Improving the Detection of Breast Cancer Using AI

Breast cancer accounts for approximately 13% of all cancer deaths in Canadian women. 1 in 8 women are expected to develop breast cancer during her lifetime, and my aunt was one of them.

5 years ago my aunt was diagnosed with Stage 3 breast cancer. The survival rate at that Stage is 72% over a 5 year period. Luckily, my aunt is alive and doing well, but her journey has not been a smooth one.

Several months before her diagnosis, my aunt started noticing symptoms that aligned with breast cancer (she also had a family history of cancer) and got a mammogram. Back then nothing was detected on the mammogram, and it would not be until months later, that signs she had seen would get even more serious and she would get another test confirming her suspicions.

So what does this mean?

Breast cancer, just like any other form of cancer, has the highest survival rates when it is detected as early as possible. Delays in diagnosis give the disease the opportunity to spread to other parts of the body, meaning the patient must go through more treatment to effectively remove the cancerous cells.

What happened with my aunt and her late diagnosis caused her a lot of avoidable pain, but the most frightening part is that she is not alone. A study conducted in 2018 found that mammograms miss about 20% of all breast cancer cases.

Along with this, mammograms are notorious for their false-positive results. A false-positive mammogram looks abnormal despite the fact that there is no cancer present. Out of 1,000 women, about 100 are called in for additional diagnostic imaging, and of these 100 women, only 4 or 5 are ultimately diagnosed with breast cancer. These false-positive tests can cause severe anxiety and lead to the need for more tests and examinations, which cost time and money.

If mammograms are not as effective as we need them to be, what is an alternative?

A new, emerging alternative in breast cancer detection is the use of Artificial Intelligence. Previously, radiologists would examine the images produced by the mammogram and determine if there were any changes in the breast tissue. However, as we already know, this process has flaws due to human error.

Over the past few years, an AI tool that utilizes supervised learning has been trained on roughly a million screening mammography images, both cancerous and non-cancerous, to understand what healthy breast tissue looks like. Doing this allows it to recognize abnormalities that radiologists cannot due to its thorough study of what it should be looking for in the images.

Now just to clarify, this is a tool, meaning mammograms will be taken, analyzed with AI and then overseen by a radiologist in order to make sure the most accurate diagnosis is made. This tech is something that will not work alone, but rather in tandem with humans. AI is able to detect pixel-level changes in breast tissue invisible to the human eye, while humans use forms of reasoning AI cannot.

The results produced by the use of AI in breast cancer detection have been very promising. This powerful tool was able to better detect T1 cancers, which is defined as early-stage invasive cancer. AI detected 91% of T1 cancers, whereas radiologists only detected 74%. It also had better overall sensitivity, with 88.8% in comparison to 75.3% of radiologists alone.

AI is the future of breast cancer detection

With more accurate results and the ability to detect cancer earlier, implementing AI into our current mammogram system has the potential to save the lives of millions, improve the scalability of breast cancer screening around the world, and make sure that no one else has to go through what my aunt did.

How can you make a difference?

Educate yourself! Whether it is understanding what breast cancer is or being able to recognize symptoms, being knowledgeable can assist you in helping others. Here are some good resources: Centres for Diseases Control and Prevention, American Cancer Society and the Breast Cancer Organization.

Donate! There are many organizations working towards advancing cancer detection and treatment that need support. Here are a few of them: Canadian Cancer Society and The Princess Margaret Cancer Society.