Top Opportunities for Artificial Intelligence to Improve Cancer Care

Using AI in Cancer Detection

Recently, medical professionals have expanded the use of AI capabilities in cancer detection. At Tulane University, researchers discovered that  AI can accurately detect and diagnose colorectal cancer by analyzing tissue scans as well or better than pathologists.

The researchers gathered over 13,000 images of colorectal cancer from 8,803 subjects and 13 independent cancer centers in China, Germany, and the United States. Then, using images that technicians selected at random, the researchers built a machine learning program.

The program can recognize images of colorectal cancer, which according to researchers, is one of the most common causes of cancer-related deaths in Europe and the US.

After creating a performance measurement tool, the team of researchers compared the machine learning technique to the work done by the pathologists. The study indicated that the average pathologist scored around 0.969 for accuracy when identifying colorectal cancer, while the program scored 0.98, proving to be slightly more accurate than the manual data by pathologists.

According to researchers, there hope is that the study will encourage pathologists to use more prescreening technology to speed up diagnosis.

Not only can AI catch cancer earlier, but it can also improve detection accuracy. New York University researchers created an AI program trained to identify patterns among thousands of breast ultrasound images to aid physicians in diagnosing.

When tested on 44,755 completed ultrasound exams, the AI tool increased radiologists’ ability to accurately identify breast cancer by 37 percent. Additionally, the tool helped to reduce the number of tissue samples and biopsies necessary to confirm tumors by 27 percent.

“Our study demonstrates how artificial intelligence can help radiologists reading breast ultrasound exams to reveal only those that show real signs of breast cancer and to avoid verification by biopsy in cases that turn out to be benign,” study senior investigator Krzysztof Geras, PhD, said in a press release.

AI can also advance existing technology to improve patient outcomes. According to a recent study, medical professionals can use AI technology to quickly and accurately sort through breast MRIs in patients with dense breast tissue to eliminate those without cancer.

While mammography plays an important role in reducing breast cancer-related deaths, it is less sensitive in women with extremely dense breast tissue. Additionally, women with extremely dense breasts are three to six times more likely to develop breast cancer than women with almost entirely fatty breasts and two times more likely than the average woman.

According to researchers, by combining mammography capabilities and AI, the technology can significantly reduce the workload for radiologists, and improve patient outcomes.  

Predictive Models Assist Decision-Making

Predictive models have become a critical element in cancer treatment. By identifying risk factors, predictive models can determine an individual’s likelihood of developing certain cancers. Medical professionals can then encourage patients to engage in preventive care strategies.

According to University of Hawaii researchers, deep learning can distinguish between the mammograms of women who will later develop breast cancer and those who will not.

Not only can mammograms assist in detecting cancer but the technology can also predict breast cancer risk by measuring breast density. While denser breasts on mammography are typically associated with a higher risk of cancer, other unknown factors hidden in the mammogram could contribute to risk.

“Conventional methods of breast cancer risk assessment using clinical risk factors haven’t been that effective,” study lead author John A. Shepherd, PhD, said in a press release. “We thought that there was more in the image than just breast density that would be useful for assessing risk.”

The research team used a data set of over 25,000 digital screening mammograms from 6,369 women. More than 1,600 of the women developed screening-detected breast cancer, and 351 developed invasive breast cancer.

The team trained the deep learning model to find details in the mammogram linked to increased cancer risks. When they tested the deep learning-based model, it underperformed in assessing the risk factors for interval cancer risk. However, researchers said the model outperformed clinical risk factors, including breast density, in determining screening-detected cancer risk.

“The results showed that the extra signal we’re getting with AI provides a better risk estimate for screening-detected cancer,” Shepherd said. “It helped us accomplish our goal of classifying women into low risk or high risk of screening-detected breast cancer.”

According to the team, these findings could significantly impact clinical practices in which breast density alone guides many management decisions.

Making medical decisions regarding a cancer diagnosis can be particularly challenging because of the high costs of treatment and there is no guarantee a patient will be cured. By gathering data insights, University of Colorado Cancer Center researchers are working to create a decision-making tool for women who recently received a breast cancer diagnosis.

In 2019, assistant professor of surgical oncology Sarah Tevis, MD, and a team of researchers began gathering data by surveying women diagnosed with breast cancer to better understand the quality of life outcomes. The study showcases data gathered from 3- and 6-month reported outcomes from patients who had lumpectomies and mastectomies.

“We’re hoping to collate a large group of patient data to get a sense of what the average patient experiences three months, six months, a year after treatment. It gives us a foundation of data to be able to tell patients, ‘Here’s what other people in situations similar to yours have experienced,’” Tevis said in a press release.

“We’ve been really good about telling patients what to expect short-term, maybe in that first month after surgery, but beyond that haven’t been able to give them good, data-based information, how they might feel one year after surgery or even more long-term.”

A decision-making tool will assist recently diagnosed breast cancer patients in deciding whether or not they want to pursue treatment.

Developing Treatment Responses

Before pursuing treatment, physicians can use AI to predict how patients may respond to certain medications. Predictive information is critical for patients and physicians when deciding treatment options.

In a collaborative study, researchers analyzed biopsy samples collected from three large, randomized clinical trials. According to researchers, physicians could potentially use genetic test scores to create a personalized treatment for men with the most aggressive forms of prostate cancer.

Two-thirds of prostate cancer deaths occur in patients with high-risk prostate cancer. Therefore, balancing survival risk with quality of life is important to consider when making treatment decisions. According to researchers, biomarkers could potentially be used to create precision medicine, treatment guidelines, and identify who might benefit from different therapy methods.

“When a man is diagnosed with high-risk prostate cancer, we don’t have a widely accepted way to sub-classify their cancer and truly personalize their therapy, but we think we will in the near future,” lead author and professor at Harvard Medical School Paul L. Nguyen, MD, said in a public statement.

Nguyen and his team used the Decipher biopsy test, which analyzed the activity of 22 genes in prostate tumors, to create scores reflecting how aggressive a patient’s cancer is.

Researchers calculated Decipher scores using RNA extracted from archival biopsy samples collected from three major prostate cancer trials. They then analyzed how the scores were related to long-term outcomes.

With predictive analytics, the genetic signatures indicated which patients were most likely to develop distant metastases, die from their prostate cancer, and die from other causes.

Everyone is different and not all patients will have the same response to medications. Machine learning and predictive analytics can also prevent unnecessary side effects from cancer treatments that may not work for the individual

The Georgia Institute of Technology and Ovarian Cancer Institute researchers used machine learning algorithms to determine how their patients would respond to cancer-fighting drugs.

For the study, professor at the School of Biological, John F. McDonald, and his team created predictive machine learning-based models for 15 distinct cancer types, using data from 499 independent cell lines from the National Cancer Institute.

The models were then validated with a clinical dataset containing seven chemotherapeutic drugs, administered either singularly or in combination to 23 ovarian cancer patients. The model showed an overall predictive accuracy of 91 percent.

As AI capabilities continue to develop, researchers can find ways to implement the technology in enhancing cancer care and patient outcomes.