Conclusions
Auscultation is an important part of a medical examination. It allows rapid screening as an essential step in practically every visit. From the analysis of the data described in this paper, it can be stated that the survey results clearly indicate the need for more training at different stages of learning and medical practice in this area. They also indicate the awareness of the medical community concerning the lack of a unified nomenclature of respiratory sounds. The results of the questionnaire are confirmed by the results obtained in the test section, in which the participants of the experiment classified the previously recorded respiratory sounds. These results were compared with the standard developed by the team of specialist physicians. The average number of correct detections of auscultation phenomena in the test (not taking into account the sound classes) is only from 24.1% for students to 36.5% for pulmonologists. It should be borne in mind that correct detection means that additional phenomena may have been identified which, according to the standard, were not present in the sound sample (P). If one looks more restrictively at the results and takes into account only those responses that are in full compliance with the standard, the average number of responses compliant with the standard (CS) is just 14.7% for all groups. If one splits the results into detailed and main sound classes, it can be noticed that the highest number of correct answers was obtained for the wheeze classes. This is confirmed by both the correct answers received by participants of different professions and students (Fig. 4 and Fig. 5), as well as the lowest number of false positives noticed for this class (Fig. 6). This is not surprising, since they are the most characteristic sounds: continuous, tonal, periodic and relatively loud, e.g. in the case of stridor, a characteristic type of inspiratory wheeze, 100% correct answers for pulmonologists was found. More problematic are the classes of crackles and rhonchi. For crackles, in the case of detailed, coarse, medium and fine crackle classes, the number of correct detections was relatively low and for the best group of physicians (pulmonologists) the number ranged between 40% of correct responses, depending on the type of crackles. This is comparable and consistent with the results obtained by US doctors (Mangione and Nieman, 1999). If all types of classes along with crackles are grouped together (i.e. each of the answers for the class of crackles: fine, medium, coarse crackles and crepitus are considered to be correct, regardless of which one was correct according to the standard), then the result is significantly improved and reaches 73.4% of the correct answers for the group of pulmonologists. For the remaining groups, this value is around 60% of the correct answers. In the case of rhonchi, which are the most capacious class of respiratory sounds, the advantage of pulmonologists is clearly visible. In their case, the number of correct detections is 51.2%, while for other groups this value does not exceed 30%. The lowest number of correct indications is visible for the breath sound. The results show that few physicians can unequivocally assess the appearance of a louder breath sound, prolongation of the expiratory phase or diminished breath sound. These are phenomena whose unambiguous analysis would require more recordings, from different recording points, and preferably the reference sound recorded in a healthy patient.
For all classes of sounds, one can see the advantage of the pulmonologist group in terms of their correct answers, which is also consistent with the results of Mangione and Nieman (1999). This shows indirectly that it is possible to improve the skills of doctors in this field, through training and practice.
The comparison of the results of the correct answers for the detailed classification (Fig. 4) and the classification with the grouped (main) classes (Fig. 5) shows that this grouping significantly increases the percentage of correct responses. It must be stressed that this grouping results from the fact that if doctors are mistaken in the detailed classification, they usually confuse the subgroups of a given class, as is also shown in Figs. 6-8. This can be interpreted on the one hand as the possibility of the semantic ambiguity of classes (nomenclature), and on the other hand as a too detailed division, which due to the acoustic characteristics of distinctive features of sounds is unrealistic, because doctors in most cases are unable to distinguish these sounds using their hearing only.
Analyzes of the sound classes that the respondents marked, together with the correct answer, show how they grouped the individual auscultation phenomena into the main sound classes. It can be seen that in the case of normal sounds the category of crackles often appears in the supplementary answers. This is most likely related to the quality of the recording. With the use of piezoelectric transducers in the Litmann 3200 electronic stethoscope, it is not possible to eliminate artifacts associated with moving the stethoscope chestpiece. These artifacts are akin to crackles and are in most cases perceptually indistinguishable from pathological sounds. In the case of the sounds chosen for the test, these samples were almost devoid of these artifacts, but as the results show, some people perceived silent sounds resembling crackles and qualified them as pathological sounds. This problem seems particularly important when listening to a remote patient, where the doctor is unable to verify whether the sound is an artifact, or if it is a cause for concern. The technical solution to this problem is to use a microphone instead of a piezo transducer.
The remaining results from the analysis of additional phenomena indicate that the responses group sounds to some general classes and this is consistent with the standardized nomenclature (Pasterkamp et al., 2016). On this basis one can distinguish three main classes of sounds: crackles (which are divided into classes of fine, medium and coarse), wheezes and rhonchi. Then in these categories physicians (mainly pulmonologists) may try to distinguish specific, detailed classes. In the case of the crackles and crepitus class, it is noteworthy that, for nomenclature and semantic meaning among physicians, the class of fine crackles and the class of crepitus are of equal quality and often confused with each other (when the correct response is fine crackles, the crepitus response is more often chosen (Fig. 6d). In turn, the class of coarse crackles (Fig. 6b) is slightly more confused with the class of rhonchi than with the other subtypes of crackles, which also indicates the difficulty in finding the acoustically distinctive features of these classes. Consequently, it can be concluded that acoustically and semantically they are often referred to the same class by physicians. The medium crackles seems unimportant, as they are confused with both crepitus and rhonchi, and the number of correct answers is half the incorrect answers (this class is classified as crepitus). In general, based on the analysis of the results presented in Fig. 8, it can be concluded that in the future with this kind of phenomena it would be good idea to use the classification according to Pasterkamp et al. (2016), which is comprised of only the following classes: fine crackles, crepitus and coarse crackles.
The most ambiguous is the class of rhonchi, which is at the boundary between the class of wheezes and the class of medium and coarse crackles. Rhonchi, like wheezes, are continuous and periodic sounds. However, their fundamental frequency is lower than that of wheezes. They are often mistaken for them (Fig. 8c). The formation of rhonchi is associated with the movements of secretions in the respiratory pathways, which often results in a stertorous, intermittent sound that may be confused with coarse and medium crackles (as shown in Figs. 8a and 8b). This is consistent with the results obtained by Willkins etal. (1990), who also showed that rhonchi are an ambiguous class, very often incorrectly identified.
As has already been mentioned, Fig. 8 also shows that wheezes are the most homogeneous class, and in the case of samples containing this phenomenon as a correct answer described by the standard they were mainly confused with respect to the respiratory cycle (inhale / exhale).
In conclusion, it can be considered that the effectiveness of physicians in the clear classification of auscultation sounds with the use of detailed sound classes is low and very heterogeneous. It seems that in the case of a screening test, which can be assumed for a respiratory examination with a stethoscope, the standardized nomenclature proposed by Pasterkamp et al. (2016) is sufficiently precise and should provide the basis for a standardized classification of sounds for every language and further detailed respiratory diagnostics. This kind of classification should be also standardized and taught during medical studies and scholarships as a globally unified and unequivocal theory with the same semantic meaning. This approach would eliminate misunderstanding between physicians which is very common even when they speak the same language.
It should also be emphasized that due to the ever-growing market of electronic stethoscopes, it is possible to record normal and pathological respiratory sounds.
Accordingly the results of the study prove that the ability of detection of different auscultation phenomena can be significantly improved by training (which is made in everyday practice by pulmonologists who obtained higher scores). Moreover the survey across the participants suggests that there is a strong need to have more training during study and further practice. It must be emphasized that this problem is a global one which was also mentioned by other authors. It seems reasonable to create a worldwide database with auscultation signals described by the best specialists that will be used during education. Additionally the possibility of recording and storage of signals during the examination may be a good verification of the quality of this subjective procedure and may lead to increase in objectivity. This is highlighted also by the results that suggest that physicians (except pulmonologists), in general, are not better than medical students. This confirms the need of practical education not only during studies but also during further medical practice.