EASE: Clinical Implementation of Automated Tumor Segmentation and Volume Quantification for Adult Low-Grade Glioma

Abstract

The growth rate of non-enhancing low-grade glioma has prognostic value for both malignant progression and survival, but quantification of growth is difficult due to the irregular shape of the tumor. Volumetric assessment could provide a reliable quantification of tumor growth, but is only feasible if fully automated. Recent advances in automated tumor segmentation have made such a volume quantification possible, and this work describes the clinical implementation of automated volume quantification in an application named EASE: Erasmus Automated SEgmentation. The visual quality control of segmentations by the radiologist is an important step in this process, as errors in the segmentation are still possible. Additionally, to ensure patient safety and quality of care, protocols were established for the usage of volume measurements in clinical diagnosis and for future updates to the algorithm. Upon the introduction of EASE into clinical practice, we evaluated the individual segmentation success rate and impact on diagnosis. In its first 3 months of usage, it was applied to a total of 55 patients, and in 36 of those the radiologist was able to make a volume-based diagnosis using three successful consecutive measurements from EASE. In all cases the volume-based diagnosis was in line with the conventional visual diagnosis. This first cautious introduction of EASE in our clinic is a valuable step in the translation of automatic segmentation methods to clinical practice.