Automated Identification of Innocent Still's Murmur in Children

Abstract

To our knowledge, our solution is the first murmur classifier that focuses singularly on Still's murmur. Following further refinement and testing, the presented algorithm could reduce the number of children with Still's murmur referred unnecessarily to pediatric cardiologists.

A Smartphone Stethoscope and Application for Automated Identification of Innocent Still’s Murmur

“…Murmur Analysis -The analysis algorithm has been developed and tested using an in-house murmur library with clinically documented diagnoses. The library consisted of 119 Still's murmurs, 84 other innocent murmurs, and 502 pathological murmurs [3].…”

“…For our second goal, we compared our automated Still's murmur identification algorithm [3] on Still's and pathological murmurs recorded using StethAid stethoscope. Utilizing the Littmann stethoscope, we built a database of different heart murmur types and their associated diagnoses as provided by an expert pediatric cardiologist.…”

“…We have created a novel digital stethoscope (see Figure 1), which connects to a smartphone, and developed a custom mobile application that records heart sounds, using the developed hardware, and analyzes them for the presence of Still's murmur [3]. We call this application and digital stethoscope StethAid.…”

An Approach for Automatic Identification of Fundamental and Additional Sounds from Cardiac Sounds Recordings

“…As the duration of S1 and S2 sounds (without splitting of subcomponents) is known to range between 80 ms to 150 ms during normal cardiac operation, sound lobes with shorter or longer durations were investigated as potentially abnormal heart sounds [5]. The duration and energy of adjacent peaks were also evaluated, as S1 and S2 sounds cannot have the same energy under normal cardiac conditions [9]. Hence, if two adjacent peaks were found to have the same energy, then one of the two sounds was considered as abnormal.…”

“…In case of splitting of S1 or S2, the interval between the two discrete consecutive sound lobes would be less than 50 ms and both sound lobes would have different energies [9]. This is because one of the split components of S1 or S2 will have a lower intensity compared to the another component [9]. Further, the consecutive sound lobes with time interval greater than 50 ms, were investigated for abnormal/additional lub/dub sounds.…”