Histopathological Cancer Detection via Transfer Learning (VGG16 & VGG19)

A deep learning model based on VGG architectures that classifies cancerous cells with over 95% accuracy by analyzing microscopic histopathology images.

As a solution to the time-consuming and error-prone nature of traditional pathology examinations, a Computer-Aided Diagnosis (CAD) system that automatically diagnoses cancer from microscopic images has been developed. In this study, instead of training a model from scratch, the Transfer Learning technique was used to maximize success on the dataset. Method and Technical Architecture: Transfer Learning: By using the weights of pre-trained VGG16 and VGG19 models in the ImageNet dataset, the model was enabled to learn complex features (feature extraction) in medical images faster and more accurately. Data Preprocessing: The dataset was diversified using image augmentation techniques, and the risk of overfitting was minimized. Classification: The model performs binary classification of images as "Benign" and "Malignant". Success Metrics: Accuracy, Precision, Recall, and F1-Score metrics were analyzed to measure the clinical reliability of the model.