Identifying Diabetes Mellitus through Deep Learning-Based DNA Sequences Devindha Permatasari, M. Isa Irawan
Sepuluh Nopember Institute of Technology
Abstract
Diabetes mellitus is a chronic disease that cannot be cured and is caused by a deficiency or absence of the hormone insulin. It is the leading cause of death worldwide, with an estimated 2-5 million lives lost to diabetes every year. Detecting and treating diabetes early is critical to preventing complications and reducing the risk of serious health problems. We propose a deep learning method to identify diabetes based on biomarkers found in DNA sequences. This method uses sequences of DNA data that are represented as spectrogram images. DNA sequences are converted into numerical values using an entropy-based mapping technique which is a fractional derivative of Shannon Entropy. We use the specgram function from the Matplotlib library which uses the Short-Time Fourier Transform (STFT) to generate a spectrogram image which is a three-dimensional plot with time, frequency and amplitude represented by a color scale. Spectrogram images of DNA sequences were then extracted and classified using deep learning methods. The results reveal that the deep learning method using the ResNet module achieves an accuracy of 91.3% in identifying diabetes.
Keywords: Diabetes Mellitus, Deep Learning, DNA Sequences
