Serkan Nural's research while affiliated with Hatay Antakya State Hospital and other places

... The highest frequency of interest was determined to be 150 Hz, and therefore, taking into account Nyquist's theorem, 500 Hz was chosen as the sampling frequency, which is equivalent to a sampling time of 2 ms [7]. On the contrary, Altan et al. preferred using Hilbert-Huang Transform to find out features from lung sound [8]. ...

... 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). ...

... Using deep belief network (DBN) along with the features extracted from the modes, the authors had achieved 93.67% classification accuracy. For COPD categorization, Altan et al. [15] had suggested another framework that uses 3 dimensional second order difference plots (3D-SODP) to extract features from lung sound data and then fed them to DBN to distinguish two extreme severities of COPD namely COPD0 and COPD4. They achieved 95.84% accuracy rate for two class severity classification which is the first work on lung sound based COPD severity detection. ...

... The upper layers of the DBN represent features that are more abstract whereas the lower layers of the DBN learn simple features [1]- [3]. The DBN was implemented to electroencephalography signal to detect the brain activities of the patients in stroke [4]; electrocardiogram signal to detect coronary artery disease [5] and arrhythmia [6]; and lung sounds to diagnose asthma [7], to separate the patients with chronic obstructive pulmonary disease (COPD) from particularly the smokers who are potential COPD patients [8]. The DBN trains the model considering the capabilities of reaching the global minimum for the supervised training and high classification performances with fast, greedy, and layerwise pre-training [3]. ...

... In this study, 3 different open source data sets [7][8][9] were used and the results were examined. The audio data were separated as 1s time data and their attributes were extracted. ...

... [12] Tager enerji operatörü, kısa-zamanlı Fourier dönüşümü değerlerinden elde ettikleri göğüs seslerinden hesapladıkları kürtosis, entropi, ortalama çaprazlama değerleri yardımıyla, Altan vd. hırıltı seslerinin üç boyutlu ikinci derece fark haritalarını bölütlenmesini derin öğrenme algoritmaları kullanarak kronik obstruktif akciğer rahatsızlığının seviyelerinin teşhisini [13], Ulukaya vd. [14] önerdikleri Q değişkeni ayarlanabilir dalgacık dönüşümünü baz alan oranlı genişleme dalgacık dönüşümüyle çıkardıkları özniteliklerle hırıltı seslerinin tespitini ve astım hastalığının analizini gerçekleştirmişlerdir. ...