Developments in AI technology that stand to benefit the field of medical imaging
Gartner, Inc. anticipates the global artificial intelligence industry will be worth an astronomic $3.9 trillion in 2022. AI has already had a big impact on the healthcare sector and will continue to do so, especially as it expands its footprint in the field of medical imaging. A white paper from Signify Research estimated that the market for AI in diagnostic imaging alone will be worth more than $2 billion by 2023. That being the case, there is no time like the present to look forward and discover the benefits of machine learning in radiology and what is to be expected as it becomes increasingly commonplace.
Sorting the Data
New technology used in medical imaging has led to a significant increase in the amount of data and images collected — which requires more of radiologists’ time to go through and examine. The growth of 3D imaging and electronic medical records has made the job of a radiologist even more labor-intensive when it comes to interpreting data.
While having access to more data can aid in treatment and provide a more holistic view of a patient’s health, overworked radiologists can be a problem. In fact, a term, “observer fatigue,” has been coined to refer to the negative impact on diagnostic accuracy brought on by increased workloads and a decrease in the amount of time spent examining each image.
Adopting AI for data interpretation comes in different forms. Some institutions such as the Ohio State University Wexner Medical Center have chosen to use their own in-house research departments to develop algorithms for their medical imaging departments. Meanwhile, Wake Radiology puts outsourced AI to use at its outpatient imaging offices throughout the Raleigh-Durham area.
Wake Radiology uses EnvoyAI, which protects patients’ privacy by first anonymizing all data, then sends the data to the cloud, where it is analyzed according to an algorithm. The results are finally sent back to radiologists, drastically reducing their workload and helping to combat observer fatigue.
Further Uses
Of course, there is much more that can be done with AI in the field of medical imaging aside from simply interpreting data. Business efficiency can be increased by incorporating AI such as the Dell EMC platforms, which can increase security with facial recognition, assure safe drug interactions, consolidate workloads to increase the speed of applications, and more.
As AI technology progresses, it will be able to provide more comprehensive assessments of the images used in radiology. Currently, algorithms are created to look for very specific anomalies — tumors, or signs of pneumonia, for instance — but as machine learning develops, it could provide a full interpretation of a CT scan or a chest X-ray that picks up on molecular markers for prognostic and preventive purposes.
In October 2018, Aidoc received FDA approval for its AI product, which assists radiologists by analyzing CT scans for acute brain bleeds. This marked the first FDA approval of a deep learning platform in the medical imaging field that helps in diagnostics. Aidoc’s technology has already been successful at Sheba Medical Center in Israel, where it was first adopted, and now it could be of use in hospitals and medical imaging centers across the United States.
The Shiley Eye Institute at UC San Diego has found yet another use for AI in radiology. Its researchers developed AI technology specifically to be used in CT scans that alert technicians if a patient’s eye lens is in the scan. Lenses are especially sensitive to the radiation used in the scan, and avoiding the danger is essential in proper patient care.
The Future of AI in Medical Imaging
One of the next steps in AI is the development of Explainable Artificial Intelligence (XAI), or Transparent AI. Whereas much of current AI operates by the “black box” theory that the algorithms used to make decisions are too complex for humans to track, XAI also features an explainable model and an explanation interface that provides reasoning.
XAI — which is currently being developed by the military in targeting systems and provides great hope for driverless cars — is of particular use in medical imaging where doctors and radiologists need to know the reason for a diagnosis. When AI platforms identify an anomaly in an image that is too small for humans to see, it’s important that this can be explained for effective treatment and prognosis.
While AI is just starting to take hold in radiology, there are plenty of reasons to be optimistic for its future. Despite the fears of lost jobs when talk of adopting machine learning arises, there clearly is a need for the assistance it provides in the medical field — as well as plenty of room for it to be successfully implemented alongside the humans it helps rather than merely replacing them.