Microstructural Imaging Data Centre (MIDaC)
Lead Research Organisation:
CARDIFF UNIVERSITY
Department Name: School of Physics and Astronomy
Abstract
Characterising microstructure in vivo is key to understanding tissue properties and processes in health and disease. Central to this mission, the UK's National Facility for In Vivo MR Imaging of Human Tissue Microstructure (NMIF) houses one of the world's three 'Connectom' MRI scanners, fitted with unique hardware and optimised for imaging tissue microstructure in unprecedented detail. The potential to enable new breakthroughs in neuroscience/medicine with such equipment is huge.
While extremely valuable, however, such resources are also expensive and scarce. Making their data open and widely available to the community in a well-documented, scientifically usable form would maximise their impact. With few exceptions, such practice is rarely adopted by medical imaging researchers. Our multi-disciplinary team, drawing on data sharing principles developed by the Cardiff Astronomy Instrumentation Group (AIG), will define and pioneer a unique scientific and technological platform for sharing NMIF data with the worldwide brain research community, extending the value and scientific applications of the resource. Ancillary information and novel processing workflows will also be published. The Microstructural Imaging Data Centre (MIDaC) will 'blaze a trail' that promotes open data in brain imaging research, and offer a unique resource of state-of-the-art data from one of the most powerful medical imaging instruments currently available.
While extremely valuable, however, such resources are also expensive and scarce. Making their data open and widely available to the community in a well-documented, scientifically usable form would maximise their impact. With few exceptions, such practice is rarely adopted by medical imaging researchers. Our multi-disciplinary team, drawing on data sharing principles developed by the Cardiff Astronomy Instrumentation Group (AIG), will define and pioneer a unique scientific and technological platform for sharing NMIF data with the worldwide brain research community, extending the value and scientific applications of the resource. Ancillary information and novel processing workflows will also be published. The Microstructural Imaging Data Centre (MIDaC) will 'blaze a trail' that promotes open data in brain imaging research, and offer a unique resource of state-of-the-art data from one of the most powerful medical imaging instruments currently available.
Planned Impact
Differential and early diagnosis of neurodegenerative diseases and syndromes, such as epilepsy, cancer, and dementia, are amongst society's biggest medical challenges. Given the variety of causes and aetiologies of these conditions, reliable identification of the underlying pathologies is a requisite for therapy effectiveness. Connectom scanners are perfectly suited to reveal early changes in structural and functional properties by exploiting unique hardware designed to characterise tissue with unprecedented level of detail. It is in the dissemination of this data that the technologies enabled by this project can have a serious impact, facilitating its full exploitation and, consequently, maximising the knowledge extracted from them. In particular, the following sectors might benefit from this proposal:
1. Healthcare sector, by exploiting novel image-based disease biomarkers to increase treatment effectiveness and reduce the associated costs.
2. Industry sector, by utilising the microstructural information to improve models and software for neurosurgery and drug and stem cell delivery into the brain, accounting for the impact of tissue microstructure on the transport of the drug/cells.
3. Education sector, by taking advantage of an imaging repository to be used in undergraduate and postgraduate courses and projects, as well as by scientific communities hosting educational events such as workshops, hackathons, and challenges.
1. Healthcare sector, by exploiting novel image-based disease biomarkers to increase treatment effectiveness and reduce the associated costs.
2. Industry sector, by utilising the microstructural information to improve models and software for neurosurgery and drug and stem cell delivery into the brain, accounting for the impact of tissue microstructure on the transport of the drug/cells.
3. Education sector, by taking advantage of an imaging repository to be used in undergraduate and postgraduate courses and projects, as well as by scientific communities hosting educational events such as workshops, hackathons, and challenges.
Description | The MIDaC project allowed the interdisciplinary team to establish the basis of a translational theme within the research imaging centre (CUBRIC) focused on open data and software practices. Despite its short duration (5 months), it enabled the project members to understand the disparity in open practices between the fields of Astronomy and Medical Imaging, and to generate a longer-term plan for the centre. Very importantly, the project members i. Got to understand the issues related to medical imaging data sharing, including data ownership and ethical concerns, ii. Established a common understanding of the problem from multiple perspectives, iii. Generated tools to facilitate the adoption of existing open standards (the BIDS toolbox), iv. Tested the feasibility of implementing and sharing open software (via BIDS Apps), and v. Noted the limitations of actual data protocols for specialised equipment and modalities, such as the scanner with ultra-strong gradients available in CUBRIC. All these points were carried forward by a subsequent STFC funded project (the Microstructural Imaging Protocol - MISP), where the team has generated a new data sharing extension to deal with special microstructural MRI data. The results from the project are expected to impact not only in the scientific domain, where open practices are becoming a central concern, but also commercially due to the possibility to establish online and remote data scanning and sharing services having data sharing at its main core. These enterprises are being led by Leandro Beltrachini and involving most of the MIDaC applicants, as well as Siemens as a key stakeholder. |
Exploitation Route | These findings are potentially important to the wider medical imaging community. There is considerable scope for making research data from expensive facilities available to the general research community, thereby enhancing their scientific productivity. |
Sectors | Healthcare Pharmaceuticals and Medical Biotechnology |
Description | The Medical Imaging Data Centre (MIDaC): market research study, Wellcome Trust ITPA |
Amount | £5,000 (GBP) |
Organisation | Wellcome Trust |
Sector | Charity/Non Profit |
Country | United Kingdom |
Start | 09/2022 |
End | 02/2023 |
Description | The Medical Imaging Data Centre (MIDaC): organisational, operational, business, and data management plans, STFC Impact Acceleration award |
Amount | £30,472 (GBP) |
Organisation | Science and Technologies Facilities Council (STFC) |
Sector | Public |
Country | United Kingdom |
Start | 09/2022 |
End | 03/2023 |
Description | The Medical Imaging Data Centre - Market research study |
Amount | £5,000 (GBP) |
Organisation | Engineering and Physical Sciences Research Council (EPSRC) |
Sector | Public |
Country | United Kingdom |
Start | 03/2021 |
End | 06/2021 |
Description | The Microstructural Imaging Sharing Protocol (MISP) - Cardiff University STFC Impact Acceleration Account |
Amount | £45,000 (GBP) |
Organisation | Science and Technologies Facilities Council (STFC) |
Sector | Public |
Country | United Kingdom |
Start | 08/2019 |
End | 03/2020 |
Description | Translating open software practices to computational electrophysiology - Cardiff University STFC Impact Acceleration Account |
Amount | £11,539 (GBP) |
Organisation | Science and Technologies Facilities Council (STFC) |
Sector | Public |
Country | United Kingdom |
Start | 01/2020 |
End | 03/2020 |
Title | Design of MRI pre-processing pipeline |
Description | The main task of the science team was to design a sound diffusion MRI pre-processing pipeline that allows to correct for the most important artefacts impacting on the data. The uniqueness of the system utilised for this project (the Connectom scanner with ultra-high gradients) imposed more challenges than regular MRI scanners for getting good quality images from which to extract microstructural information. For this reason, we have arrived at a robust pipeline consisting on the following five steps: a. Compute brain mask using the BET2 algorithm as implemented in FSL. This allows to save processing time by analysing data of interest only. b. Rician noise removal using the MPPCA method as implemented in MRtrix. c. Correction of signal drift due to system heating, utilising Vos' technique re-implemented in Python (the original algorithm is written in Matlab). d. Reduction of Gibbs ringing artefact based on Kellner's algorithm and as implemented in MRtrix. This artefact appears mostly in high contrast areas and is generated by the limited basis functions used in the discrete Fourier transform needed for MR image reconstruction. These steps were carefully selected after a thorough literature review. However, there are more steps we intend to do in future versions: i. To evaluate other algorithms for similar pre-processing steps. The modularity of the open source implementation of the pipeline (see Research Databases and Models section) will allow an easy change of algorithms correcting for a specific artefact, making it possible to make a systematic comparison between available techniques. ii. To study the impact of the algorithms' order on the resulting pre-processed images and the corresponding microstructural estimates. Although we considered the potential issues into account, a thorough analysis is definitely needed. Once again, the modularity of the pipeline design makes this analysis possible with minimal efforts. iii. To analyse the impact of different pre-processing steps in particular diffusion MRI sequences. It is worth noting that there is still no agreement in the medical imaging community regarding the list of steps and/or algorithms to use for pre-processing diffusion MRI data. We intend to perform these experiments in the near future and submit the results to a top journal in the field. |
Type Of Material | Improvements to research infrastructure |
Year Produced | 2019 |
Provided To Others? | Yes |
Impact | The pre-processing pipeline is made available to the entire centre, who are now able to use it in the internal computer cluster. We are currently working in a manuscript describing its advantages, which we expect to publish soon. |
URL | http://www.astro.cardiff.ac.uk/~spxap/MIDAC_2019/MIDAC_PIPELINE/ |
Title | Multidisciplinary team structure |
Description | One of the main challenges of this project was to establish a working agenda for researchers working in very different fields (as neuroimaging and astronomy) and with different areas of expertise within them. The dependence of the project on the working structure was even more pronounced given the short span of the grant (i.e. 5 months). To solve this task, we presented an organisational structure derived from that developed for the Cardiff-led Herschel-SPIRE space instrument. This structure comprised separate teams with well-defined tasks, overseen by a management team. Each team was led by a specialist in each of the technical sections needed for the success of the project: a. Operations team, in charge of the MRI acquisition setup, led by the lead MR operator in CUBRIC (Dr John Evans). b. Science team, responsible for the design of MRI pre-processing pipelines, led by a specialist in the field (Dr Leandro Beltrachini) c. Software development team, responsible for defining the data processing framework and strategies, led by an expert in the area of astronomy software (Dr Andreas Papageorgiou) d. Processing and archiving team, in charge of designing a data sharing interface, led by CUBRIC's IT manager (Dr Cyril Charron) Each of these teams comprised people working in very specific tasks as described in the case for support. Team leaders were responsible for managing the available resources (e.g. investigator time) and setting their own agenda and timelines, usually involving weekly meetings and regular discussions with other teams as required. The overall team structure was coordinated by a management team comprised of all the investigators and researchers of the project. The management team met on a monthly basis with the objectives of: a) summarising achievements during the period, b) overseeing the progress made by each team and provide feedback, c) aligning the work made by each team in the context of the grant, and d) setting the tasks for the coming period. Before the management meetings, LB was in charge of collating all the updates from each team in a single presentation, as well as sharing the corresponding documents to be discussed. The coordination of activities was managed using Microsoft Teams through Cardiff University subscription. The software was successfully employed to a) communicate internally between members, b) share presentations and material for the management meetings, c) share data and preliminary results between team members, and d) generate individual channels for communication within teams. |
Type Of Material | Improvements to research infrastructure |
Year Produced | 2018 |
Provided To Others? | No |
Impact | The resulting structure was crucial for obtaining tangible outcomes after only five months, allowing each team member to keep track of all individual tasks being done at the same time of understanding their fit within the overall project. During the management meetings, all members had time to make suggestions and provide feedback, which initiated very interested discussion points, ranging from technical details to further work beyond the scope of this project. In particular, our successful grant application to the STFC Impact Acceleration Account is a result of such enriching meetings. |
Title | Advanced diffusion MRI database |
Description | One of the original goals of the MIDaC project was to generate an exemplar dataset for testing the entire data centre proof-of-concept. Moreover, we planned the acquisition of unique diffusion MRI data in the National Microstructural Imaging Facility (NMIF) available in CUBRIC, comprising one of the three Connectom scanners with ultra-high gradients in the world. This scanner has been designed to apply gradient fields of up to 300mT/m, i.e. between 7 and 8 times larger than any other MRI scanner. This allows to characterise smaller structures than any other MRI facility, such as the size of neurons and other cells in brain tissue. We used the Connectom scanner to acquire data from a single subject during a number of sessions and employing both standard and special sequences uniquely available with the NMIF. The objective was to present researchers with a resource that enables direct comparison between regularly acquired data and that obtained by employing optimised sequences not yet available through clinical scanners. More specifically, we measured the following images: a. Session number 1: Single pulsed gradient spin echo (sPGSE) with standard parameters. This data is an example of the most utilised setup in the diffusion MRI community. Its purpose is to provide data from the same subject with which to compare. b. Session number 2: sPGSE with varying diffusion times. This data is not novel but goes beyond the standard sequence utilised in session 1 since it has different values for the diffusion times, not generally done in clinical systems. c. Session number 3: sPGSE with multiple gradient amplitudes and diffusion shells. This dataset is a state-of-the-art since it was acquired with ultra-high gradient values. d. Session number 4: Oscillating gradient spin echo (OGSE) with different frequencies. This sequence is a state-of-the-art which was shown to increase the sensitivity to microstructures and for which there is no publicly available data. All the details of the images acquired are available together with ancillary data in the URL. We are now describing the usefulness of the dataset in a report we expect to submit to a peer-reviewed journal in the coming months. |
Type Of Material | Database/Collection of data |
Year Produced | 2019 |
Provided To Others? | Yes |
Impact | Given its uniqueness, the dataset is expected to reach high levels of adoption in the field. We are now preparing a manuscript describing the data and showing exemplars of its use. |
URL | http://www.astro.cardiff.ac.uk/~spxap/MIDAC_2019/MIDAC_Public_Archive/ |
Title | Modular pre-processing workflow |
Description | From the outset, the software included a crucial and novel aspect to ensure its usability and flexibility for further upgrades: the modularity. This refers to the degree to which a system's components may be separated and recombined. This is a very important property since allows to define independent processing steps performing actions on the data without any other requirement than utilising some predefined inputs and generating predefined outputs. This allows to easily test a) different arrangements of processing steps; b) multiple implementations of processing algorithms (e.g. in different programming languages); and c) unrelated algorithms for solving the same processing task. These aspects have been shown crucial for allowing the community to experiment different pre-processing setups and promote both reproducibility and accessibility to the pipeline (e.g. see the fMRIprep report at https://www.nature.com/articles/s41592-018-0235-4). We have defined a generic modular structure in Python that allows researchers to normalise the usually ad hoc process of defining the pre-processing pipeline. |
Type Of Material | Data analysis technique |
Year Produced | 2019 |
Provided To Others? | Yes |
Impact | The modularity concept was already introduced in several pipelines in the centre. It is worth noting that the modular structure created in MIDaC may have a great impact in other areas in which multiple sequential data processing steps are required. Examples include other medical imaging modalities (such as electro/magnetoencephalography and multimodal MRI), astronomy data, and geophysics. |
URL | http://www.astro.cardiff.ac.uk/~spxap/MIDAC_2019/MIDAC_PIPELINE/ |
Description | Collaboration with researchers working at Valladolid, Spain |
Organisation | University of Valladolid |
Country | Spain |
Sector | Academic/University |
PI Contribution | We have set up a joint working plan on the basis of the MIDaC project. More explicitly, we have designed a programme for presenting new diffusion MRI data sharing protocols. The role of the Cardiff team is to present the basis of such standard. |
Collaborator Contribution | Our partner will be responsible of exploring collaboration opportunities for the protocol with Philips Iberia, as well as contributing in the design of the protocol itself. |
Impact | We have received funding from a STFC Impact acceleration account project through Cardiff University. Title: The microstructure Imaging Sharing protocol (MISP) Total award value: £96k Investigators: Dr. Leandro Beltrachini (PI, Cardiff), Dr. Andreas Papageorgiou (Cardiff), Prof. Matt Griffin (Cardiff), Prof. Derek Jones (Cardiff), Prof. Peter Hargrave (Cardiff), Prof. Santiago Aja Fernandez (Valladolid, Spain). |
Start Year | 2019 |
Title | MIDaC Public Interface |
Description | The public interface is intended to allow users with no access to powerful computers, or without the know-how to install and operate specialised software, to perform science on intermediate-processed science data. These intermediate-processed data have most of the common systematic effects removed and are ready for higher-level analysis (see Research tools, point 2). For the public interface we generated a dynamic front-end, implemented in Javascript and PHP, with the aim of providing an intuitive way for visitors to enquire for the available datasets available for download, along with a description of the scan and the parameters that correspond to each dataset. For each resource, the user is offered a choice of several levels of processing to download, as well as a summary of processing diagnostics. Advanced users are also permitted to download the raw (BIDS) data in order to perform their own specialised data reduction. On the back-end, the server is implementing user-access control and collection of user statistics (visitor statistics) and science product statistics (e.g. download frequency, location etc.) |
Type Of Technology | Webtool/Application |
Year Produced | 2019 |
Impact | We were not able to collate any impact information regarding this entry yet. |
URL | http://www.astro.cardiff.ac.uk/~spxap/MIDAC_2019/MIDAC_Public_Archive/ |
Title | Open Source Data Processing Pipeline |
Description | It is common in the field of neuroimaging that processing pipelines are implemented and depended on propitiatory software (Matlab, IDL). One of the outcomes of this project is to create a data processing pipeline implemented in open source, freely available software as Python, and depended only on open-source, freely available libraries (MRtrix3, FSL). The MIDaC pipeline is used to process raw dMRI data in BIDS format into intermediate processing levels with common systematic effects removed. In addition to the processing, the pipeline generates diagnostic products in a format compatible with the MIDaC Public Interface so that the output can be readily made publicly available. |
Type Of Technology | Software |
Year Produced | 2019 |
Impact | The pipeline has been recently introduced in the centre. We expect to collate user feedback for new versions. |
URL | http://www.astro.cardiff.ac.uk/~spxap/MIDAC_2019/MIDAC_PIPELINE/ |
Title | The BIDS Toolbox |
Description | Neuroimaging data is very heterogeneous. In its most general form, it may comprise information in plenty different formats, containing from single scalar quantities to strings and multidimensional data arrays. The wide variety of existing protocols, nomenclatures, and instruments make data sharing a demanding challenge in the field. Addressing this problem is crucial for facilitating collaborations between colleagues and centres, as well as to enhancing reproducibility and transparency of results. Moreover, it becomes a crucial organisational aspect for arranging large databases based on numerous subjects, each of them scanned with multiple imaging instruments providing complementary information of brain structure and function. To tackle this issue, Gorgolewsky et al. proposed the Brain Imaging Data Structure (BIDS) format. BIDS is a community-led standard for organising and describing neuroimaging data and behavioural information, maximising their usability and, consequently, open data practices. In few years, it has found an increasingly important role in neuroimaging communities, including fMRI, MEG, and EEG. However, despite of the efforts of the community to define the standard, it has not been widely embraced by the MRI community in general. The reason, we think, is mostly based on the lack of a comprehensive and simple-to-use tool for managing and converting MRI raw data to BIDS format. Existing tools in the field lack of some key functionalities required by scientists, such as the possibility to modify an existing BIDS structure (e.g. by adding new data) or to automatically categorise the medical images without additional information other than the raw data. To solve this problem, we propose the BIDS Toolbox, an open source software tool that simplifies the adoption of BIDS for researchers and institutions working the field of neuroimaging. The BIDS Toolbox aims at being a software piece that is easy to integrate in existing data centres and research environments willing to adopt BIDS as a format to share neuroimaging data. To that end, we chose common design practices in software engineering and adopted a microservice design, which enables modularity and facilitates integration with other services, like an image management platform. The Toolbox functionality is exposed through a REST API and uses JSON as communication format. In the current implementation of the Toolbox, we have used and modified parts of the open source software bidskit for some of the dataset-creation features of the toolbox, and the Flask framework to create the web services. All the codebase is Python v3. In the design of the BIDS Toolbox we assumed that the user might not know the scan modality and type for a given set of DICOM files, or that the Toolbox could be part of a processing pipeline which could not have that information. Given that this information is required to create a BIDS dataset, we developed an algorithm that infers the type of scan based on the properties of the DICOM files. We refer the interested reader to see more details in the conference paper available at https://arxiv.org/abs/1906.09996. |
Type Of Technology | Software |
Year Produced | 2019 |
Open Source License? | Yes |
Impact | Preliminary work was presented at the 2019 edition of the IEEE SmartWorld congress, in Leicester. The presentation was entitled "The BIDS Toolbox: A Web Service to Manage Brain Imaging Datasets" and delivered as part of the Workshop on Data Pre-processing for Big Biomedical Data. It had a very good reception and was suggested for expansion and submission at a Special Issue on "Internet of People: Human-driven Artificial Intelligence and Internet for Smarter Hyper-Connected Societies" (https://www.journals.elsevier.com/future-generation-computer-systems/call-for-papers/special-issue-on-internet-of-people-human-driven-artificial). The toolbox is nowadays utilised within the centre (composed of 200 researchers in the field) and is expected to facilitate the adoption of open science practices in the field. |
URL | http://arxiv.org/abs/1906.09996 |
Description | Presentations of the project outcomes at different stages at CUBRIC |
Form Of Engagement Activity | A formal working group, expert panel or dialogue |
Part Of Official Scheme? | No |
Geographic Reach | Local |
Primary Audience | Other audiences |
Results and Impact | Team members have presented, on three opportunities, the project aim and status at different stages. These talks were open to people working in CUBRIC and related centres within Cardiff University, and were attended by 30 people approximately. Although CUBRIC is a data generation centre, there was no culture on data sharing policies, and these talks highlighted the benefits of doing so. Since the last talk, several people have contacted the MIDaC team to provide advice on data sharing formats and procedures. As a result, CUBRIC has developed a data infrastructure policy based on the results of this project (more explicitly, the adoption of the BIDS format for data storage and sharing). |
Year(s) Of Engagement Activity | 2018,2019 |