Machine Learning based Medical Image Deepfake Detection: A Comparative Study

View on arXiv ← Back to list

Authors: Siddharth Solaiyappan, Yuxin Wen

Published: 2021-09-27 05:10:55+00:00

AI Summary

This paper investigates the detection of deepfakes in medical images, specifically focusing on the detection of artificially injected or removed tumors. It compares the performance of eight machine learning algorithms, including conventional methods and deep learning models, in distinguishing between tampered and untampered CT scans.

Abstract

Deep generative networks in recent years have reinforced the need for caution while consuming various modalities of digital information. One avenue of deepfake creation is aligned with injection and removal of tumors from medical scans. Failure to detect medical deepfakes can lead to large setbacks on hospital resources or even loss of life. This paper attempts to address the detection of such attacks with a structured case study. Specifically, we evaluate eight different machine learning algorithms, which including three conventional machine learning methods, support vector machine, random forest, decision tree, and five deep learning models, DenseNet121, DenseNet201, ResNet50, ResNet101, VGG19, on distinguishing between tampered and untampered images.For deep learning models, the five models are used for feature extraction, then fine-tune for each pre-trained model is performed. The findings of this work show near perfect accuracy in detecting instances of tumor injections and removals.

Key findings

The study achieved near-perfect accuracy in detecting tumor injections and removals, particularly when using deep learning models combined with data augmentation and region of interest localization. Deep learning models outperformed conventional methods, especially when augmented data was used.

Approach

The research evaluates eight machine learning algorithms (three conventional and five deep learning models) for detecting deepfakes in medical images. Deep learning models were used for feature extraction and fine-tuning. Different image preprocessing techniques, including localization and data augmentation, were explored to improve detection accuracy.

Datasets

LIDC-IDRI (untampered data) and CT-GAN dataset (tampered data).

Model(s)

SVM, Random Forest, Decision Tree, DenseNet121, DenseNet201, ResNet50, ResNet101, VGG19.

Author countries

USA

← Previous