Cancer is a complex and devastating disease that has long been a global health challenge. In the case of colorectal cancer (CRC), it is currently the fourth most common cancer and the second deadliest. However, with the emergence of artificial intelligence (AI), a new research frontier has opened that could change the trajectory of cancer care as we know it. 

AI’s remarkable ability to analyze vast amounts of data and identify intricate patterns to predict, diagnose, and recommend treatment options shows promise to revolutionize the world of oncology. Exploring how AI is transforming cancer care across multiple disciplines – with a particular focus on colorectal cancer – will be a critical first step to understanding AI’s potential to save lives as we work to end CRC. 
 

AI in Cancer Detection and Diagnosis
 
There are several different ways that AI has shown potential in finding and diagnosing cancer, including medical imaging software, biopsy analysis, and categorizing risks. If combined, these tools could have a critical impact on health outcomes. 

Medical Imaging Software 

Simply put, AI algorithms make up a set of rules that determine how to solve problems. Medical imaging software is one application of an AI algorithm, programmed to analyze medical images, such as X-rays, mammograms, CT scans, MRIs, and colonoscopy images. The software then detects subtle abnormalities and potentially cancerous lesions. Like a second set of eyes in the cancer screening process, medical imaging software has the potential to lead to more precise detection and diagnosis.

AI-based models have become more commonplace in breast imaging procedures, with use across various medical establishments and at least five FDA-approved algorithms for breast-imaging detection and diagnosis. In one study, 211 mammograms were completed over a 10-year period. It was found that AI had a higher rate of cancer detection than radiologists alone (75% vs 67%), and, when combined, the study showed an even higher accuracy than AI detection alone (83% vs 72%).

In the case of colorectal cancer, two computer-aided polyp detection systems have been approved by the FDA – EndoScreener and GI Genius. A clinical trial showed the detection rate as 27% higher for endoscopists who used SKOUT, an AI polyp detection device with 510(k) clearance by the FDA (an approval to market and sell their product, but without full FDA approval). In another research study, artificial intelligence showed an increase in the number of smaller adenomas found during colonoscopy screenings. 

FDA-approved medical imaging software is also being used for lung cancer and a variety of other oncology applications. As the technology continues to rapidly change, it’s important to reflect on what steps need to be taken to ensure medical institutions are using the most up-to-date information and resources for their patients. 

Biopsy Analysis 

A biopsy is a procedure where a small sample of tissue, cells, or bodily fluids is taken from the body to determine the presence of abnormalities or disease. Biopsies are sent to pathologists for review, trained medical professionals who specialize in the study of diseases. As the medical field works to enhance the accuracy of cancer detection and diagnostics, AI may have a crucial role to play in assisting pathologists with tissue sample analysis. 

The FDA has authorized a biopsy analysis software, Paige Prostate, that helps to identify prostate cancer with a 70% reduction in detection errors. Additionally, researchers are currently looking at new ways to use biopsy software analysis to improve speed and accuracy within various oncology diagnostic settings. 

When a polyp is removed during a colonoscopy, your doctor will likely have it sent off to a pathologist for review. Biopsy analysis AI software is currently being researched in colorectal cancer pathology. One study showed a 99.17% accuracy when using AI to analyze multispectral biopsy images from CRC patients. 

By leveraging machine learning, AI research has demonstrated its potential to rapidly and accurately classify cells, identify specific tumor markers, and determine the stage and grade of cancers. This may help to expedite diagnosis and enhance precision medicine in the future. 

Categorizing Risks

Right now, the standard recommendation made by the U.S. Preventive Services Task Force for people at average risk is to start getting screened for CRC at age 45. However, certain medical and family histories may require earlier screening for CRC. 

Using AI, a patient’s risk of developing colorectal cancer can be more precisely determined through integration of various data sources, including genetic information, lifestyle factors, and medical records. With young-onset CRC on the rise, new AI technologies could be a valuable tool in helping to prevent colorectal cancer in the future. 

This same technology has the potential to be utilized across a variety of systems to help identify high-risk individuals who may benefit from targeted screening and prevent and detect cancer in its earlier stages. 

Researchers in Boston are currently using this technology to detect early signs of lung cancer with up to 94% accuracy. Similarly, recent research led by investigators at Harvard Medical School has used AI to identify people at the highest risk for pancreatic cancer up to three years before actual diagnosis. 

While more studies are needed to refine the role of AI in oncology, there is great promise in its ability to improve the detection and treatment of various cancers, with a notable amount of FDA-approved devices being used in clinical care today. 
 

AI in Oncology Precision Medicine & Treatment
 
While AI can accomplish a lot when it comes to preventing or catching cancer early, that is not its only potential in the world of oncology. AI is currently being used in research studies for targeted therapies, drug development, and treatment planning. 

Targeted Therapy

AI algorithms aid in identifying unique molecular patterns and genetic mutations associated with specific types of cancer, also known as biomarkers. By analyzing large-scale databases, AI has the potential to provide personalized treatment selection, matching patients with targeted therapies (FDA-approved and/or clinical trials) that are more likely to be effective. These personalized recommendations may minimize patient risks or negative side effects, and improve overall health outcomes. 

Drug Discovery & Development

The field of AI is accelerating the drug discovery and development process through machine learning algorithms that analyze vast databases of chemical compounds. The analysis then predicts potential effectiveness and identifies new drug candidates for specific cancer types. This AI-driven approach saves time by identifying potential drug candidates more efficiently and facilitates the development of new cancer treatments. 

Treatment Planning & Decision Support

Additionally, AI systems are being developed to assist oncologists with individualized treatment planning by integrating patient-specific data. One study for breast cancer patients showed an 83% clinically-accepted outcome in providing treatment planning for radiation using artificial intelligence. While AI can’t replace your oncologist, it could be an invaluable tool to assist in oncology care moving forward.  

In order to continue building on these technologies, it’s important to understand that AI algorithms are not set in stone. They require constant maintenance and an intricate understanding of the algorithm to develop software that can be relied upon in medical practice. While there is certainly a lot of promise for these AI applications moving forward, there are still setbacks that companies are facing as they work to crack the code on cancer treatment and AI. 
 

AI in Oncology Monitoring, Prognosis, Survivorship & Quality of Life 
 
When thinking about cancer, it can be helpful to categorize things in order of importance. Preventing cancer is a primary goal within the medical community. However, after that, early detection and precision treatment is key. Finally, there comes a time when looking at ongoing monitoring, prognosis, and survivorship plays a role in oncology planning. 

Researchers are currently investigating how oncology care can be improved through real-time monitoring systems, prognostic models, and tools to address survivorship. As the medical field continues to advance, new AI technologies could change the landscape as we know it. 

Real-time Monitoring

While not currently available in clinical settings, real-time monitoring has the potential to enable continuous monitoring of cancer patients. By analyzing data from wearable devices, electronic health records, and patient-reported outcomes, real-time monitoring may be able to facilitate the early detection of treatment-related complications, predict disease progression, and provide timely interventions.

Prognostics

AI-based prognostic models can use patient data to predict disease outcomes and survival rates. By synthesizing numerous variables, AI algorithms have the potential to enhance the accuracy of prognostic assessments. While still in its research phase, this technology shows promise of developing more personalized treatment plans and supportive care strategies in the future.

Survivorship & Quality of Life

AI tools also have the potential to assist in addressing survivorship issues by providing personalized care interventions to improve quality of life for cancer survivors. By analyzing patient-reported outcomes and a large library of medical research and information, AI systems are helping to address physical, emotional, and psychosocial needs in various research settings. This may aid clinicians in providing comprehensive and patient-centered care as the technology continues to advance towards more clinical settings. 
 

A Deeper Look at AI & Colorectal Cancer Prevention
 
When taking a more in-depth look at AI and its application for colorectal cancer prevention, it’s important to remember that colorectal cancer usually starts in the form of a polyp. While not all polyps are cancerous, it is necessary to remove polyps before they have the chance to manifest into cancer. This is why on-time screening is so important. 

According to a study published in the medical journal Endoscopy in 2012, it was observed that as high as 25% of polyps go undetected during a colonoscopy screening. Another study published in the medical journal Gut in 2019 investigated how AI systems affected the detection of polyps during a colonoscopy. The authors found that the computer-assisted systems improved the detection of potentially cancerous polyps compared to a standard colonoscopy alone (29.1% vs. 20.3%). A more recent study completed in 2022 showed that AI was able to cut missed polyps in half. 

Considering this data and everything else we have learned about AI, it is reasonable to assume that these technologies will continue to play a crucial role in the prevention of colorectal cancer and oncology treatment. 
 

Challenges, Considerations & Forward Thinking
 
As we continue to refine AI applications in our medical systems and beyond, they have the potential to become more efficient. However, there are still certain challenges to consider, including data privacy and security, algorithm interpretability, data transferability, health equity, regulatory approvals, and ethical considerations surrounding patient consent and AI decision-making. 

There is also the complication of the black box issue to be considered, a phrase that describes any AI technology that is so complex its decision-making process cannot be explained to or understood by humans. If we can’t verify the answer to the problem, how can we know it is true? 

“Things change rapidly in the world of healthcare and modern technology,” said David Fenstermacher, the Alliance’s Senior Director of Research and Medical Affairs. “One data element that changes has the potential to ‘crash’ the algorithm, making the results of the algorithm less accurate or inaccurate. If you don’t have people who are constantly monitoring and working to understand if the AI algorithms have been compromised, the software will degrade and not be useful anymore.” 

It’s important that industry leaders continue having conversations about these challenges and work to overcome barriers that may prevent AI from being used as efficiently and ethically as possible. 

“AI in oncology is not a standard of care – yet,” said David. “However, we’re not far away from seeing these tools more readily available in the healthcare industry and utilized in standard practice. mCode is a data model that has been developed to standardize the collection of oncology data. Additionally, the FDA has developed regulation surrounding AI algorithms in medical practice. We will need to see more of this type of standardization and guidance to improve the use of AI in oncology beyond imaging.” 

Despite these concerns, there is certainly a lot to be celebrated about the role AI plays in cancer detection, treatment, and monitoring. As ongoing research aims to refine issues still present with AI technology, collaborative efforts among healthcare providers, researchers, and AI developers will likely drive the adoption of new AI technologies in cancer care moving forward. 

The future of AI in oncology prevention and care looks bright, with the potential to further optimize patient care and improve health outcomes. As the Colorectal Cancer Alliance works to end CRC in our lifetime, we are committed to staying on top of emerging health technologies to better assist our community in overcoming this disease.