AI outperforms radiologists at predicting cancer treatment response based on imaging
Artificial intelligence may be more suited for predicting cancer treatment responses based on imaging findings than radiologists.
According to new research published in Frontiers, AI can more accurately predict how lung cancer patients will respond to treatment. Unlike many studies that compare the detection capabilities of radiologists and AI, this latest work puts a spotlight on how each uses imaging to determine treatment efficacy. The findings have the potential to change how treatment decisions are made in the future, experts involved in the analysis contend.
The group conducted a meta-analysis of 11 retrospective studies to determine whether AI could match the performance of radiologists in predicting treatment effectiveness based on imaging alone. Two reviewers assessed the studies’ data, calculating the pooled sensitivity, specificity and accuracy for both AI and radiologists.
Following numerous comparative analyses, the group determined that AI achieved a superior performance, yielding a sensitivity of 0.90, specificity of 0.80 and accuracy 0.90, respectively. Risk difference favored AI by 0.06 for sensitivity and 0.04 for specificity, but these results were more notable depending on the modality used to predict treatment responses, the group noted.
The largest difference in performance between AI and radiologists was observed when CT and PET/CT scans were used; AI showed superiority using MRI scans as well, but the difference was less notable. The group described AI’s edge over radiologists as “modest but [with] statistically significant superiority.”
“Artificial intelligence has emerged as a promising adjunct to radiologist interpretation in oncology imaging,” Nehemias Guevara Rodriguez, with the Deparment of Medicine, Division of Hematology Oncology and Bone Marrow Transplants at Saint Louis University, and colleagues suggested. “However, generalizability is limited by retrospective study dominance, incomplete demographic reporting, lack of regulatory clearance, and minimal cost-effectiveness evaluation," they added.
The group acknowledged that their analysis is subject to limitations. They suggested that future prospective studies that incorporate “explainable AI, equity assessments, and formal economic analyses” would be beneficial in determining AI’s specific role in managing treatment decisions.
Read more about the findings here.
In addition to her background in journalism, Hannah also has patient-facing experience in clinical settings, having spent more than 12 years working as a registered rad tech. She began covering the medical imaging industry for Innovate Healthcare in 2021.