Pengaruh Contrast Limited Adaptive Histogram Equlization dalam Klasifikasi CT-Scan Tumor Ginjal menggunakan Deep Learning

Abstract

The human excretory system, comprising the kidneys, ureters, and bladder, plays a crucial role in maintaining overall body health by filtering blood and eliminating waste products, including water and toxins. However, kidneys are susceptible to various diseases, such as kidney tumors, which present a significant global health challenge, with over 430,000 new cases reported in 2020. This research focuses on using CT-scan imaging techniques to analyze and assess kidney tumors. The study employs the Image Enhancement Contrast Limited Adaptive Histogram Equalization (CLAHE) method to enhance the quality of Kidney Tumor CT-Scan images for deep learning classification using the MobileNetV2 Architecture. The dataset, consisting of 4,560 images, is divided into training, validation, and testing sets in an 80:20 ratio. Applying CLAHE with a clip limit of 20 and an 8x8 tile grid significantly improves evaluation metrics compared to non-CLAHE datasets, achieving an impressive f1-score of 99.56% and accuracy of 99.56%. This improvement is achieved using the Adam optimizer with a learning rate of 0.01. These findings underscore the efficacy of CLAHE in enhancing the model's performance in kidney tumor classification. They are particularly valuable for radiologists as they enhance diagnostic accuracy and efficiency, potentially reducing diagnostic errors and improving patient outcomes.