New research finds that the use of artificial intelligence (AI) along with imaging can help predict how effective immunotherapy will be for melanoma patients. A team led by researchers from Columbia University sought to validate-using radiomics and machine learning-the performance of quantitative computed tomography (CT) imaging features for estimating overall survival (OS) in patients with advanced melanoma treated with immunotherapy. The study authors developed a machine learning algorithm that evaluates patients' CT scans and creates a biomarker, or radiomic signature, that correlates with patient outcome.
"There is a critical clinical need for early radiographic markers of treatment efficacy in patients who are being treated for metastatic melanoma. This need is evident in patients treated with immune checkpoint inhibitors, which have been used successfully to enhance patients' immune response against cancer," stated Laurent Dercle, MD, PhD, Associate Research Scientist in the Department of Radiology at Columbia University Irving Medical Center, and one of the study's co-authors.
"This is due to atypical patterns of response and progressions to these new drugs," Dercle continued. "Pseudoprogression is one example of an atypical response, meaning that the cancer would appear to be progressing using our current methods of measuring tumor growth to assess treatment response, when in fact it is responding to the treatment."
Study Details
The study used radiomics and machine learning to retrospectively analyze CT images obtained at baseline and first follow-up and their associated clinical metadata, the authors wrote in JAMA Oncology (2022; doi:10.1001/jamaoncol.2021.6818). Data were prospectively collected in the KEYNOTE-002 and KEYNOTE-006 multicenter clinical trials.
Research included 575 patients with a diagnosis of advanced melanoma who were randomly assigned to training and validation sets, according to the authors, who collected data for the study from Nov. 20, 2012, to June 3, 2019. Data was analyzed between July 1, 2019, and Sept. 15, 2021.
Overall, the performance of the signature CT imaging features for estimating OS at the 6-month post-treatment landmark in patients who received pembrolizumab was measured using an area under the time-dependent ROC curve (AUC).
A random forest model combined 25 imaging features extracted from tumors segmented on CT images to identify the combination that best estimated OS with pembrolizumab in 575 patients. The signature combined four imaging features, two related to tumor size and two reflecting changes in tumor imaging phenotype. In the validation set-287 patients treated with pembrolizumab-the signature reached an AUC for estimation of OS status of 0.92 (95% CI: 0.89-0.95).
The findings suggest that "the radiomic signature discerned from conventional CT images at baseline and on first follow-up may be used in clinical settings to provide an accurate early readout of future OS probability in patients with melanoma treated with single-agent programmed cell death 1 blockade," the authors wrote.
Ultimately, the study's findings were not particularly surprising, said Dercle, noting that "the most striking result was that the signature could be used to identify pseudoprogressors, since it is estimated a favorable overall survival in 82 percent of patients whose tumors were categorized as pseudoprogression at month three."
The team's next research steps are to "move the field forward, and to go one step closer to using AI in clinical practice," he said, adding that "we are prospectively validating our signatures and applying them to new clinical trials to understand how they can be used."
Dercle foresees radiomics being used to measure the effectiveness of numerous treatment options for many different cancers in the near future. For example, he said a physician treating a patient with metastatic melanoma could potentially determine that a treatment isn't working and switch to another therapy early in the treatment process, instead of waiting for the tumor to grow bigger or new tumors to appear. "This could save a patient [from] months of ineffective treatment and improve the overall outcome."
Dercle also predicts that radiomics will become part of a precision medicine approach to cancer treatment, "by helping physicians choose the most appropriate therapy for each individual patient and modify it more quickly when it's not working. And the amazing thing about it is that it doesn't require more tests or more scans. We are simply applying computing power to analyze the medical images that patients are already getting."
AI is "transforming the field of radiology," Dercle noted, "but it's important to understand that AI will be used in conjunction with human decision making for narrow and specific tasks, rather than as a replacement for radiologists."
Mark McGraw is a contributing writer.