Why AI Is the Solution to Radiology Data Problem

Radiologists are under increasing pressure to read and interpret hundreds of images a day. One study estimated that the average radiologist needs to interpret one image every 3–4 seconds in an 8-hour workday to meet workload demands.

As reimbursement rates for image reviews are declining, the burden on radiologists to read and interpret more images is only increasing, forcing them to increase their daily productivity if they are to stay afloat. Image reading reimbursement rates can vary drastically depending on whether the read was a "limited" or "complete" and by the type of insurer.

While the increased number of imaging reads is creating massive amounts of data for researchers to study, it is also leading many providers to feel burnt out as well as taking time away from patients. Instead of counseling patients on the results of their images and care options, radiologists are spending time in front of a computer.

The pressure to read images quickly also can impact the accuracy with which radiologists are interpreting results. Because radiology involves visually interpreting results and making decisions around uncertainty, errors are inevitable. But under constrained conditions, false positives and negatives can be even more prevalent. Misdiagnoses can be harmful and costly to patients both physically and emotionally.

Artificial intelligence is seen by many as the silver bullet to radiologists’ workload demands and data problems. Deep learning algorithms are taught to take a large dataset of medical images, read, and interpret the results as a radiologist would. The algorithm is trained to sort through regular and abnormal results, identifying suspicious findings, tumors, and brain bleeds.

Deep learning methods can be particularly helpful in identifying rare or difficult to diagnose diseases as they are built on large datasets containing images of these diseases. While a trained radiologist might only see one case of a rare disease in the entirety of his career, AI was trained off robust datasets to identify these unique conditions quickly, increasing its diagnostic accuracy compared to a physician.

Every day, new research reports an algorithm that was developed to read and interpret different radiological images from CT scans to MRIs. These studies often report that the algorithm can more accurately and quickly identify abnormal results compared to radiologists.  

AI, therefore, has the potential to improve diagnostic accuracy, ease the burden for providers, and overcome challenges to limited reimbursement.  

Improving the Accuracy of Imaging Reads

False-positive and false-negative image reads can significantly impact patients. Repeated and unnecessary tests can be emotionally draining as patients wait for a diagnosis, and medical bills from repeated testing can stack up, creating substantial cost burdens for patients.

Artificial intelligence has repeatedly demonstrated lower false positive and false negative read rates compared to radiologists. The improved accuracy means fewer patients are undergoing unnecessary testing and the health care system is incurring fewer avoidable costs.

Increasing the accuracy of image reading also means that patients in need of treatment can receive treatment sooner than before. Having a diagnosis earlier and starting treatment quickly after diagnosis can lead to better results for the patient.

Creating Time for Providers

Studies of artificial intelligence have shown that the algorithms can often read images faster than the average radiologist. Integrating artificial intelligence into routine clinical workflow would mean that more images could be read in less time. This more efficient process would increase the amount of time radiologists have, resulting in less burnout and more time with patients.

Rather than feeling overwhelmed with a need to read and interpret hundreds of images throughout the workday, radiologists could spend more time focusing on patient care. The routine tasks radiologists need to perform could be automated, giving radiologists time to spend on things computers cannot do: counseling patients, creating a care plan, helping patients understand their diagnosis and treatment options, and coordinating care with other providers.

Solving the Reimbursement Dilemma

Part of the reason providers felt pressure to read and interpret more images was because reimbursement for individual reads was declining. Providers need to read more images if they are to be reimbursed at the same rate.

Artificial intelligence studies have shown the algorithms can read and interpret images faster than providers, so using artificial intelligence in clinical practice would eliminate the burden providers feel to maintain an unrealistic workload. 

As the healthcare industry shifts away from fee-for-service to outcomes-based reimbursement models, leveraging artificial intelligence for routine tasks can help providers begin to focus on the quality rather than the quantity of services rendered. This change in mindset would allow providers more time to focus on counseling patients, emphasizing high-quality, value-based care.

___________________________________________________________________________________________________

About Pure Storage

Pure Storage (NYSE: PSTG) helps innovators build a better world with data. Pure's data solutions enable SaaS companies, cloud service providers, and enterprise and public sector customers to deliver real-time, secure data to power their mission-critical production, DevOps, and modern analytics environments in a multi-cloud environment.  One of the fastest-growing enterprise IT companies in history, Pure Storage enables customers to quickly adopt next-generation technologies, including artificial intelligence and machine learning, to help maximize the value of their data for competitive advantage. And with a certified NPS customer satisfaction score in the top one percent of B2B companies, Pure's ever-expanding list of customers are among the happiest in the world.