AI- Machine Learning

AI and Clinical Intelligence in Oncology Research

Artificial intelligence and machine learning are poised to revolutionize many human endeavors, including cancer research. Here, AI and ML offer a number of opportunities to better understand existing data and make predictions based on vast datasets.

Implementing these technologies in oncology research is not without its challenges. Yet many of these challenges are surmountable, and the results offer promise to patients, providers and clinical trial teams.

Top Challenges in Cancer Research

Oncology research poses a number of challenges, from detection and diagnosis to treatment and prognosis. Artificial intelligence offers ways to address several of these challenges.

Risk Assessment, Detection and Diagnosis

Risk assessment, detection and diagnosis are essential for early treatment of cancer, yet can be among the most difficult steps in the process. A study in Radiology by John A. Shepherd, Ph.D. and fellow researchers, for example, focuses on ways to use artificial intelligence to improve breast cancer risk assessment.

“Conventional methods of breast cancer risk assessment using clinical risk factors haven’t been that helpful,” Shepherd explains. By using artificial intelligence, however, the researchers were able to explore mammogram images more deeply in order to find ways to assess risk that focus on more than just breast density.

Breast cancers are not the only conditions that currently suffer from unhelpful methods of assessing risk. Predicting risk, detecting cancers and providing the right diagnosis are consistently challenging. As they are essential to patient outcomes, however, these challenges remain a focus of attention for AI and machine learning research in oncology.

Data Integration

Current limitations in data integration pose challenges across all clinical trials, including oncology. Yet data integration is essential for the optimal use of AI and machine learning, making it another significant area of study for oncology research and digital intelligence.

For example, in an article on the challenges of data integration, researchers Helen Le Sueur, Ian N. Bruce, and Nophar Geifman note that “differences in the way data were captured and represented across different studies” continue to pose challenges for clinical trial teams seeking to pool data from multiple sources. In some cases, they had no choice but to organize data by hand, as available digital tools still lacked the ability to translate across studies.

For cancer patients, management of their conditions often “requires care from a multidisciplinary team, leading to a lack of a single aggregated data source in real-world settings,” write Michael Grabner and fellow researchers in JMIR Cancer.

Grabner, et al. examined the use of data integration in research of non-small-cell lung cancer. While they found that the integrated data could be used in a number of ways, they also found it challenging to create that pooled data set due to differences in the data collected and emphasized by various specialists involved in the patients’ care.

Because massive data sets are used to train AI and machine learning algorithms, creating high-quality data sets is essential. These data sets can also help clinical trial teams better contextualize new research.

clinical trial teams

How Digital Intelligence Boosts Oncology Research

Artificial intelligence offers a number of possibilities when applied to cancer research, write Olivier Elemento and fellow researchers in an article in Nature Reviews Cancer. Opportunities to apply AI and machine learning in oncology research include:

  • Detection and diagnosis of cancers.
  • Classifying subtypes of cancers.
  • Optimizing cancer treatment for individual patients and conditions.
  • Identifying new therapeutic targets in drug research.

Artificial intelligence is already making a significant impact on cancer diagnostics, write researchers Claudio Luchini, Antonio Pea and Aldo Scarpa in the British Journal of Cancer. Use of AI-based devices is also showing promise in the diagnosis and treatment of breast, lung and prostate cancers.

In cancer diagnostics, AI plays various roles. One way AI is being used is to identify cancer in MRI images. The artificial intelligence is first trained on a dataset of images, then given new images to analyze. “The model finds the prostate and outlines cancer-suspicious areas without any human supervision,” says radiologist Ismail Baris Turkbey. While the AI cannot yet replace experienced radiologists, its development can help practitioners zero in on images most likely to be cancer.

While data integration continues to develop, its use already shows promise in some contexts. Continued efforts to improve data integration through artificial intelligence may further improve clinical trials.

Grabner et al., for instance, noted that despite the challenges of using data integration to research non-small cell lung cancer, “the availability of integrated clinical data from [abstracted medical records], health plan claims, and other sources of clinical care may improve the ability to assess emerging treatments.” Standardizing and sharing data poses a challenge, but it also creates opportunities for better application of AI and machine learning to oncology research.

AI and machine learning

Choosing the Right Digital Tools for Clinical Trials

Currently, artificial intelligence and machine learning lack the ability to stand in for healthcare providers or clinical researchers. Rather, these tools exist as a way to support the efforts and expertise of these professionals.

For example, applying AI as an initial screening of imaging tests prior to radiologist review may help streamline the process and provide an additional perspective. “Our hope is that AI will be able to reduce human errors and radiologist burnout,” says Gopal Vijayaraghavan, M.D., director of breast imaging services at the University of Massachusetts Medical School.

In order for AI and machine learning to fulfill this supplemental role in oncology research and patient care, the algorithms must be designed and trained effectively. The more data that is available for the training process, and the better that data is, the more effective the resulting algorithms will be.

Just as AI needs to learn from the data it receives, clinical trial teams will need to learn how to use AI and machine learning tools in their work. Currently, “there is a relative ignorance of the medical community related to AI and its methods and applications,” write Eduardo Farina and fellow researchers in a 2022 article in Future Science.

This lack of knowledge is understandable: Healthcare providers and clinical trial teams focus on their areas of expertise, which typically do not include artificial intelligence. Ensuring that teams have the information they need to understand AI and machine learning tools can help them make connections between their work, the tools’ capabilities and the limitations of existing technologies.

In many ways, artificial intelligence and machine learning are still in their infancy. As research continues to work out the ways in which these tools can be applied to oncology research and clinical trials, new applications for them will continue to be unlocked. The result may be a transformation of how clinical trial teams think about the data they collect in oncology research and how they apply it to the problems they face.

Want to stay up to date with our news?

To top