Image Enhancement Techniques on Chest X-Ray Images to Improve COVID-19 Detection

Abstract

The COVID-19 pandemic has claimed many lives. The diagnosis is made to prevent the spread of COVID-19. One of the diagnostic methods that have now become the gold standard is RT-PCR, but this method still has shortcomings in terms of accuracy so it is at risk of causing inaccurate decision-making. The use of medical imaging techniques such as CXR and chest CT scans in the diagnosis of COVID-19 is considered to be able to increase the accuracy of COVID-19 detection so that the risk of making inappropriate decisions can be minimized. Compared to a chest CT scan, CXR is considered superior in terms of price and availability so with these advantages the use of CXR is more effective in diagnosing COVID-19. However, it should be noted that in terms of performance, the chest CT scan far outperformed CXR. For CXR to be better utilized, image enhancement techniques are applied and combined with several classification algorithms. The experiments on two datasets showed that applying BCET (Balance Contrast Enhancement Technique) prior to classifying consistently outperforms other classification methods without enhancement techniques on other compared methods. Moreover, the SVM algorithm achieved the best classification results for all image types in both datasets by scoring the highest AUC compared to other algorithms.