Automated imaging response evaluation system: Prototype software tool for standardized radiographic response assessment in glioblastoma clinical trials
Neuro-Oncology Practice
Ashley Teraishi , Francesco Sanvito , Catalina Raymond , Nicholas S Cho , Jason Norris , Sani Gandhi , Alexander R Cohen , Yashaar Hafizka , Audrey Luo , Hao-Wen Sim , Lucas P Wachsmuth , Sonia Yip , Eng-Siew Koh , Merryn Hall , David M Ashley , Mark A Rosenthal , Elizabeth J Hovey , Hui K Gan , Margaret O Johnson , Elizabeth H Barnes , John R Simes , Zarnie Lwin , Mustafa Khasraw , Noriko Salamon , Timothy F Cloughesy , Benjamin M Ellingson

Summary
The clinical management of patients with brain tumors, as well as the results of clinical trials testing new medications, require the evaluation of tumor changes on MR imaging. After tumor measurements are obtained, the AIRES software application can automatically apply the standardized RANO rules, to ultimately define when the tumor can be considered in a progression phase, as opposed to a stable phase, or in treatment response. This automated tool enables fast, reliable, and reproducible assessments, and therefore has the potential to reduce errors, streamline workflows, and lower costs by minimizing the resources required for training personnel and conducting reads. The standardized Response Assessment in Neuro-Oncology (RANO), modified RANO (mRANO), and RANO 2.0 criteria are utilized for imaging endpoints in glioblastoma clinical trials. After measuring tumor size at each timepoint, the crucial step of categorizing progression, stable disease, and response based on numerical thresholds is currently done manually. “Manual” RANO evaluations are time-consuming, error prone, and require dedicated training. Here, we present the features, installation, and usage of Automated Imaging Response Evaluation System (AIRES), a software application prototype that applies mRANO criteria using externally measured tumor sizes, clinical status, and corticosteroid dose. AIRES labels each timepoint with RANO categories and calculates progression-free survival and objective response rate. As a real-world application on clinical trial datasets, mRANO reads were performed on measurements from 367 brain MRI scans across 41 patients from NUTMEG (NCT04195139), a phase II multicenter clinical trial. Four readers with varying levels of experience achieved a percentage of fully correct RANO reads ranging from 49% to 76%, while AIRES reached 100%, compared to a fifth expert reader. AIRES-based reads were significantly faster than “manual” reads by the expert (37.1 ± 8.6 vs 107.7 ± 58.6 s, P < .0001). AIRES is a reliable and time-efficient tool for mRANO assessments and may be adapted for RANO 2.0.
Citation
Teraishi, Ashley, et al. “Automated imaging response evaluation system: Prototype software tool for standardized radiographic response assessment in glioblastoma clinical trials.” Neuro-Oncology Practice (2026): npag039.
