Case Western Reserve University scientists are developing artificial intelligence (AI) tools to help surgeons and oncologists identify the subtle but critical differences between a recurring tumor and damaged non-cancerous tissue on post-operative MRI scans of certain cancer patients.
The work is being led by Pallavi Tiwari, PhD, and Satish Viswanath, PhD. Both are faculty members in the Case Western Reserve School of Medicine and lead researchers in the Center for Computational Imaging and Personal Diagnostics (CCIPD) at the Case School of Engineering.
Tiwari, Viswanath and several collaborators were recently awarded a $1.15 million grant from the National Cancer Institute (NCI)'s Informatics Technology in Cancer Research (ITCR) program to pursue development and dissemination of the AI-informed tools.
Avoiding second surgeries
The potential benefit for doctors and their patients: Fewer unnecessary surgeries to remove suspect tissue which now can only be confirmed to be non-cancerous after initial therapy.
Doctors often end up performing those surgeries for one simple reason: Tissue that has been scarred and damaged--even killed--by chemotherapy or radiation resembles a recurring tumor on an MRI scan, the researchers said.
"They look very similar on the image, at least from what the human eye can perceive," said Viswanath, who specializes in colorectal cancers, while Tiwari focuses on brain cancers.
For a colorectal cancer patient, that can often mean getting a proctectomy (a portion of the rectum removed), a radical procedure that significantly reduces quality of life, Viswanath said.
So, until now, if you don't take the lesion out, you can't tell if it's a tumor. But you really don't want to keep hitting cancer patients with unnecessary surgeries--and that's especially true in both brain and colorectal cancers."
Pallavi Tiwari, PhD., Lead Researchers, Center for Computational Imaging and Personal Diagnostics (CCIPD), Case School of Engineering
Their proposed tool would harness the interpretive power of the center's deep-learning computers, which will use the AI tools being designed and developed in this project to tease out miniscule variations between the tumors and damaged tissue on MRI scans.
Those previously unseen variations differentiate tumors from dead tissue (known as necrosis, when most or even all of the cells in the tissue have died) or severely damaged scar tissue (known as fibrosis).
The research covers brain and colorectal cancer because they are similar in "terms of over-treatment," Viswanath said, referring to decisions by some surgeons to not risk a second surgery when it is actually necessary, or the earlier example of an unnecessary surgery.
Spreading the word
The NCI grant also calls for the researchers to begin making the tool available to other scientists, with an eye on future dissemination among clinicians.
"Dissemination of this information is a key to this grant," Tiwari said.
"The research community is starting to appreciate the importance of radiomics, and there is a lot of excitement. Hopefully, the next step is to really get this into the clinical community as well."
Radiomics refers to the method of extracting certain features from radiographic medical images using data-characterization algorithms. These features, when interpreted by the computer, could uncover disease characteristics that fail to be appreciated by the naked eye.
Other collaborators on the project include neuro-oncologist Manmeet Ahluwalia, MD (Cleveland Clinic); colorectal surgeon Sharon Stein, MD, (University Hospitals); imaging scientist Nicole Seiberlich, PhD (University of Michigan); Andrew Janowczyk (Research Faculty, CCIPD); and Anant Madabhushi, the F. Alex Nason Professor II of Biomedical Engineering at Case Western Reserve and director of the CCIPD.
This new work also meshes with previous projects under the auspices of the CCIPD and the Case Comprehensive Cancer Center at the School of Medicine, and will allow for the centers to better connect pathology and radiology, Madabhushi said.
"The CCIPD will now have three concurrent NCI/ITCR grants focusing on AI tools for cancer diagnosis and prognosis--two focused on AI in pathology," Madabhushi said.
"Ultimately, that synergy will allow for more precise understanding and prognosis of cancer, in turn, leading to the betterment of the cancer patient."
Some of that work is referenced in academic research, said Viswanath, the corresponding author of a recently published paper in Radiomics and Cancers. A second paper is due soon in Radiology.
This type of research and development of tools using AI and cancer treatment will be among the topics covered in a virtual conference this fall.
Madabhushi, Tiwari, Viswanath and several other faculty from Case Western Reserve and Cleveland Clinic are also hosting the 2nd Artificial Intelligence in Oncology (AIO) Symposium Oct. 19-20 under the auspices of the Case Comprehensive Cancer.