By Frieda Wiley, Texas Health Fellowship
A panel of experts convened to share their insights on leveraging artificial intelligence in cancer to impact patient outcomes during “How AI is revolutionizing cancer detection and care” at Health Journalism 2024.
Laurie Margolies, M.D., vice-chair for breast imaging at the Icahn School of Medicine at Mount Sinai, said she originally relied on computer-based detection and shared that she has observed fewer false positive and negative test results since implementing an AI tool that supports tissue analysis two years ago.
“It gives patients and the physicians reading them additional confidence because it reduces the anxiety clinicians feel when telling a patient they have cancer,” Margolies told the audience.
Using AI in this way also ameliorates the impact of physician shortages and burnout. In turn, it helps standardize cancer diagnostics, particularly among clinicians who generally screen a variety of tissues.
For example, some clinicians may specialize in radiology of a specific tissue, such as breast tissue. However, others may regularly evaluate other tissues in addition to the breast, such as knee tissue. Clinicians who generalize in a variety of tissue types may be more likely to miss cancer in certain tissue.
“AI can help bring all clinicians up to the expert level in diagnostics,” Margolies said.
AI helps clinicians distinguish between subtypes of breast tissue. This is a huge benefit, given how breast tissue varies among individuals, including fatty, white, dense, and gray tissue. A growing body of work has examined patterns in breast tissue, prompting researchers to examine breast density and family history to help identify at-risk patients.
In addition to mammography, AI offers a useful decision-support tool in ultrasounds by serving as a clinician’s “second opinion.” Instead of asking another clinician for an opinion on a suspicious spot, the doctor can ask AI.
Diagnostics aside, AI may offer its greatest impact in clinical trials, as only 5-8% of all cancer patients participate in clinical trials for cancer.
“The main reason why cancer clinical trial enrollment is low is because most people do not offer patients the trials,” said Ravi Parikh, M.D., assistant professor, medical ethics, health policy and medicine and associate director, Penn Center for Cancer Care Innovation.
Clinical research coordinators sometimes mistakenly deem patients ineligible for a clinical trial, further driving down enrollment, Parikh said.
Regardless of the data paucity, AI tools fill voids in data gaps and analytics in ways humans cannot accomplish. By using data captured in spreadsheets, AI can flag potential patients earlier than human users. One AI tool, natural language processing (NLP) extracts data from free text. Large language models (LLM) can be used to design a holistic foundational model that can be customized to carry out numerous tasks.
Looking forward, AI may one day help speed up clinical trials in addition to continuing to expand knowledge generation.
Frieda Wiley is an independent journalist based in the Dallas-Fort Worth metroplex. She was a 2024 AHCJ-Texas Health Journalism Fellow.
