Contributions
Abstract: P518
Type: Poster
Abstract Category: Pathology and pathogenesis of MS - Imaging
Background: There is an increasing number of MRI analysis methods in MS to calculate brain volume, cortical thickness, lesion characteristics and other imaging parameters. Additionally, increasingly large data sets are acquired in clinical trials and research projects. It is difficult and cumbersome to manually (or semi-automatically) process such large datasets with different, although linked, post-processing methods. Variations between software versions and pipeline parameters can bias the results, and controlling for inter-operator variability requires time and effort. We developed an integral system that offers scalable, automated, and distributed processing of MS imaging data and tested it using a large research data set.
Methods: We customized an open-source imaging informatics software platform (XNAT) that facilitates data management for neuroimaging into an integrated system that automatically carries out the required processing over a scalable cluster of computers. We tested the efficacy of the system in 2 different MS research data sets with a total number of 202 MRI sessions. The test system architecture was modest and comprised 3 processing nodes with a total number of 10 processing cores.
Results: The following processes were completed by the presented system: 130 brain parcellations, 54 lesion fillings, 65 segmentations of hippocampi, 65 cortical thickness calculations, 117 brain sodium maps, 56 Diffusion Tensor Imaging (DTI) maps, and 129 instances of Neurite Orientation Dispersion and Density Imaging (NODDI). The processing was completed in less than 8 weeks without operator intervention. In our calculations of brain total sodium concentration, we obtained significantly lower coefficients of variation (p< 0.05), and more significant differences between patients and healthy controls (p=0.002 vs. 0.033 for the GM) using the presented system as compared with a manual method.
Conclusions: Our newly developed fully automated system for the analysis of imaging data is a more time- and resource-efficient method than the manual/semi-automatic solutions currently in use. The introduction of this service will help to standardize results across studies (including audit-trail and storage), as the same processing pipelines, software versions and parameters will be used for all the processes. Other benefits of the proposed architecture include its cost effectiveness and potential to be extended to clinical trials, clinical setting and other research centres.
Disclosure:
Baris Kanber: nothing to disclose
Ferran Prados: nothing to disclose
Niamh Cawley: nothing to disclose
Arman Eshaghi: Arman Eshaghi has received MAGNIMS and Multiple Sclerosis International Federation McDonald fellowships (MSIF, www.msif.org).
Sara Collorone: nothing to disclose
Claudia A.M. Wheeler-Kingshott: Claudia A.M. Gandini Wheeler-Kingshott serves as a consultant for Biogen and receives research support from the UK MS Society, UCL/UCLH NIHR BRC, EPSRC, ISRT, Wings for Life, New Zealand Brain Research Centre, Novartis, and Biogen.
Frederik Barkhof: nothing to disclose
Olga Ciccarelli: Olga Ciccarelli is an Associate Editor of Neurology and serves as a consultant for GE Healthcare, Novartis, Roche, Biogen, Genzyme and Teva.
Sebastien Ourselin: nothing to disclose
Abstract: P518
Type: Poster
Abstract Category: Pathology and pathogenesis of MS - Imaging
Background: There is an increasing number of MRI analysis methods in MS to calculate brain volume, cortical thickness, lesion characteristics and other imaging parameters. Additionally, increasingly large data sets are acquired in clinical trials and research projects. It is difficult and cumbersome to manually (or semi-automatically) process such large datasets with different, although linked, post-processing methods. Variations between software versions and pipeline parameters can bias the results, and controlling for inter-operator variability requires time and effort. We developed an integral system that offers scalable, automated, and distributed processing of MS imaging data and tested it using a large research data set.
Methods: We customized an open-source imaging informatics software platform (XNAT) that facilitates data management for neuroimaging into an integrated system that automatically carries out the required processing over a scalable cluster of computers. We tested the efficacy of the system in 2 different MS research data sets with a total number of 202 MRI sessions. The test system architecture was modest and comprised 3 processing nodes with a total number of 10 processing cores.
Results: The following processes were completed by the presented system: 130 brain parcellations, 54 lesion fillings, 65 segmentations of hippocampi, 65 cortical thickness calculations, 117 brain sodium maps, 56 Diffusion Tensor Imaging (DTI) maps, and 129 instances of Neurite Orientation Dispersion and Density Imaging (NODDI). The processing was completed in less than 8 weeks without operator intervention. In our calculations of brain total sodium concentration, we obtained significantly lower coefficients of variation (p< 0.05), and more significant differences between patients and healthy controls (p=0.002 vs. 0.033 for the GM) using the presented system as compared with a manual method.
Conclusions: Our newly developed fully automated system for the analysis of imaging data is a more time- and resource-efficient method than the manual/semi-automatic solutions currently in use. The introduction of this service will help to standardize results across studies (including audit-trail and storage), as the same processing pipelines, software versions and parameters will be used for all the processes. Other benefits of the proposed architecture include its cost effectiveness and potential to be extended to clinical trials, clinical setting and other research centres.
Disclosure:
Baris Kanber: nothing to disclose
Ferran Prados: nothing to disclose
Niamh Cawley: nothing to disclose
Arman Eshaghi: Arman Eshaghi has received MAGNIMS and Multiple Sclerosis International Federation McDonald fellowships (MSIF, www.msif.org).
Sara Collorone: nothing to disclose
Claudia A.M. Wheeler-Kingshott: Claudia A.M. Gandini Wheeler-Kingshott serves as a consultant for Biogen and receives research support from the UK MS Society, UCL/UCLH NIHR BRC, EPSRC, ISRT, Wings for Life, New Zealand Brain Research Centre, Novartis, and Biogen.
Frederik Barkhof: nothing to disclose
Olga Ciccarelli: Olga Ciccarelli is an Associate Editor of Neurology and serves as a consultant for GE Healthcare, Novartis, Roche, Biogen, Genzyme and Teva.
Sebastien Ourselin: nothing to disclose