Dive Brief:

• The success of emerging artificial intelligence techniques in image analysis is fostering rapid growth in the use of deep learning for medical imaging, with the resulting computer vision systems able to perform some clinical interpretation tasks at the level of expert physicians, according to the authors of a paper published in Radiology, the official journal of the Radiological Society of North America.
• "Machine learning algorithms will transform clinical imaging practice over the next decade" by reducing diagnostic errors, improving patient outcomes, enhancing efficiency and lowering costs, the authors predict.
• The report is intended to be a blueprint for professional societies, funding agencies, research labs and others in the field to accelerate research on AI innovations that benefit patients, the authors said.

Dive Insight:

The effort to create a roadmap for industry collaboration on machine learning applications for diagnostic medical imaging and to prioritize research needs arose from an August 2018 workshop held at the National Institutes of Health in Bethesda, Maryland, that was co-sponsored by RSNA and the Academy for Radiology and Biomedical Imaging Research.

Imaging research laboratories are developing artificial intelligence systems to improve medical image reconstruction, noise reduction, quality assurance, triage, segmentation, computer-aided detection, computer-aided classification and radiogenomics.

A key area of focus is the reduction of diagnostic errors that may cause patient harm. Diagnostic errors play a role in up to 10% of patient deaths, and between 3% and 6% of image interpretations by radiologists contain clinically important errors, the report said.

The report looks at strategies to produce more publicly available, validated and reusable datasets for evaluating new algorithms and techniques. The authors advocate for the development of pre-trained model architectures for clinical imaging data and methods for distributed training that reduce the need for data exchange between institutions. They outline several key research priorities, including:

• new image reconstruction methods that efficiently produce images suitable for human interpretation from source data,
• automated image labeling and annotation methods, including information extraction from the imaging report, electronic phenotyping and prospective structured image reporting,
• new machine learning methods for clinical imaging data, such as tailored, pre-trained model architectures, and distributed machine learning methods,
• machine learning methods that can explain the advice they provide to users, and
• validated methods for image de-identification and data sharing to facilitate wide availability of clinical imaging datasets.

Earlier this month FDA outlined in a white paper its vision for a framework to regulate artificial intelligence algorithms that change based on real-world learning. 

"Artificial intelligence and machine learning have the potential to fundamentally transform the delivery of health care," now former FDA Commissioner Scott Gottlieb said in a statement.