How AI Will Augment And Change Breast Imaging

A recent study published in The Lancet Oncology examined the ability to detect breast cancer on mammograms comparing the performance of two radiologists with multiple years of experience with that of AI and a single radiologist. In this first of its kind randomized controlled trial in Sweden, over 80,000 women were screened with mammograms to detect breast cancer.

Over a period spanning nearly 16 months, AI-supported mammography resulted in a slight increased rate of cancer detection of six per 1,000 screened women, compared to five per 1,000 when two radiologists read the study without AI. Furthermore, the false-positive rate (when a breast cancer is diagnosed on mammogram when none exists) was similar amongst both groups (AI-assisted and non AI-assisted screens) in the study.

Although the study did not explicitly measure the time a radiologist could save with AI, the authors postulated that radiologists could reduce their workload by as much as 44%.

Given the potential of AI to transform the way breast imaging is practiced and cancer is detected, it behooves us to understand how AI will do so. Here are five ways AI will enhance breast imaging and its practice.

Save More Lives

Data shows that breast cancer screening with annual mammography starting at the age of 40 saves lives, reducing mortality by up to 40% when compared to no screening. If AI-assisted mammography is able to detect even one more cancer per 1,000 compared to non-assisted AI as shown in the study in The Lancet Oncology, then this will result in thousands of more lives saved when considering millions of women will receive screening mammograms. Healthcare systems and physicians are always striving to provide the most effective care possible to patients, and AI could certainly help attain that goal.

Improve Radiologist Efficiency

The study in The Lancet Oncology suggests that radiologists could reduce their workload by 44% if they embraced AI as part of their workflow in reading mammograms. In Sweden, screening mammograms require a double read by two radiologists, but that is not the case in the United States where one radiologist can read a screening mammogram. Thus, the time savings in workload would not be as dramatic for practicing radiologist physicians in America, but would certainly still be present. Other published studies have also supported decreased workload for breast radiologists for screening mammograms in the presence of AI by up to 40%. If AI can decrease some time in reading screening mammograms, breast radiologists can improve their efficiency and spend more time completing other tasks such as with patient interactions and communications.

Bolster Radiologist Confidence

Detecting cancer on screening mammograms is not an easy task for radiologists. The density of breast cancer on a mammogram is similar to the density of normal breast tissue, making detection very difficult. It is of no surprise that nearly 20% of breast cancers are missed by screening mammograms according to the National Cancer Institute. Thus, if AI can aid in breast cancer detection, this should be welcomed by both radiologists and patients. In addition, AI can strengthen the confidence of a radiologist in instances where the radiologist may not be sure if a lesion is in fact a cancer but the AI tool diagnoses that same lesion as cancer. In this way, AI would add great value to a radiologist’s interpretation of a mammogram.

Decrease Burnout

Physicians, and breast radiologists in particular, are burning out at alarming rates. In a survey administered by the Society of Breast Imaging to breast radiologists across the globe, more than 78% of participants reported at least one measured dimension of burnout. It is well known that the increasing workload for radiologists in the last twenty years remains a major reason for radiologist burnout and stress. If AI can increase radiologist efficiency and allow for less time to be spent interpreting various studies such as screening mammograms, then AI also has the potential to decrease burnout and the emotional exhaustion that is pervading breast radiologists on a daily basis.

Cultivate Creativity

AI may also allow breast radiologists to think outside the box during their interpretation of images. AI has the potential to suggest certain diagnoses that may not have been considered by the interpreting radiologist when viewing a study. In this way, AI can foster creativity in a radiologist and allow a radiologist to consider a multitude of diagnoses when interpreting studies.

The Future

AI has enormous applications in medicine and could provide much utility for breast imaging. When discussing AI, Dr. Laura Heacock, a breast radiologist at NYU Langone Perlmutter Cancer Center, states, “Think of it as a tool like a stethoscope for a cardiologist.”