Breast Cancer Awareness: How Artificial Intelligence Is Improving Accuracy of Mammograms

When you think of artificial intelligence (AI), maybe your first thought is of popular voice assistants like Siri and Alexa. Or, maybe you think of customer support chatbots and driverless cars.

No matter what your first thoughts are, we cannot discount the one area of our lives that AI has drastically improved: our healthcare. Medical professionals are beginning to utilize machine learning to carry out a number of tasks, including completing complex surgeries and making better-informed decisions focused on top-quality patient care.

One particular medical area that artificial intelligence has improved is the interpretation and accuracy of mammograms. With Breast Cancer Awareness Month now underway, here are the great strides we’ve made with AI to improve test results and strive to detect breast cancer at the earliest stage possible.

What Are Mammograms & Why Are They Flawed?

Starting at age 40, it is recommended that women get a mammogram every 1-2 years to detect breast cancer at its earliest stages. This type of x-ray captures four images of breast tissue that is then analyzed by a radiologist to determine if there are any abnormalities that may indicate cancer.

Although mammograms have facilitated early detection of the most common cancer in women, the screening remains a flawed practice for a number of reasons. For one, doctors can misinterpret the results to be either a false-negative or false-positive. In the instance of a false-negative, doctors fail to detect a tumor from the scan and relay inaccurate information to the patient. In the case of a false-positive, doctors read the screening as cancer and are then proven wrong after additional testing. One last glaring flaw of the screening is that current mammograms cannot detect fast-developing and often deadly cancers that arise between annual visits.

Out of the 30 million mammograms performed each year in the U.S. alone, 50 percent of test results are false-positives, often requiring women to undergo additional (and expensive) testing for no reason. Healthcare spending due to these false-positive results amounts to $4 billion annually. Another 20 percent of these mammograms are false-negatives, which can lead to treatment delays and emotional distress for the patients involved. This misdiagnosis of breast cancer is a glaring reality that we shouldn’t ignore.

How Artificial Intelligence Improves Mammograms

Accuracy of Results
Companies and researchers are focused on the fact that inaccurate readings are a common flaw of mammograms and have developed software to address the problem. The Houston Methodist Research Institute in Texas utilizes its own AI program to interpret mammograms with a 99 percent accuracy rate. The program is also 30 times faster at analyzing the scans than a human doctor. French startup Therapixel is also making strides with its own AI software that the company claims can reduce the rate of false-positive results by five percent.

Proper Training for Medical Professionals
Artificial intelligence is also serving as an educational tool for doctors involved in mammogram interpretations. In a study published by the Journal of Medical Imaging, researchers found that radiologists are susceptible to context bias, meaning that they refer to their past mammogram diagnoses to make decisions on the results of new scans. To limit the influence of context bias on radiologists’ diagnostic skills, AI was used to analyze eye movements that will indicate and address any cognitive processes that may be negatively impacting mammogram interpretations.

Improved Efficiencies within Healthcare Systems
AI is also improving efficiencies within the healthcare system as it pertains to breast cancer diagnoses. With multiple versions of mammogram testing (think 2D and 3D imaging) being utilized, AI can automatically collect data from each mammography type on a per-facility basis or across multiple facilities within one healthcare system. This data collection then allows for analysis of multiple factors in mammogram procedures and results, such as radiation dosage and breast positioning for optimal screening. AI also stands to improve workflow issues to speed up diagnostic procedures and make use of desired machines that will put the patient first.

Final Thoughts

Before we say goodbye to another Breast Cancer Awareness Month, take time to encourage the women in your life get their annual mammogram if they meet the recommended age requirements. Our ability to use these screenings to detect the earliest possible form of breast cancer will only continue to improve thanks to the power and careful use of artificial intelligence.