Artificial Intelligence in Radiology

Abstract

Artificial intelligence is the next big thing in radiology. Artificial intelligence will change everything about the radiology field, from the way reports are reviewed to patient care. Patient care will be at the forefront of the artificial intelligence movement. Data patterns in studies analyzed by artificial intelligence algorithms will give preliminary reports to the radiologist. Along with preliminary reports, artificial intelligence will be able to keep track of a patients’ medical history. The future is here and only beginning.

Keywords: artificial intelligence, radiologyArtificial Intelligence in Radiology

Artificial intelligence: is it the future of radiology? Artificial intelligence is the top trending topic in radiology for 2018.

Artificial intelligence is moving from a testing phase to now being implemented into the medical field. Artificial intelligence ranges from algorithms trained to detect abnormalities on images to keeping a full medical history on a patient. According to Merriam-Webster dictionary, artificial intelligence is: (a) a branch of computer science dealing with the simulation of intelligent behavior in computers, and (b) the capability of a machine to imitate intelligent human behavior (2018). Dr. Schier stated “intelligence refers to the ability to solve problems” (2018). To understand artificial intelligence and what it means for the future, we have to understand how artificial intelligence works.

There are many different branches and types of artificial intelligence. King and King (2018) reported the following: A major component of AI is machine learning, which is a subfield of computer science that enables computers to learn without being explicitly programmed. This exciting technology incorporates computational models and algorithms that are similar to the structure and function of our brain’s biologic neural networks. These computational models are often referred to as “artificial neural networks. ” When these artificial neural networks process information (digital data) from numerous input flows, they have the ability to “learn” and alter their structure in much the same way the neurons in our brain are altered with memory (p501). Deep learning is an extensive network of machine learning, with data recognizing objects in images. Radiology is all about what is within the image. Neural networks are algorithms designed to analyze thousands and thousands of images, taking the data and organizing it to reveal patterns.

Combining physicians with artificial intelligence will impact the field of radiology. Artificial intelligence software generates a preliminary report of the scan, allowing the radiologist to review the scan and add to the report. Any critical findings will be reported and alarm the reading radiologist of an emergent case. Artificial intelligence software will review a study and decide where on the list the exam should be placed, whereas now PACS dictates the work list based off the time a scan is ordered. “Artificial intelligence not only provides your preliminary reports with findings but can also actively scan your report as you dictate for errors of context (right versus left discrepancy)” (Sana, 2018). Artificial intelligence will provide recommended follow-ups based on protocols, making it easier for the radiologist. Mohan states the pros of this approach would be dramatically improving the skewed ratio of the number of scans to the number of radiologists available (2018). When used correctly, artificial intelligence will be able to decrease the turn-around time for reports.

Although there are many good things about artificial intelligence, there are also flaws. At the present time, artificial intelligence algorithms are powerful but fragile and any noise in the image can disrupt the findings. When scans are not of technical quality due to incapacitated patients breathing or patients with large body habitus, artificial intelligence will not be able to analyze the study. Another flaw is earning the patient’s trust with a machine. Will they trust a machine to provide a report for potentially life-threatening results? Next, will insurance pay to have a machine analyze a costly study. Mohan poses a question to the artificial intelligence algorithm developers asking, “Are there softwares good enough to not “miss” anything “and if at all there is a “miss,” who is responsible- is it the software developers, the institute administration or the treating physicians who will follow the results to plan the patient’s treatments (2018). Radiologists and artificial intelligence machines still have guidelines to follow, and an ethical duty to the patient. Artificial intelligence has the capability to not only analyze a patient’s scan but also keep up with said patient’s medical history over time. There needs to be a balance between maintaining personal information privacy and the advancement of intelligent machines (Kohli & Geis, 2018). Patients can sign a waiver allowing a third-party to review and contribute to their health records (Kohli & Geis, 2018).

According to Miller and Brown, simple neural networks have been used in medicine since the early 1990s to interperet electrocadiograms and diagnose myocardial infarction (2018). Artificial intelligence in the radiology field is beginning to be used. Sana reports GE has partnered with MGH and IBM Watson’s partnership with Radiology Partners (2018). Artificial intelligence is the future of the radiology field. Will artificial intelligence eliminate the need for radiologists? No it will not, but artificial intelligence will greatly enhance a patient’s medical care. Radiologists can let delevopers know where and how artificial intelligence can be of benefit to them. Radiologist and computer developers can work together to really make artificial intelligence great and effective. Faster reoprt turn arounds, notifying the ordering physician, and keeping track of a patient’s care are all ways that artificial intelligence will benefit the radiology field and the patient.