Şentürk, Zehra Karapınar2023-07-262023-07-2620222214-7853https://doi.org/10.1016/j.matpr.2022.10.223https://hdl.handle.net/20.500.12684/12288Cardiovascular diseases (CVDs) are the first cause of death around the world. High-accuracy heart disease classification can support the decisions about the patients of the doctors and increase the success of the treatment, and so it can increase the lifetime of the patients. Emerging deep learning (DL) studies enable automated classification of a broad range of problems without requiring harsh feature extraction processes. Automated CVD diagnosis is one of the fields in that DL is exploited. Since the death rates of CVD is very high, there is an urgent need to improve the detection rates of the current diagnosis methods. This paper proposes efficient preprocessing to enable dl-based classification of ECG signals, which is the most popular way of automated diagnosis nowadays, without analyzing the peaks in the signal for heart disease diagnosis. The signals were converted to scalogram images to perform classification via DL. Continuous wavelet transform (CWT) was exploited and CWT coefficients of one-dimensional ECG signals were arranged as the scalogram to form a two-dimensional image. A public dataset, Physionet was used as an ECG signal dataset. The proposed preprocessing was shown to be quite effective for dl-based ECG signal classification for arrhythmia (ARR), congestive heart failure (CHF), and normal sinus rhythm (NSR) with an accuracy of 98.7%. AlexNet-based DL solution with the proposed preprocessing performed more than 12% better than self-created CNN. The proposed architecture missed only 1.33% of the samples. This approach can easily be adapted to other biomedical signals and different DL models and may provide an opportunity for easy, fast, and accurate classification of crucial problems. © 2023 Elsevier Ltd. All rights reserved.en10.1016/j.matpr.2022.10.223info:eu-repo/semantics/closedAccessContinuous Wavelet TransformDeep LearningECGHeart Disease ClassificationAutomationBiomedical signal processingCardiologyComputer aided diagnosisComputerized tomographyDeep learningDiseasesImage classificationPatient treatmentWavelet transformsCardiovascular diseaseCauses of deathContinuous Wavelet TransformDeep learningDisease classificationECG signalsHeart diseaseHeart disease classificationHigh-accuracyLearning studiesElectrocardiogramsArticle2-s2.0-85142263870Q2