Automatic assessment of glioma burden: a deep learning algorithm for fully automated volumetric and bidimensional measurement

Abstract

Background: Longitudinal measurement of glioma burden with MRI is the basis for treatment response assessment. In this study, we developed a deep learning algorithm that automatically segments abnormal fluid attenuated inversion recovery (FLAIR) hyperintensity and contrast-enhancing tumor, quantitating tumor volumes as well as the product of maximum bidimensional diameters according to the Response Assessment in Neuro-Oncology (RANO) criteria (AutoRANO).

Methods: Two cohorts of patients were used for this study. One consisted of 843 preoperative MRIs from 843 patients with low- or high-grade gliomas from 4 institutions and the second consisted of 713 longitudinal postoperative MRI visits from 54 patients with newly diagnosed glioblastomas (each with 2 pretreatment "baseline" MRIs) from 1 institution.

Results: The automatically generated FLAIR hyperintensity volume, contrast-enhancing tumor volume, and AutoRANO were highly repeatable for the double-baseline visits, with an intraclass correlation coefficient (ICC) of 0.986, 0.991, and 0.977, respectively, on the cohort of postoperative GBM patients. Furthermore, there was high agreement between manually and automatically measured tumor volumes, with ICC values of 0.915, 0.924, and 0.965 for preoperative FLAIR hyperintensity, postoperative FLAIR hyperintensity, and postoperative contrast-enhancing tumor volumes, respectively. Lastly, the ICCs for comparing manually and automatically derived longitudinal changes in tumor burden were 0.917, 0.966, and 0.850 for FLAIR hyperintensity volume, contrast-enhancing tumor volume, and RANO measures, respectively.

Conclusions: Our automated algorithm demonstrates potential utility for evaluating tumor burden in complex posttreatment settings, although further validation in multicenter clinical trials will be needed prior to widespread implementation.

Keywords: RANO; deep learning; glioma; longitudinal response assessment; segmentation.

Fig. 1

Example of manual vs automatic FLAIR hyperintensity segmentation (A) and enhancing-tumor segmentation (B) for the testing set in the postoperative patient cohort. (C) Examples of AutoRANO applied to automatic enhancing segmentations on the postoperative patient cohort.

Fig. 2

Volume and RANO measures are highly repeatable. Repeatability of (A) manual FLAIR hypertintensity volume, (B) automatic FLAIR hypertintensity volume, (C) manual contrast-enhancing tumor volume, (D) automatic contrast-enhancing tumor volume, (E) manual RANO, and (F) AutoRANO in the postoperative patient cohort. Training and testing sets are shown in light blue and dark blue, respectively, in B, D, and F. Line of identity (x = y) is shown in all plots.

Fig. 3

Automatically and manually derived volumes are highly correlated. Correlation between manually and automatically derived volumes for (A) FLAIR hypertintensity in the preoperative patient cohort, (B) FLAIR hyperintensity in the postoperative patient cohort, and (C) contrast-enhancing tumor in the postoperative patient cohort. Training and testing sets are shown light blue/red/gray and dark blue/red/gray, respectively. Line of identity (x = y) is shown in all plots.

Fig. 4

There was moderate interrater and manual–algorithm agreement for RANO measures. Agreement between RANO measures for (A) Rater 6 vs Rater 4, (B) AutoRANO vs Rater 4, and (C) AutoRANO vs Rater 6 in the postoperative patient cohort. Training and testing sets are light blue and dark blue, respectively, in B and C. Line of identity (x = y) is shown in all plots.

Fig. 5

There was high agreement between manually and automatically derived longitudinal changes in volume and RANO measures. Agreement between automatic and manual delta measures for (A) FLAIR hypertintensity volume, (B) contrast-enhancing tumor volumes, and (C) RANO measure in the postoperative patient cohort. Training and testing sets are shown light blue/red and dark blue/red, respectively. Line of identity (x = y) is shown in all plots.

Fig. 6

AutoRANO had higher agreement with manual contrast-enhancing volume than manual RANO measures. Correlation between manual contrast-enhancing volume and RANO measures for (A) manual RANO and (B) AutoRANO in the postoperative patient cohort. Training and testing sets are shown in light blue and dark blue, respectively in B. Linear fit is shown in all plots.