UCM Professor Looks at ‘Multifaceted Radiomics’ as Way to Predict Risk of Cancer Metastasis
In fighting cancer, one of the challenges doctors face is how to effectively use radiotherapy to control the metastasis or spread of the disease without harming the patient in other ways. New research by a University of Central Missouri faculty member uses an innovative Artificial Intelligence (AI) technique to allow physicians to predict which patients are at low risk of distant metastasis in order to help minimize radiation’s severe side effects. The research conducted by Zhiguo Zhou, Ph.D., assistant professor of computer science, is titled “Multifaceted radiomics for distant metastasis prediction in head-and-neck cancer.”
It was published in the journal, Physics in Medicine and Biology, and subsequently reported in the July 2020 issue of Physics World. Zhou, who has explored AI in medicine for 10 years, joined the UCM faculty in 2019. He began working on this recently published study more than three years ago while serving in the Department of Radiation in oncology at the University of Texas Southwestern Medical Center in Dallas. One of his UT colleagues, Jing Wang, Ph.D., served as co-author on the journal article. Zhou said the research proposes a novel model for predicting metastasis in head-and-neck cancer after radiotherapy with “outstanding results.”
It is a study he believes could provide a general framework which could be extended to predict treatment outcomes for primary cancers in other parts of the human body. While the research now undergoes a validation process that involves a multi-institutional prospective study, Zhou is hopeful that it can be applied in clinical settings within the next two to three years. “Nowadays, radiotherapy has become one of the most important treatment methods in cancer therapy,” Zhou said. “The basic principle is to use radiation to kill the (primary) tumor and minimally deliver the dose (of radiation) to the surrounding normal organs. However, radiation is also harmful to the human body and it is very difficult to achieve this ideal situation.”
Physicians working with cancer patients must weigh different outcomes to address an optimal treatment plan, Zhou said. He believes the research he and others are doing will assist in this effort. “We think the solution is, if we can accurately predict the treatment outcome or response before radiotherapy, we can optimally make the treatment plan. This is the basic idea of why we need to do this research,” he said. As noted in the Physics World article, “As with cancers elsewhere in the body, early-stage cancers of the head and neck are treated using radiotherapy with increasing success. When treatment fails, it is often down to the growth of new tumors far from the site of the initial disease.
Predicting which patients are most likely to develop distant metastasis (DM) is vital so that low-risk patients can be spared the severe side effects that accompany the systemic treatments used to control cancer proliferation.” In seeking to develop a reliable model to predict distant metastasis, Zhou and his research collaborators utilized PET and CT diagnostic and treatment planning images of 188 patients with head-and-neck cancer that were obtained from different institutions. These patients had received follow-up consultations with their care providers, and the images were already seen by physicians. The researchers were able to extract from each patient 257 features that included intensity, geometric and textural characteristics in addition to other data related to patient age, gender, and progression of the disease.
Zhou said since 2012 a prediction model called radiomics has existed, which uses a characterization algorithm to extract data to help further understanding of a patient’s likelihood of experiencing the spread of cancer from the initial tumor to other organs or lymph nodes. His research, called “M-radiomics,” takes a multifaceted approach to radiomics in order to produce a more reliable and accurate prediction model. Three different algorithms are used in this process to help address challenges related to the integration of data from multiple imaging modalities, sensitivity-specificity optimization, and use of multiple data machine learning classifiers simultaneously. “In M-radiomics we can integrate these three challenges into one framework.,” Zhou said, adding that “the results are very promising for distant metastasis prediction in head-and-neck cancer.”
Throughout the summer, Zhou is continuing his research in M-radiomics, in addition to pursuing other interests related to artificial intelligence in medicine. His work at UCM also involves teaching an undergraduate introduction to biomedical informatics course and a graduate-level course on artificial intelligence. Supportive of student research, Zhou mentors and instructs five graduate students who work with him on research, each of whom is also doing their own research project. He said he is looking forward to getting more undergraduate students involved in research as his career as UCM progresses.
This article comes from information provided by Jeff Murphy, Assistant Director for Media Relations, Integrated Marketing and Communications for the University of Central Missouri.