Unsplash
I’ve never really thought about how important personalized medical care is until I was diagnosed with breast cancer myself - it gets real personal real quick. October is Breast Cancer Awareness Month and also my anniversary of being diagnosed and now cancer free! So, I felt it appropriate to focus on AI developments in early detection.
I am healthy, active and do not have the gene for breast cancer. Matter of fact, on the human medical doctor prediction scale it was a 3% chance my abnormal mammogram would come back anything but precancerous tissue. If it weren’t for me being proactive instead of waiting and seeing, I would be in a much different place health wise.
Breast cancer is the main source of cancer-related deaths among women, and unfortunately is very hard to detect. About one of every 10 cancers are misdiagnosed as not cancerous, meaning that a patient can miss out on important treatment opportunities.
Then again, the more mammograms a woman has, the almost certain it is she will see a false positive outcome. Following 10 years of yearly mammograms, about two out of three patients who don't have cancer will be informed that they do and be exposed to an invasive treatment, undoubtedly a biopsy.
Notwithstanding significant advances in hereditary qualities and present day imaging, a diagnosis catches most breast cancer patients off guard. For a few, the diagnosis comes past the point of no return. Thus, recognizing patients early has been a focal point of breast cancer research.
Personalized Data
As opposed to taking a one-size-fits-all methodology, today medical researchers are utilizing AI to create customized screening around a woman’s likelihood of becoming ill.
As opposed to manually distinguishing images from a mammogram that predict future malignancy, MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) and Massachusetts General Hospital (MGH) researchers prepared a deep learning model to interpret the images using AI.
Utilizing data from in excess of 90,000 mammograms, the model identified malignant growth the human eye was unable to recognize. It was also shown to predict if a patient would develop breast cancer as much as five years out.
The researchers’ model was altogether better at foreseeing danger than existing methodologies: It precisely put 31 percent of all cancer patients in its most elevated hazard classification, contrasted with just 18 percent for conventional models.
This shows personalized risk-based, instead of age-based, examinations can have a significant effect, and AI makes this possible.
“Since the 1960s radiologists have noticed that women have unique and widely variable patterns of breast tissue visible on the mammogram,” says Constance Lehman, MD, professor of radiology and chief of breast imaging at MGH.
“These patterns can represent the influence of genetics, hormones, pregnancy, lactation, diet, weight loss, and weight gain. We can now leverage this detailed information to be more precise in our risk assessment at the individual level.”
Synthetic Data for Real-World Results
Cancerous breast tissue has two key properties: heterogeneity, meaning a few regions are delicate and some are firm. Breast ultrasound elastography is an emerging imaging technique that provides information about a potential breast lesion by evaluating its stiffness in a non-invasive way.
At the crux of this procedure, however, is a complex computational problem that can be time-consuming and cumbersome to solve. But what if instead we relied on the guidance of an algorithm to streamline the steps to diagnosis?
Assad Oberai, USC Viterbi School of Engineering Hughes Professor in the Department of Aerospace and Mechanical Engineering, asked this exact question in his recent research paper, "Circumventing the solution of inverse problems in mechanics through deep learning: application to elasticity imaging."
During his research, due to limited data sources to create an accurate algorithm, Oberai trained a machine to interpret real-world images using synthetic data.
"If you had enough data available, you wouldn't," said Oberai. "But in the case of medical imaging, you're lucky if you have 1,000 images. In situations like this where data is scarce, these kinds of techniques become important."
Oberai and his team used about 12,000 synthetic images to train their machine learning algorithm, and achieved about 100 percent exactness on other synthetic images. They then tested it on real images to decide how precise it could be in giving a diagnosis, comparing outcomes against biopsy-affirmed judgments.
Oberai explained, “We had about an 80 percent accuracy rate. Next, we continue to refine the algorithm by using more real-world images as inputs.”
Rethinking How We Approach Diagnosis
There are two predominant points that make machine learning an important tool to advance the landscape for cancer detection and diagnosis. First, machine learning algorithms can detect patterns that may be opaque to humans. By observing and understanding many such patterns, the algorithm can make an accurate diagnosis.
However, does AI replace a radiologist's job in deciding diagnosis or treatment?
None of the researchers mentioned in this article anticipate an algorithm would replace doctors or technicians.
"The general consensus is these types of algorithms have a significant role to play, including from imaging professionals whom it will impact the most. However, these algorithms will be most useful when they do not serve as black boxes," said Oberai.
"What did it see that led it to the final conclusion? The algorithm must be explainable for it to work as intended."
Like other emerging technologies, AI breast cancer detection systems require a reliable assessment of clinical efficacy before they can be widely used. As we enter this exciting and rapidly developing new frontier, it will be important to validate AI systems for breast cancer screening on multiple, different image datasets that are representative of screening populations.
Additionally, the optimal interface between radiologists and AI systems needs to be identified, with broader stakeholder compliance identified for inevitable missed cancers and false positives.
These findings do not even begin to address the emotional side of living with being cancer free. It doesn’t quite feel free.
As a cancer survivor myself, I am always wondering what is going on inside my body. So, equipping doctors with AI-enabled cancer surveillance technology to reduce the fear and thoughts in the back of your mind seems to be a no-brainer.
