Part I: Survey
In this part, each participant responded to a number of questions regarding:
- education, the specialization started or held,
- assessment of their own skills in adult and child auscultation,
- type of stethoscope (electronic / analog) and frequency of auscultation in their medical practice,
- opinion on scale a 5-grade on the number of hours devoted to studying auscultation during their study and specialization, the need for additional training, and the scale of the problem of ambiguity in the nomenclature used in the classification of auscultation sounds.
Part II: Classification of hearing sounds
This part consisted of 24 sounds (see Tab. 1 for details) that the test participant listened to (they could replay each sound), evaluated and assigned to specific classes (details below). Nine sounds were selected from the demonstration recordings included on a CD in the Fundamentals of Lung and Heart Sounds (Willkins et al., 2004). The rest were recorded with the Littmann 3200 electronic stethoscope. These were the records of the respiratory sounds of children aged 5 months to 14 years (average 7.6 years).
The choice of sounds for the test was two-step. First, sounds from a database of over 2000 sounds were selected by acousticians. Those sounds were chosen because they contained the smallest number of distortions and artifacts. The artifacts of Littmann 3200 stethoscopes appear mainly when the chestpiece is moved and during the application or deposition of the chestpiece from the body. These disturbances are mainly caused by the acousto-electric transducer. The other equally important criterion was the choice of sound class to make the sounds as diverse as possible, but at the same time unambiguous. This way 50 different sound examples were selected. The final selection of sounds was made by the team of specialists composed of eight experienced pediatricians and pulmonologists working at the Karol Jonscher Clinical Hospital in Poznań, Karol Marcinkowski Poznań University of Medical Sciences. At the meeting of those physicians, the Fostex PH-50 headset coupled with high quality professional headphones (Sennheiser HD600) enabled simultaneous listening of sound samples to all physicians. After listening to each signal, they classified the sound. Then there was a discussion about it and a common position was held. Finally, only the sounds which no physicians had any doubt over were chosen for the test. The set of those signals with their descriptions are called “standard” here.
The sounds represented certain classes (Table 1) and were presented to the participants as a collection of signals from 24 different patients. The terms used in the classes were chosen to include that the entire spectrum of nomenclature that is used in modern medical literature (Szczeklik i Szczeklik, 1979, Rowińska-Zakrzewska and Kus, 2004, Willkins et al., 2004, Mangione and Nieman 1999, Pasterkamp et al., 2016) , and thus by physicians in their daily practice. During the experiment more than one class could be assigned to each sound by a listener. Additionally, with each sound there was information about the place on the chest that the chestpiece of the stethoscope was placed on during recording, as well as the age, height and weight of the patient. Each participant was informed about the need for high quality headphones. No feedback was used, which means that the participants were not informed about the correctness of their response during the test. Moreover, the order of presentation was randomly chosen, but it was the same for all participants. As a result, the participants could not learn the correct answers and patterns as they performed the listening task.