The Diagnosis of Asthma using Hilbert-Huang Transform and Deep Learning on Lung Sounds

RDsLINet: A Novel Lightweight Inception Network For Respiratory Disease Classification Using Lung Sounds

“…For task 2, though [2] provides decent amount of accuracy, the overall algorithm including the preprocessing steps, is heavily computationally expensive since it includes EMD, which is computationally heavy due to iterative decomposition of multicomponent signal [35], thereby increases the preprocessing time and the same has been discussed by Subho et.al [2]. Likewise, for task 3, we have outperformed the only existing work by Altan et al [13] by a margin of 7%. Additionally, we have reduced the burden of hand crafted feature extraction process proposed by Altan et al [13] by employing the deep learning architecture.…”

“…Likewise, for task 3, we have outperformed the only existing work by Altan et al [13] by a margin of 7%. Additionally, we have reduced the burden of hand crafted feature extraction process proposed by Altan et al [13] by employing the deep learning architecture.…”

“…In recent years, various studies have been conducted to classify respiratory diseases using lung sound signals taken from either D1 [1], [2], [20] or private volunteer database [13]. However, we have utilized all public available databases (D1, D2, and D3) to classify wide spectrum of respiratory diseases.…”

“…In [2], [20], D1 is not used for asthma classification as it is one of the minority classes, however as we are using D2 also for asthma classification task, we have taken into account both databases. Additionally, to the best of our knowledge, this is the second work on asthma classification based on lung sounds following to Altan et al [13]. To evaluate the performance of the proposed RDsLINet, we have employed the following performance metrics: accuracy (Acc), specificity (Spe), recall (Rec) or sensitivity (Sen), and precision (Prc) to evaluate our proposed framework similar to [28], [29], [30], [31].…”

“…Asthma is one of the minority classes in the ICBHI datset with only two sound recordings, making deep learning approaches unreliable for its inclusion in most classification methods used in [2], [10], [12]. Recently, Altan et al [13] has shown an interesting result on asthma classification by utilizing lung sound data from their own recorded database. Hilbert transform (HHT) based time and frequency domain feature extraction process followed by a deep belief network (DBN) has been utilized and an accuracy rate of 84.61% has been achieved for binary class classification (asthma vs healthy).…”