Search

Introduction: Inherent variations between inter-center data can undermine the robustness of segmentation models when applied at a specific center (dataset shift). We investigated whether specialized center-specific models are more effective compared to generalist models based on multi-center data, and how center-specific data could enhance the performance of generalist models within a particular center using a fine-tuning transfer learning approach. For this purpose, we studied the dataset shift at center level and conducted a comparative analysis to assess the impact of data source on glioblastoma segmentation models., Methods & Materials: The three key components of dataset shift were studied: prior probability shift-variations in tumor size or tissue distribution among centers; covariate shift-inter-center MRI alterations; and concept shift-different criteria for tumor segmentation. BraTS 2021 dataset was used, which includes 1251 cases from 23 centers. Thereafter, 155 deep-learning models were developed and compared, including 1) generalist models trained with multi-center data, 2) specialized models using only center-specific data, and 3) fine-tuned generalist models using center-specific data., Results: The three key components of dataset shift were characterized. The amount of covariate shift was substantial, indicating large variations in MR imaging between different centers. Glioblastoma segmentation models tend to perform best when using data from the application center. Generalist models, trained with over 700 samples, achieved a median Dice score of 88.98%. Specialized models surpassed this with 200 cases, while fine-tuned models outperformed with 50 cases., Conclusions: The influence of dataset shift on model performance is evident. Fine-tuned and specialized models, utilizing data from the evaluated center, outperform generalist models, which rely on data from other centers. These approaches could encourage medical centers to develop customized models for their local use, enhancing the accuracy and reliability of glioblastoma segmentation in a context where dataset shift is inevitable., Competing Interests: Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2024 The Authors. Published by Elsevier B.V. All rights reserved.)

Background: Neuroimaging analysis is currently crucial for an early assessment of glioblastoma, to help improving treatment and tumor follow-up. To this end, multiple functional and morphological MRI sequences are usually employed, requiring the development of automated tools capable to extract the relevant information from these sources. In this work we present ONCOhabitats (https://www.oncohabitats.upv.es): an online open access system for glioblastoma analysis based on MRI data., Methods: ONCOhabitats provides two main services for untreated glioblastomas: (1) malignant tissue segmentation, and (2) vascular heterogeneity assessment of the tumor. The segmentation service implements a deep patch-wise 3D Convolutional Neural Network with residual connections. The vascular heterogeneity assessment service implements the Hemodynamic Tissue Signature (HTS) method patented in P201431289, which aims to identify habitats within the tumor with early prognostic capabilities., Results: The segmentation service was validated against the BRATS 2017 reference dataset, showing comparable results with current state-of-the-art methods (whole tumor Dice segmentation: 0.89). The vascular heterogeneity assessment service was validated in a retrospective cohort of 50 patients, in a study focused on predicting patient overall survival based on the HTS habitats. Cox proportional hazard regression analysis and Kaplan-Meier survival study showed significant positive correlations (p-value <.05) between the HTS habitats and patient overall survival. ONCOhabitats system also generates radiological reports for each service, including volumetries and perfusion measurements of the different regions of the lesion., Conclusion: ONCOhabitats system provides open-access services for glioblastoma heterogeneity assessment, implementing consolidated state-of-the-art techniques for medical image analysis. Additionally, we also give access to the scientific community to our computational resources, offering a computational capacity of about 300 cases per day., (Copyright © 2019 Elsevier B.V. All rights reserved.)