ijimsrInternational Journal of Innovation in Multidisciplinary Scientific ResearchCoVolume -2 | Issue - 1 | 2024MultidisciplinaryEnglishMAR-MAYNNY20240515Basic and Applied SciencePathological Lung Sound Detection using Deep Transfer Learning Pathological Lung Sound Detection using Deep Transfer LearningEnglishY17SmithaRaveendranEnglishNJitendraSonawanEnglishYGajanan K.BirajdarEnglishYMukesh DPatilEnglishYIJIMSR.2024.2110Lung sound analysis has gained prominence as a non-invasive method for diagnosing respiratory conditions. Recent development in deep transfer learning models have signified the potential to enhance the accuracy of lung sound detection, enabling early and accurate diagnosis. This paper presents an approach for lung sound detection using deep transfer learning techniques. A deep neural network architecture pretrained on a large external dataset and fine-tuned on a specialized lung sound dataset to leverage both general and domain-specific features. Firstly, input lung sound recordings are transformed into three spectrogram images. Two transfer learning models AlexNet and GoogleNet captures intricate patterns within lung sounds, differentiating between normal respiratory sounds and those indicative of pathological conditions. To evaluate the effectiveness of our approach, several comprehensive experiments are performed on a ICBHI 2017 dataset of lung sounds. The results showcase the improved performance of deep transfer learning model compared to conventional methods and standalone deep learning architectures with a significant reduction in false positives and false negatives. Highest detection rate of 94.64% is attained by GoogleNet model on ICBHI database.EnglishDeep learning, Lung sound, Spectrogram, AlexNet, GoogleNet, performance metricshttps://ijimsr.org/admin/abstract?id=27References1619