This paper aims to present a Convolutional Neural Network (CNN)-based approach to classify kidney-related conditions, specifically tumors, cysts, and stones using coronal (lateral) and axial (cross-sectional) CT scan images. We used a dataset taken from hospitals in Dhaka, Bangladesh so that the model built by us could accurately discriminate different types of kidney diseases. We used the following categories of normal, cyst, stone, and tumor in this dataset with 9955 training images and 2491 testing images. To generalize the model, we searched for categorical variations using real kidneys instead of applying artificial data augmentation techniques. The accuracy of our model in classifying different kidney conditions is high and almost always finds the intended result which also encourages the use of CNN as a reliable and powerful tool for medical diagnosis. The results are presented through performance metrics of the model (precision, recall, confusion matrices) and future endeavors for clinical application. Major contributions include the use of a unique dataset about kidneys, high accuracy achieved by the CNN model with an innovative classification approach.