Automatic segmentation, classification, and follow‐up of optic pathway gliomas using deep learning and fuzzy c‐means clustering based on MRI

Abstract

Purpose Optic pathway gliomas (OPG) are low‐grade pilocytic astrocytomas accounting for 3‐5% of pediatric intracranial tumors. Accurate and quantitative follow‐up of OPG using magnetic resonance imaging (MRI) is crucial for therapeutic decision making, yet is challenging due to the complex shape and heterogeneous tissue pattern which characterizes these tumors. The aim of this study was to implement automatic methods for segmentation and classification of OPG and its components, based on MRI. Methods A total of 202 MRI scans from 29 patients with chiasmatic OPG scanned longitudinally were retrospectively collected and included in this study. Data included T2 and post‐contrast T1 weighted images. The entire tumor volume and its components were manually annotated by a senior neuro‐radiologist, and inter‐ and intra‐rater variability of the entire tumor volume was assessed in a subset of scans. Automatic tumor segmentation was performed using deep‐learning method with U‐Net+ResNet architecture. A fivefold cross‐validation scheme was used to evaluate the automatic results relative to manual segmentation. Voxel‐based classification of the tumor into enhanced, non‐enhanced, and cystic components was performed using fuzzy c‐means clustering. Results The results of the automatic tumor segmentation were: mean dice score = 0.736 ± 0.025, precision = 0.918 ± 0.014, and recall = 0.635 ± 0.039 for the validation data, and dice score = 0.761 ± 0.011, precision = 0.794 ± 0.028, and recall = 0.742 ± 0.012 for the test data. The accuracy of the voxel‐based classification of tumor components was 0.94, with precision = 0.89, 0.97, and 0.85, and recall = 1.00, 0.79, and 0.94 for the non‐enhanced, enhanced, and cystic components, respectively. Conclusion This study presents methods for automatic segmentation of chiasmatic OPG tumors and classification into the different components of the tumor, based on conventional MRI. Automatic quantitative longitudinal assessment of these tumors may improve radiological monitoring, facilitate early detection of disease progression and optimize therapy management.

Automated tumor segmentation and brain tissue extraction from multiparametric MRI of pediatric brain tumors: A multi-institutional study

“…Several of these existing models are focused on a particular histology, such as optic pathway gliomas (OPGs). 17 , 20 The variety of histologies and tumor locations included in our model allowed us to build a model that can be used more generally. This may be specifically useful if the tumor diagnosis, in a prospective patient, is not yet known or if a dedicated model for a particular histology does not exist, as is typically the case in clinical contexts.…”

“…NEC label was generated to validate the performance of the model in distinguishing non-enhancing from enhancing areas more globally, particularly as edema and cystic components are infrequent in many subjects. To further evaluate the agreement between the predicted and expert tumor segmentations, a few semantic radiomic features were calculated, some of which are included among VASARI (Visually AcceSAble Rembrandt Images) 20 features, including proportion of the whole tumor volume that is enhancing tumor, non-enhancing tumor, cystic components, or edema. Pearson’s correlation coefficient (significance level P < .05) was calculated between the VASARI features obtained via predicted or expert segmentations.…”

Automated Tumor Segmentation and Brain Tissue Extraction from Multiparametric MRI of Pediatric Brain Tumors: A Multi-Institutional Study

“…These results are significantly higher than other pediatric deep learning models [11][12][13] , which achieved Dice scores ranging from 0.71 to 0.76. Several of these existing models are focused on a particular histology, such as optic pathway gliomas (OPGs) 17,20 . The variety of histologies and tumor locations included in our model allowed us to build a model that can be used more generally.…”

“…NEC label was generated to validate the performance of the model in distinguishing non-enhancing from enhancing areas more globally, particularly as edema and cystic components are infrequent in many subjects. To further evaluate the agreement between the predicted and expert tumor segmentations, a few semantic radiomic features were calculated, some of which are included among VASARI (Visually AcceSAble Rembrandt Images) 20 features, including proportion of the whole tumor volume that is enhancing tumor, non-enhancing tumor, cystic components, or edema. Pearson's correlation coefficient (significance level p < 0.05) was calculated between the VASARI features obtained via predicted or expert segmentations.…”

“…Such models do not generalize well to PBTs 15 due to the different radiological appearance of tumors compared to adult brain tumors, and the anatomical differences as a result of the developing brain in children 16 . A few studies have proposed different DL solutions to the PBT segmentation problem, with whole tumor Dice scores ranging between 0.68-0.76 [17][18][19][20] . These approaches show lower performance than the models proposed for adult brain tumor segmentation, are often only designed for a particular histology, only segment the whole tumor without subregions, or segment the tumors based on one or two MRI sequences.…”