Case Western Reserve University works with NYU and pharmaceutical companies Bristol Myers Squibb, AstraZeneca to validate imaging-based solutions for predicting response to therapy for lung cancer patients

For artificial intelligence (AI) tools being developed at Case Western Reserve University to have impact in the fight against cancer, they’re going to have to be validated in rigorous human clinical trials.

That validation may be a step closer following two recent agreements among bioengineering pioneer Anant Madabhushi, a longtime collaborator at New York University, and select large pharmaceutical companies:

  • In April, Madabhushi entered into a contract with AstraZeneca (LSE/STO/NYSE: AZN), a global, science-led biopharmaceutical company that focuses on the discovery, development and commercialization of prescription medicines, primarily for the treatment of diseases in three therapy areas—oncology; cardiovascular, renal and metabolism; and respiratory and immunology. 
  • Earlier this year, Madabhushi inked a similar deal with United States-based Bristol-Myers Squibb Company (NYSE: BMY), a global biopharmaceutical company whose mission is to discover, develop and deliver innovative medicines that help patients prevail over serious diseases.

“This is an important step in not only validating our research, but in further advancing efforts to get the right treatment to the patients who will benefit the most,” said Madabhushi, the F. Alex Nason Professor II of Biomedical Engineering at Case Western Reserve and director of the Center for Computational Imaging and Personalized Diagnostics (CCIPD). “We have shown that our AI, our computational-imaging tools, can have the potential to predict an individual cancer patient’s response to immunotherapy.”

Recent research by CCIPD scientists has demonstrated that AI and machine learning can be employed with potential to predict which lung cancer patients will benefit from immunotherapy.

The researchers essentially teach computers to seek and identify changes in patterns in CT scans taken when lung cancer is first diagnosed, compared to scans taken during immunotherapy treatment.

The team has also been training AI algorithms to look at patterns from tissue biopsy images of cancer patients to identify the likelihood of a favorable response to treatment and is also looking beyond lung cancer. Researchers showcased these computational approaches for predicting immunotherapy response to gynecologic cancers at the 2020 American Society of Clinical Oncology (ASCO) meeting in May.

While immunotherapy has benefited many cancer patients, researchers are seeking a better way to identify patients who are mostly likely to respond to and derive the most benefit from those treatments.

Immunotherapy is a treatment that uses drugs to help the immune system fight the cancer, while chemotherapy uses drugs to directly kill cancer cells, according to the National Cancer Institute.

“One of the goals in any clinical trial is to choose patients who will actually benefit from the immunotherapy, and there is much more to learn by investigating how those biomarkers inform that selection,” Madabhushi said. “But the question has always been: ‘How do you actually identify a subset that will benefit most?’ We can help answer that question with the image-based biomarkers we are developing.”

Assessing immunotherapy response

Both AstraZeneca and Bristol Myers Squibb will provide the CCIPD with data—chest CT scan and/or digital pathology images—from completed clinical trials in which their specific immunotherapy drugs were tested on lung cancer patients.

Madabhushi is working with long-time collaborator Dr. Vamsidhar Velcheti, director of Thoracic Oncology at NYU Langone’s Perlmutter Cancer, who had previously worked in Cleveland.

photo of Dr. Velcheti
Dr. Vamsidhar Velcheti

“We believe this novel approach can be a significant improvement over traditional 2-dimensional and subjective evaluations of tumor responses using RECIST criteria,” Velcheti said.

RECIST is “response evaluation criteria in solid tumors,” the standard rules that define when tumors in cancer patients either improve, stay the same, or worsen due to various treatments.

While this new computational analysis by Madabhushi and Velcheti will be done retrospectively using already concluded clinical trial data—the goal is to demonstrate that the AI software may help to predict which patients could respond to treatment using prospectively defined algorithms and applying them to data.

“If we can show with these datasets and images that we can do that before a clinical trial, that would obviously have great value to us and to them—and to the cancer patients,” Madabhushi said.


For more information, contact Mike Scott at mike.scott@case.edu.