An EfficientNet-B4 Based Medical Deepfake Detection in Healthcare Image Analysis

Abstract: Deepfake technology, powered by artificial intelligence and deep learning, can now create highly realistic fake images, audio, and videos. While this innovation has many uses, it also poses serious risks in healthcare, where medical images like X-rays and CT scans can be altered. Such manipulation may lead to wrong diagnoses, affecting patient safety and hospital operations. This study focuses on building a reliable deep learning approach to identify fake medical images. Two datasets—knee X-rays and lung CT scans—were prepared, preprocessed, and labeled as real or fake. The EfficientNet-B4 model was then applied to detect manipulations. Results show that the model performs very well, achieving high accuracy in both datasets, especially in knee X-ray images. It also maintains a good balance between speed and performance, making it suitable for real-time use. Overall, the study demonstrates that EfficientNet-B4 is an effective solution for detecting medical deepfakes quickly and accurately.

Index Terms: Medical deepfake image detection, deep learning, EfficientNet-B4, convolutional neural networks.