Automated Tuberculosis Drug Resistance Prediction Using Computerized Tomography Images: A Deep Learning Framework

Abstract

The goal of this study is to use genetic data to predict TB medication resistance and sensitivity using a deep learning Convolutional Neural Network (CNN). The current methods for identifying drug-resistant TB strains are time-consuming and involve cultivating germs while they are exposed to medications. It may take a long time before you get the results of this. Furthermore, these methods aren't always precise and might generate false positives or negatives. To remedy these deficiencies, the proposed study would provide a more rapid and accurate method for identifying TB strains with medication resistance. When it came to making predictions, the suggested method was 97.27 percent accurate. Improved patient outcomes after treatment could result from using this model in clinical settings to detect drug resistance in TB more rapidly. The following terms are used as index terms: TB, drug sensitivity, drug resistance, deep learning, CNN, CT.