Chest X-Ray Image Segmentation Using 2D V-Net Algorithm to Improve Diagnosis of Lung Disease

• Main Authors: Wahyuningrum, Rima Tri, Maulana, Firman, Kusumaningsih, Ari, Satoto, Budi Dwi, Sari, Amillia Kartika, Sensusiati, Anggraini Dwi
• Format: Conference Proceeding
• Language: English
• Subjects: Accuracy; Convolutional neural networks; Data models; Deep Learning; Image segmentation; Lung Disease; Measurement; Seminars; Tuberculosis; V-Net; X-Ray
• ISSN: 2769-5492
• DOI: 10.1109/ISITIA63062.2024.10667916

Various kinds of lung diseases can occur such as asthma, pneumonia, bronchitis, tuberculosis (TB), and many others. This disease is typically characterised by symptoms including wheezing, chest pain, shortness of breath, and chronic cough. The world has recently been impacted by the deadly COVID-19 outbreak. WHO stated that pandemic status was increasing with the number of cases reaching 118,000 infections and more than 4000 deaths in 114 countries. Therefore, detection tools are needed to find out which citizens are infected with this virus, so that they can suppress and reduce the growth rate of daily cases by immediately providing assistance. In this research, the method used to obtain the system algorithm with the best performance is a Convolutional Neural Network (CNN) using the V-Net algorithm as a segmentation method which was tested on the chest x-ray dataset. This method can be correlated with the Reverse Transcription-Polymerase Chain Reaction examination or known as RT-PCR. In this research we have obtained the best model that can produce accurate, fast and efficient segmentation using epoch 15, 2D V-Net obtained evaluation results of Dice Coefficient and IoU metric values of 0.95769 and 0.91882, with a testing time of 1 second on the Qatar University (QU) Kaggle dataset.