Pengaruh Image Enhancement Contrast Stretching dalam Klasifikasi CT-Scan Tumor Ginjal menggunakan Deep Learning

Abstract

Kidney tumors are the third most common after prostate and bladder tumors, accounting for around 208,500 cases (2%) of all cancer cases globally. Renal Cell Carcinoma constitutes 85% of these cases, transitional cell cancer 12%, and other types 2%. In Indonesia, the incidence is 3 per 100,000 people, with a male-to-female ratio of 3.2:1. Ultrasound, CT scans, and MRI are used to detect, diagnose, and assess kidney tumors, with CT scans being crucial for evaluating complex lesions, both cystic and solid. This study uses the Image Enhancement Contrast Stretching technique to improve CT-Scan image quality for deep learning classification using the EfficientNet-B0 architecture. The dataset is split into training, validation, and testing sets in an 80:20 ratio. Hyperparameters include Adamax and RAdam optimizers with learning rates of 0.01, 0.001, and 0.0001. The highest performance was achieved using the Image Enhancement Contrast Stretching technique with the RAdam optimizer and a learning rate of 0.01, resulting in 100% accuracy, precision, recall, and F1-score. For the original dataset using the Adamax optimizer with a 0.01 learning rate, the highest performance was 99.12% accuracy, 98.28% precision, 100% recall, and 99.13% F1-score. This technique significantly enhances the performance of kidney tumor classification models.