Radiology departments are feeling the strain as workflow demands increase year by year – over 40 million imaging tests were carried out in 2017, helped by sophisticated systems supporting clinical expertise and gathering information along treatment pathways.
Recent media reports highlight the pressure to provide accurate and timely diagnoses, a theme that has inevitably led to clinical decisions relying more and more on advances in technology. Navigating masses of data presents challenges not only to medical personnel, but also on how sophisticated systems are organised to meet these demands.
Artificial Intelligence (AI) and machine-learning point the route to many of those opportunities in a similar way that the revolutionary CT scanner did when it was introduced in the 1970’s.
New software supports radiologists by enhancing decision-making; algorithm programs, an integral part of AI, are able to make sense of data more proficiently than humans in detecting early anomalies so that diagnosis can be hastened, allowing sooner treatment opportunities to be considered.
Processing vast amounts of data can only efficiently be done by technology; otherwise, radiologists and their integrated healthcare teams would be unable to cope with the workflow. Technology frees them from the burden of routine and burdensome effort so they can concentrate on the abnormalities.
It is this human interpretation that is so important in breast imaging: AI must gain equivalent trust if it is to be relied on for making life-changing decisions.
Author Ellen B. Mendelson, MD, Feinberg School of Medicine at Northwestern University in Chicago highlighted 4 specific areas that called for vigilance in the use of AI:
Diagnosis
Radiologists and indeed, other physicians still need to confirm diagnoses themselves due to the possibility of an error. The ability of AI to analyse images to a high level of proficiency will nevertheless, it was acknowledged, continue its inexorable progress as scientific advancements feed through.
Clinical Decision Making
Assessment frameworks are an essential part of clinical responsibility, as are the classification systems that are incorporated into AI data analysis. Measures of malignancy, however, must eliminate doubts about serious misdiagnosis - the opinion of a breast radiologist therefore remains overriding.
Patient Outcomes
An important driving force in the quest to improve patient outcomes is the desire for certainty in analysis. Whilst there is understandable emphasis on the degree to which AI should be restrained within safe borders, there is a commitment to exploit the advantages that machine learning and neural networks will bring.
The quality of predictability is expected to be a crucial factor as increasing amounts of breast imaging data is collected, even incorporating possibilities of identifying previously unspecified clinical associations.
Workflow
Of all the benefits forecast from the use of AI, the most acknowledged relates to the benefit that flows down to overburdened radiologists trying to ensure that the images they study give a true and certain insight into a patient’s evaluations.
Many hours of scrutiny and analysis can leave a radiologist struggling to make an assured diagnosis. AI can peruse the data with the same level of efficiency no matter what time of day it is.
