Kidney Cancer Classification Using Deep + Machine Learning

In this project, I developed an intelligent system to classify kidney CT scans into two states: healthy kidneys and tumor-affected kidneys.

I relied on a combination of deep learning and machine learning to achieve high prediction accuracy.

The steps I followed:

Using three pre-trained CNN models: ResNet50, MobileNetV2, and EfficientNetB0 to extract features from medical images.

Training 10 machine learning algorithms on the extracted features, including SVM, Random Forest, XGBoost, and LightGBM.

Applying optimization techniques such as Data Augmentation and Stratified Cross-Validation to ensure generalization and prevent overfitting.

Analyzing the performance results using metrics such as Accuracy, F1-Score, Precision, and Recall, all of which reached approximately 1.00.

Using Grad-CAM to interpret the regions the model focuses on and confirming that the learning was clinically meaningful.

Working tools:

Python – TensorFlow – Keras – Scikit-learn – OpenCV – Pandas – NumPy

Result:

An accurate and reliable model capable of distinguishing between healthy and diseased kidney images with near-perfect accuracy, enhancing the application of artificial intelligence in the medical field.