Paper Title
CLASSIFICATION MODEL USING CONVOLUTIONAL NEURAL NETWORKS AND CONVEX HULL FOR THE STUDY OF RENAL LITHIASIS

Abstract
Renal lithiasis is characterized by the presence of stones in the kidney, being chronic and morbid. In this study, samples of computed tomography (CT) scans on renal lithiasis from patients with and without renal lithiasis were used, provided by the “Instituto Mexicano del Seguro social” (IMSS), León, Gto. The studies are classified by stone size: less than 10 mm, ranging from 10 mm to 19 mm, and larger than 20 mm. The studies were contrasted with the control sample of patients without renal lithiasis. For the classification model using convolutional neural networks (CNNs), two methodologies were employed: samples with augmentation techniques using Convex Hull contours and without Convex Hull to homogenize sample sizes. The CNNs used in this study are VGG16, ResNet 50, and Inception ResNet V2. The results obtained show better accuracy in the models that employ a Convex Hull contour, reaching an accuracy of 94% and precision of 93% for the Naive Bayes Multinomial statistical test, being the highest among all classifiers. Keywords - Artificial Intelligence, Kidney Stones, Automatic Classification, Convex-Hull, Precision, Accuracy.