MRI with Artificial Intelligence Can Improve Glioblastoma Prognosis, Treatment

Analyzing MRI brain scans with an artificial intelligence (AI) tool can help providers improve prognosis and treatment for patients with glioblastoma.

By using a trained convolutional neural network, radiologists can measure the amount of skeletal muscle present in the heads of patients with glioblastoma more accurately and quickly than a provider could alone, according to new research. Results from a study presented at the National Cancer Research Institute Virtual Showcase reveal that these measurements are an indicator of the patient’s overall condition and could be used to predict their length of survival.

Glioblastoma is particularly aggressive – on average, patients survive between 12 months to 18 months post-diagnosis. Fewer than 5 percent live past five years. A condition called sarcopenia – a degenerative loss of skeletal muscle – can exacerbate this situation, making it harder for these patients to tolerate surgery, chemotherapy, or radiation and setting them up for adverse events, early treatment discontinuations, faster disease progression, and death.

“Finding a better way to assess patients’ physical condition, general well-being and ability to carry out everyday activities is important in glioblastoma, and, indeed, in many cancers,” said Ella Mi, M.D., clinical research fellow at Imperial College London, “because, at present, it’s often evaluated subjectively, resulting in inaccuracy and a high degree of variability depending on who is looking at it. So, indicators that can be assessed objectively, such as measures of sarcopenia, are needed.”

Related Content: AI-Assisted MRI Could Eliminate Pre-Treatment Surgery for Brain Tumor​

Realistically, some patients fare better than others, Mi said, and being able to objectively assess an individual's physical condition could help inform decisions on treatment, diet, and exercise, potentially improving his or her prognosis.

Consequently, said David Harrison, M.D., pathology professor at the University of St. Andrews and chair of the NCRI Cellular Molecular Pathology Initiative, getting an accurate picture of a patient's physical condition is crucial. But, to date, making those evaluations has not be easy.

“It has been very difficult to develop objective measurements of patients’ performance status. This is where AI can help,” said Harrison, who was not involved with this project. “At this stage, [Mi and her colleagues] show that there is an association between the temporalis muscle size and a patient’s frailty and how their disease progresses.”

To reach this goal in determining whether measuring skeletal muscle could help with glioblastoma treatment, Mi and her colleagues from the Computational Oncology Group at Imperial, examined 152 brain MRI scans from 45 patients who were diagnosed with glioblastoma between January 2015 and May 2018. This is the first study, in any type of cancer, to apply AI to sarcopenia measurements as a way unearth significant associations with clinical outcomes.

The scans, taken as a part of routine clinical care during diagnosis and follow-up, looked at cross-sections of the head, and Mi’s group focused on the temporalis muscle, the broad fan-shaped muscles on the sides of the head. These muscles are used for chewing, and they have also been identified as an accurate way to estimate the body’s skeletal muscle mass.

“We realized that sarcopenia could be identified by quantifying muscle in cross-sectional imaging that cancer patients routinely undergo,” she said. “This would allow for opportunistic screening of sarcopenia as part of cancer care without additional scanning time, radiation dose or cost.”

For more coverage based on industry expert insights and research, subscribe to the Diagnostic Imaging e-Newsletter here.

Using their convolutional neural network, the team identified and quantified the cross-sectional area (CSA) at the thickest part of the temporalis muscle. They, then, analyzed the association of the CSA with overall survival and how long patients survived without disease progression.

Based on their analysis, they determined CSA could significantly predict both types of survival. Higher CSA pointed to the presence of more muscle. When compared to patients with lower CSA, these patients had half the risk of dying, as well as a two-thirds reduction in the risk of disease progression at any given time. Specifically, the average overall survival for high CSA patients was 21.3 months, and it was 14 months for patients with low CSA.

The CSA effect was most strongly seen in patients over age 55 and men, the team said. But, the team is now expanding their work, incorporating additional factors, such as age, sex, tumor location, and genetic characteristics that can frequently predict chemotherapy responses. And, they are finding similar results.

“We have found that patients with high CSA had around 60-percent reduction in risk of death and 75-percent reduction in risk of disease progression compared to patients with low CSA, even when all these other factors are accounted for,” Mi said.

Ultimately, she said, these findings show that the presence of higher temporalis muscle prior to surgery, chemotherapy, or radiotherapy does correctly predict significantly longer overall and progression-free survival, potentially improving the accuracy of prognosis estimates and paving the way for better treatment plans.

“Previous evidence has shown that frail patients might benefit from shorter courses of radiotherapy or chemotherapy with temozolomide alone,” she said. “It could also guide therapeutic interventions for muscle preservation, including nutritional support, exercise therapy, and drugs.”

Additional work is already underway, she said, to validate these findings in a different patient group. Mi and her colleagues are also conducting investigations to characterize changes to the temporalis muscle during follow-up for these glioblastoma patients, and they hope to establish their own repository for glioblastoma brain MRI scans in collaboration with other academic centers.