Deepfakes have gained a lot of negative attention recently. Be it the hugely criticized DeepNude AI app which removes clothing from pictures of women or the FakeApp that swaps the faces of celebrities with porn stars in videos. Deep-learning algorithms are excellent at detecting matching patterns in images. This capability can be used to train neural nets to detect different types of cancer in a CT scan, identify diseases in MRIs, and spot abnormalities in an x-ray. While the idea of implementing deepfake AI for medical purposes sounds great, researchers don't have enough data to train a model -- simply because of privacy concerns.

Artificial intelligence (AI) and machine learning are driving a great deal of the healthcare innovation in precision medicine, according to a new Chilmark Research report. The report reveals achieving the full potential of precision medicine is impossible to realize without applying AI and machine learning. Specifically, leveraging advanced machine learning and deep learning technology can rapidly analyze large datasets that outperform clinicians and researchers. The concept of precision medicine is starting to become a reality due to new medical data from the All of Us research program, CAR-T therapies, increasingly accessible genetic testing, and other apps. As these new data-driven, personalized treatment plans begin to enter clinical practice in specialty care settings such as oncology and mental health, it is now time to assess the limits of current health IT ecosystems to broader clinical adoption, and where the opportunities lie for innovative solutions to bring precision medicine into the mainstream.

Deep learning is an important element of AI that's helping advance diagnostics and treatment, but it also remains relatively uncharted territory. The most successful applications of the tech to date have been in medical imaging, first author Fei Wang, PhD, and colleagues at Weill Cornell Medicine in New York, wrote in JAMA Internal Medicine. Other applications of the AI tech are vast, but scientists are still facing significant barriers. "The potential of deep learning to disentangle complex, subtle discriminative patterns in images suggests that these techniques may be useful in other areas of medicine," Wang and co-authors said. "Substantial challenges must be addressed, however, before deep learning can be applied more broadly."

Transfer learning from natural image datasets, particularly ImageNet, using standard large models and corresponding pretrained weights has become a de-facto method for deep learning applications to medical imaging. However, there are fundamental differences in data sizes, features and task specifications between natural image classification and the target medical tasks, and there is little understanding of the effects of transfer. In this paper, we explore properties of transfer learning for medical imaging. A performance evaluation on two large scale medical imaging tasks shows that surprisingly, transfer offers little benefit to performance, and simple, lightweight models can perform comparably to ImageNet architectures. Investigating the learned representations and features, we find that some of the differences from transfer learning are due to the over-parametrization of standard models rather than sophisticated feature reuse.