), PLN 412.82 dB (p = 0.002), SL 0.63 dB/octave (p = 0.003). Mean difference at 2 h: TP 0.99 dB/Hz (p = 0.004), PW 1.54 Hz (p = 0.01), PLN 200.56 dB (p = 0.03), SL 0.34 dB/octave (p = 0.01). Conclusion(s): It is feasible to use digital stethoscope technology to assess breath sounds in the transitioning newborn. We were able to depict a difference in breath sound characteristics between infants delivered via elective CS and NVD. This may be associated with differential lung fluid clearance.

Background and Aims: VATS is commonly performed in children with empyema, however, there is no data regarding the post-operative outcomes (after discharge) following the procedure. Method(s): A 5-year (2013-2018) retrospective review was performed on children treated with VATS and chest drain for empyema. The study was approved by the local ethics committee (RES-18-0000-071Q). Statistics: results are reported as number of cases (%) and median (range). Correlation and logistic regression analysis were performed. Result(s): We identified 62 children, 36 (58%) males, with a median age of 3.6 (0.4-14.5) years treated over the 5-year period. Presentation was: autumn 12 (19%), winter 19 (31%), spring 19 (31%) and summer 12 (19%) patients. Median post-operative follow-up was 47 (9-489) days. Streptococcus species were identified in 35 (56%) of cultures from pleural fluid. Chest drain was left in situ for a median of 4 (2-13) days. Four (6%) children required a second procedure (3 VATS, 1 thoracotomy). Median duration of pre-operative symptoms before the VATS procedure was 9 (4-17) days. Median post-operative length of stay was 8 (3-28) days and duration of oral antibiotics therapy after discharge was 28 (3-42) days. Thirty-nine (63%) required post-operative oxygen therapy for 3 (1-16) days. There was no correlation between the duration of pre-operative symptoms at the time of VATS and post-operative length of stay (r = -.03 [95% CI-0.3-0.2]; P = 0.7] (Figure) Post-operative x-ray was available for 56 (90%) patients and documented persisting lung or pleural abnormalities in 47 (84%). Logistic regression analysis failed to identify any predisposing factors for persisting x-ray abnormalities (P = 0.63). Conclusion(s): We found a significant incidence of persisting abnormalities on post-operative x-ray following VATS with chest drain for empyema in children. Long-term follow-up of children following VATS might be indicated to identify patients at risk of chronic respiratory pr

Background and Aims: VATS is commonly performed in children with empyema, however, there is no data regarding the post-operative outcomes (after discharge) following the procedure. Method(s): A 5-year (2013-2018) retrospective review was performed on children treated with VATS and chest drain for empyema. The study was approved by the local ethics committee (RES-18-0000-071Q). Statistics: results are reported as number of cases (%) and median (range). Correlation and logistic regression analysis were performed. Result(s): We identified 62 children, 36 (58%) males, with a median age of 3.6 (0.4-14.5) years treated over the 5-year period. Presentation was: autumn 12 (19%), winter 19 (31%), spring 19 (31%) and summer 12 (19%) patients. Median post-operative follow-up was 47 (9-489) days. Streptococcus species were identified in 35 (56%) of cultures from pleural fluid. Chest drain was left in situ for a median of 4 (2-13) days. Four (6%) children required a second procedure (3 VATS, 1 thoracotomy). Median duration of pre-operative symptoms before the VATS procedure was 9 (4-17) days. Median post-operative length of stay was 8 (3-28) days and duration of oral antibiotics therapy after discharge was 28 (3-42) days. Thirty-nine (63%) required post-operative oxygen therapy for 3 (1-16) days. There was no correlation between the duration of pre-operative symptoms at the time of VATS and post-operative length of stay (r = -.03 [95% CI-0.3-0.2]; P = 0.7] (Figure) Post-operative x-ray was available for 56 (90%) patients and documented persisting lung or pleural abnormalities in 47 (84%). Logistic regression analysis failed to identify any predisposing factors for persisting x-ray abnormalities (P = 0.63). Conclusion(s): We found a significant incidence of persisting abnormalities on post-operative x-ray following VATS with chest drain for empyema in children. Long-term follow-up of children following VATS might be indicated to identify patients at risk of chronic respiratory pr

Introduction/Aim: The audio characteristics of abnormal paediatric breath sounds are not well-defined and digital techniques to capture them have not been widely investigated. Our study aimed to define the audiological features of abnormal paediatric breath sounds objectively and compare the ability of digital stethoscopes to detect them against auscultation using standard bell-and-diaphragm stethoscopes. Method(s): Twenty children with normal breath sounds, generalized wheeze or crackles, and some with a diagnosis of cystic fibrosis were auscultated by a paediatric consultant and digitally recorded using both the LittmanTM 3200 Digital Electronic Stethoscope and a ClinicloudTM Digital Stethoscope. We used spectrographic analysis and recording playback to detect abnormal breath sounds and define their audio waveform characteristics. Result(s): Digital stethoscopes were more sensitive than standard auscultation in detecting wheeze in our study. Patients with clinicallydescribed wheeze had periodic waveform segments of increased intensity and frequency compared to background breath sounds, mostly spanning expiration for a period of 0.03-1.2 seconds at frequencies of 100-1050Hz, and occasionally spanning shorter inspiratory segments. Recordings of patients with crackles revealed brief (6-20 millisecond) discontinuous sounds with a distinguishing waveform identifiable within them. Conclusion(s): Digital breath sound analysis may be more sensitive than manual auscultation in detecting breath sound abnormalities in children, with potential applications for improved diagnosis, data sharing and disease monitoring. Further research and development is needed.

Introduction/Aim: The audio characteristics of abnormal paediatric breath sounds are not well-defined and digital techniques to capture them have not been widely investigated. Our study aimed to define the audiological features of abnormal paediatric breath sounds objectively and compare the ability of digital stethoscopes to detect them against auscultation using standard bell-and-diaphragm stethoscopes. Method(s): Twenty children with normal breath sounds, generalized wheeze or crackles, and some with a diagnosis of cystic fibrosis were auscultated by a paediatric consultant and digitally recorded using both the LittmanTM 3200 Digital Electronic Stethoscope and a ClinicloudTM Digital Stethoscope. We used spectrographic analysis and recording playback to detect abnormal breath sounds and define their audio waveform characteristics. Result(s): Digital stethoscopes were more sensitive than standard auscultation in detecting wheeze in our study. Patients with clinicallydescribed wheeze had periodic waveform segments of increased intensity and frequency compared to background breath sounds, mostly spanning expiration for a period of 0.03-1.2 seconds at frequencies of 100-1050Hz, and occasionally spanning shorter inspiratory segments. Recordings of patients with crackles revealed brief (6-20 millisecond) discontinuous sounds with a distinguishing waveform identifiable within them. Conclusion(s): Digital breath sound analysis may be more sensitive than manual auscultation in detecting breath sound abnormalities in children, with potential applications for improved diagnosis, data sharing and disease monitoring. Further research and development is needed.