Post-operative glioblastoma multiforme segmentation with uncertainty estimation

Abstract

Segmentation of post-operative glioblastoma multiforme (GBM) is essential for the planning of Tumor Treating Fields (TTFields) treatment and other clinical applications. Recent methods developed for pre-operative GBM segmentation perform poorly on post-operative GBM MRI scans. In this paper we present a method for the segmentation of GBM in post-operative patients. Our method incorporates an ensemble of segmentation networks and the Kullback–Leibler divergence agreement score in the objective function to estimate the prediction label uncertainty and cope with noisy labels and inter-observer variability. Moreover, our method integrates the surgery type and computes non-tumorous tissue delineation to automatically segment the tumor. We trained and validated our method on a dataset of 340 enhanced T1 MRI scans of patients that were treated with TTFields (270 scans for train and 70 scans for test). For validation, we developed a tool that uses the uncertainty map along with the segmentation result. Our tool allows visualization and fast editing of the tissues to improve the results dependent on user preference. Three physicians reviewed and graded our segmentation and editing tool on 12 different MRI scans. The validation set average (SD) Dice scores were 0.81 (0.11), 0.71 (0.24), 0.64 (0.25), and 0.68 (0.19) for whole-tumor, resection, necrotic-core, and enhancing-tissue, respectively. The physicians rated 72% of the segmented GBMs acceptable for treatment planning or better. Another 22% can be edited manually in a reasonable time to achieve a clinically acceptable result. According to these results, the proposed method for GBM segmentation can be integrated into TTFields treatment planning software in order to shorten the planning process. To conclude, we have extended a state-of-the-art pre-operative GBM segmentation method with surgery-type, anatomical information, and uncertainty visualization to facilitate a clinically viable segmentation of post-operative GBM for TTFields treatment planning.

A Deep Learning Approach for Automatic Segmentation during Daily MRI-Linac Radiotherapy of Glioblastoma

“…Eliminating the comparison of networks trained on pre-and post-operative images, Gazit et al analyzed T1ce scans from 340 patients and developed a method for segmentation of post-operative glioblastoma lesions [21]. Their results-average DSC for resection cavity and T1ce enhancing lesions of 0.71 and 0.68, respectively-compare favorably with the results from this study-average DSC for resection cavity and tumor lesion 0.84 and 0.67, respectively.…”

“…Multiplied by 31 timepoints (1 pre-RT and 30 fractions), this is 15.5-31 h of labor per patient. Although software packages and networks already exist for automatic contouring of brain structures on MRI, almost all utilize pre-operative MRI (normal brain structures) [18][19][20], and although one published study exists for postoperative auto-segmentation of glioblastoma, it is on high-field MRI [21]. One of two current commercially available MRI-linacs is low-field MRI, which has its own unique challenges.…”

Improving the Generalizability of Deep Learning for T2-Lesion Segmentation of Gliomas in the Post-Treatment Setting

“…This study sheds light on the dichotomy between typical training sets utilized and utility for clinical implementation, offers insight into effectively leveraging the widespread availability of pre-treatment data with smaller amounts of post-treatment data, and demonstrates the benefit of incorporating relatively simple but effective modifications to training strategies to tailor T2-lesion segmentation of gliomas to effectively monitor response to treatment. Overall, our model achieved a performance that was on par with results from similar deep learning-based studies of segmenting gliomas post-treatment as shown in Table 4, while using a single MR contrast and minimal processing [10,[66][67][68][75][76][77][78].…”

“…The utility of deep learning models used in monitoring longitudinal tumor progression and treatment response [10,11] is directly dependent on the accuracy of these models to perform well on treated gliomas. Although a few more recent studies have achieved equivalent performance in segmenting treated gliomas [66][67][68][69], they still either require multiple (4) image contrasts as input to segment multiple tumor compartments simultaneously, necessitate multiple image preprocessing steps (i.e., co-registration/skull stripping), use very few post-operative patients for training and testing, neglect edema and infiltration seen on T2-weighted images, or report low Dice scores (<0.65).…”