COVID-19 Classification through Deep Learning Models with Three-Channel Grayscale CT Images

COVID-19, an infectious coronavirus disease, has triggered a pandemic that claimed many lives. Clinical institutes have long considered computed tomography (CT) as excellent and complementary screening method to reverse transcriptase-polymerase chain reaction (RT-PCR). Because of the limited dataset available on transfer learning-based models become go-to solutions for automatic COVID-19 detection. However, CT images are typically provided in grayscale, thus posing challenge detection using pre-trained models, which were previously trained RGB images. Several methods been proposed literature converting grayscale (three-channel) use with deep-learning such pseudo-colorization, replication, colorization. The most common is where one-channel image repeated three-channel image. While this technique simple, it does not provide new information can lead poor performance due redundant features fed into DL model. This study proposes novel pre-processing medical utilize Histogram Equalization (HE) Contrast Limited Adaptive (CLAHE) create representation provides different each channel. effectiveness evaluated six other including InceptionV3, MobileNet, ResNet50, VGG16, ViT-B16, ViT-B32. results show significantly improves classification InceptionV3 model achieving accuracy 99.60% recall (also referred sensitivity) 99.59%. addresses limitation potentially improve early control disease. Additionally, be applied imaging tasks input, making generalizable solution.