Tag Archives: data sharing

Data sharing and reuse case study: the Mammographic Image Society database

Uncategorized, , , , ,

The Mammographic Image Society (MIAS) database is a set of mammograms put together in 1992 by a consortium of UK academic institutions and archived on 8mm DAT tape, copies of which were made openly available and posted to applicants for a small administration fee. The mammograms themselves were curated from the UK National Breast Screening Programme, a major screening program that was established in the late 80s offering routine screening every three years to women aged between 50-64.

The motivations for creating the database were to make a practical contribution to computer vision research – which sought to improve the ability of computers to interpret images – and to encourage the creation of more extensive datasets. In the peer-reviewed paper bundled with the dataset, the researchers note that “a common database is a positive step towards achieving consistency in performance comparison and testing of algorithms”.

Due to increased demand, the MIAS database was made available online via third parties, albeit in a lower resolution than the original. Despite no longer working in this area of research, the lead author, John Suckling – now Director of Research in the Department of Psychiatry, part of Cambridge Neuroscience –  started receiving emails asking for access to the images at the original resolution. This led him to dig out the original 8mm DAT tapes with the intention of making the images available openly in a higher resolution. The tapes were sent to the University Information Service (UIS), who were able to access the original 8mm tape and download higher resolution versions of the images. The images were subsequently deposited in Apollo and made available under a CC BY license, meaning researchers are permitted to reuse them for further research as long as appropriate credit is given. This is the most commonly used license for open datasets and is recommended by the majority of research funding agencies.

Motivations for sharing the MIAS database openly

The MIAS database was created with open access in mind from the outset. When asked whether he had any reservations about sharing the database openly, the lead author John Suckling noted:

“There are two broad categories of data sharing; data acquired for an original purpose that is later shared for secondary use; data acquired primarily for sharing. This dataset is an example of the latter. Sharing data for secondary use is potentially more problematic especially in consortia where there are a number of continuing interests in using the data locally. However, most datasets are (or should be) superseded, and then value can only be extracted if they are combined to create something greater than the sum of the parts. Here, careful drafting of acknowledgement text can be helpful in ensuring proper credit is given to all contributors.”

This distinction – between data acquired for an original purpose that is later shared for secondary use and data acquired primarily for sharing – is one that is important and often overlooked. The true value of some data can only be fully realised if openly shared. In such cases, as Suckling notes, sufficient documentation can help ensure the original researchers are given credit where it is due, as well as ensuring it can be reused effectively. This is also made possible by depositing the data on an institutional repository such as Apollo, where it will be given a DOI and its reuse will be easier to track.

Impact of the MIAS database

As of August 2020, the MIAS database has received over 5500 downloads across 27 different countries, including some developing countries where breast cancer survival rates are lower. Google Scholar currently reports over 1500 citations for the accompanying article as well as 23 citations for the dataset itself. A review of a sample of the 1500 citations revealed that many were examples of the data being reused rather than simply citations of the article. Additionally, a systematic review published in 2018 cited the MIAS database as one of the most widely used for applying breast cancer classification methods in computer aided diagnosis using machine learning, and a benchmarking review of databases used in mammogram research identified it as the most easily accessible mammographic image database. The reasons cited for this included the quality of the images, the wide coverage of types of abnormalities, and the supporting data which provides the specific locations of the abnormalities in each image.

The high impact of the MIAS database is something Suckling credits to the open, unrestricted access to the database, which has been the case since it was first created. When asked whether he has benefited from this personally, Suckling stated “Direct benefits have only been the citations of the primary article (on which I am first author). However, considerable efforts were made by a large number of early-career researchers using complex technologies and digital infrastructure that was in its infancy, and it is extremely gratifying to know that this work has had such an impact for such a large number of scientists.”. Given that the database continues to be widely cited and has been downloaded from Apollo 1358 times since January 2020, it is still clearly the case that the MIAS database is having a wide impact.

The MIAS Database Reused

As mentioned above, the MIAS database has been widely reused by researchers working in the field of medical image analysis. While originally intended for use in computer vision research, one of the main ways in which the dataset has been used is in the area of computer aided diagnosis (CAD), for which researchers have used the mammographic images to experiment with and train deep learning algorithms. CAD aims to augment manual inspection of medical images by medical professionals in order to increase the probability of making an accurate diagnosis.

A 2019 review of recent developments in medical image analysis identified lack of good quality data as one of the main barriers researchers in this area face. Not only is good quality data a necessity but it must also be well documented as this review also identified inappropriately annotated datasets as a core challenge in CAD. The MIAS database is accompanied by a peer-reviewed paper explaining its creation and content as well as a read me PDF which explains the file naming convention used for the images as well as the annotations used to indicate the presence of any abnormalities and classify them based on their severity. The presence of this extensive documentation combined with it having been openly available from the outset could explain why the database continues to be so widely used.

Reuse example: Applying Deep Learning for the Detection of Abnormalities in Mammograms

This research, published in 2019 in Information Science and Applications, looked at improving some of the current methods used in CAD and attempted to address some inherent shortcomings and increase the competency level of deep learning models when it comes the minimisation of false positives when applying CAD to mammographic imaging. The researchers used the MIAS database alongside another larger dataset in order to evaluate the performance of two existing convolutional neural networks (CNN), which are deep learning models used specifically for classifying images. Using these datasets, they were able to demonstrate that versions of two prominent CNNs were able to detect and classify the severity of abnormalities on the mammographic images with a high degree of accuracy.

While the researchers were able to make good use of the MIAS database to carry out their experiments, due to the inclusion of appropriate documentation and labelling, they do note that since it is a relatively small dataset it is not possible to rule out “overfitting”, where a deep learning model is highly accurate on the data used to train the model, but may not generalise well to other datasets. This highlights the importance of making such data openly available as it is only possible to improve the accuracy of CAD if sufficient data is available for researchers to carry out further experiments and improve the accuracy of their models. ­

Reuse example: Computer aided diagnosis system for automatic two stages classification of breast mass in digital mammogram images

This research, published in 2019 in Biomedical Engineering: Applications, Basis and Communications, used the MIAS database along with the Breast Cancer Digital Repository to test a CAD system based on a probabilistic neural network – a machine learning model that predicts the probability distribution of a given outcome –  developed to automate classification of breast masses on mammographic images. Unlike previously developed models, their model was able to segment and then carry out a two-stage classification of breast masses. This meant that rather than classifying masses into either benign or malignant, they were able to develop a system which carried out a more fine-grained classification consisting of seven different categories. Combining the two different databases allowed for an increased confidence level in the results gained from their model, again raising the importance of the open sharing of mammographic image datasets. After testing their model on images from these databases, they were able to demonstrate a significantly higher level of accuracy at detecting abnormalities than had been demonstrated by two similar models used for evaluation. On images from the MIAS Database and Breast Cancer Digital Repository their model was able to detect abnormalities with an accuracy of 99.8% and 97.08%, respectively. This was also accompanied by increased sensitivity (ability to correctly classify true positives) and specificity (ability to correctly classify false negatives).

Conclusion

Many areas of research can only move forward if sufficient data is available and if it is shared openly. This, as we have seen, is particularly true in medical imaging where despite datasets such as the MIAS database being openly available, there is a data deficiency which needs to be addressed in order to improve the accuracy of the models used in computer-aided diagnosis. The MIAS database is a clear example of a dataset that has enabled an important area of research to move forward by enabling researchers to carry out experiments and improve the accuracy of deep learning models developed for computer-aided diagnosis in medical imaging. The sharing and reuse of the MIAS database provides an excellent model for how and why future researchers should make their data openly available.

Published 20th August 2020
Written by Dominic Dixon

CCBY icon

Charities’ perspective on research data management and sharing

Uncategorized, , , , , , ,

In 2015 the Cambridge Research Data Team organised several discussions between funders and researchers. In May 2015 we hosted Ben Ryan from EPSRC, which was followed by a discussion with Michael Ball from BBSRC in August. Now we have invited our two main charity funders to discuss their views on data management and sharing with Cambridge researchers.

David Carr from the Wellcome Trust and Jamie Enoch from Cancer Research UK (CRUK) met with our academics on Friday 22 January at the Gurdon Institute. The Gurdon Institute was founded jointly by the Wellcome Trust and CRUK to promote research in the areas of developmental biology and cancer biology, and to foster a collaborative environment for independent research groups with diverse but complementary interests.

This blog summarises the presentations and discusses the data sharing expectations from Wellcome Trust and CRUK. A second related blog ‘In conversation with Wellcome Trust and CRUK‘ summarises the question and answer session that was held with a group of researchers on the same day.

Wellcome Trust’s requirements for data management and sharing

Sharing research data is key for Wellcome’s goal of improving health

David Carr started his presentation explaining that the Wellcome Trust’s mission is to support research with the goal of improving health. Therefore, the Trust is committed to ensuring research outputs (including research data) can be accessed and used in ways that will maximise health and societal benefits. David reminded the audience of benefits of data sharing. Data which is shared has the potential to:

  • Enable validity and reproducibility of research findings to be assessed
  • Increase the visibility and use of research findings
  • Enable research outputs to be used to answer new questions
  • Reduce duplication and waste
  • Enable access to data to other key communities – public, policymakers, healthcare professionals etc.

Data sharing goes mainstream

David gave on overview of data sharing expectations from various angles. He started by referring to the Royal Society’s report from 2012: Science as an open enterprise, which sets sharing as the standard for doing science. He then also mentioned other initiatives like the G8 Science Ministers’ statement, the joint report from the Academy of Medical Sciences, BBSRC, MRC and Wellcome Trust on reproducibility and reliability of biomedical research and the UK Concordat on Open Research Data with a take-home message that sharing data and other research outputs is increasingly becoming a global expectation, and a core element of good research practice.

Wellcome Trust’s policy for open data

The next aspect of David’s presentation was Wellcome Trust’s policy on data management and sharing. The policy was first published almost a decade ago (2007) with subsequent modifications in 2010. The principle of the policy is simple: research data should be shared and preserved in a manner which maximises its value to advance research and improve health. Wellcome Trust also requires data management plans as a compulsory part of grant applications, where the proposed research is likely to generate a dataset that will have significant value to researchers and other users. This is to ensure that researchers understand the importance of data management and sharing and to plan for it from the start their projects.

Cost of data sharing

Planning for data management and sharing involves costing for these activities in the grant proposal. The Wellcome Trust’s FAQ guidance on data sharing policy says that: “The Trust considers that timely and appropriate data management and sharing should represent an integral component of the research process. Applicants may therefore include any costs associated with their proposed approach as part of their proposal.” David then outlined the types of costs that can be included in grant applications (including for dedicated staff, hardware and software, and data access costs). He noted that in the current draft guidance on costing for data management estimated costs for long-term preservation that extend beyond the lifetime of the grant are not eligible, although costs associated with the deposition of data in recognised data repositories can be requested.

Key priorities and emerging areas in data management and sharing

Infrastructure

The Wellcome Trust also identified key priorities and emerging areas where work needs to be done to better support of data management and sharing. The first one was to provide resources and platforms for data sharing and access. David pointed out that wherever available, discipline-specific data repositories are the best home for research data, as they provide rich metadata standards, community curation and better discoverability of datasets.

However, the sustainability of discipline-specific repositories is sometimes uncertain. Discipline-specific resources are often perceived as ‘free’. However, research data submitted to ‘free’ data repositories has to be stored somewhere and the amount of data produced and shared is growing exponentially – someone has to pay for the cost of storage and long-term curation in discipline-specific data repositories. An additional point for consideration is that many disciplines do not have their own repositories and therefore need to heavily rely on institutional support.

Access

Wellcome Trust funds a large number of projects in clinical areas. Dealing with patient data requires careful ethical considerations and planning from the very start of the project to ensure that data can be successfully shared at the end of the project. To support researchers in dealing with patient data The Expert Advisory Group on Data Access (a cross-funder advisory body established by MRC, ESRC, Cancer Research UK and the Wellcome Trust) has developed guidance documents and practice papers about handling of sensitive data: how to ask for informed consent, how to anonymise data and the procedures that need to be in place when granting access to data. David stressed that balance needs to be struck between maximising the use of data and the need to safeguard research participants.

Incentives for sharing

Finally, if sharing is to become the normal thing to do, researchers need incentives to do so. Wellcome Trust is keen to work with others to ensure that researchers who generate and share datasets of value receive appropriate recognition for their efforts. A recent report from the Expert Advisory Group on Data Access proposed several recommendations to incentivise data sharing, with specific roles for funders, research leaders, institutions and publishers. Additionally, in order to promote data re-use, the Wellcome Trust joined forces with the National Institutes of Health and the Howard Hughes Medical Institute and launched the Open Science Prize competition to encourage prototyping and development of services, tools or platforms that enable open content.

Cancer Research UK’s views on data sharing

The next talk was by Jamie Enoch from Cancer Research UK. Jamie started by saying that because Cancer Research UK (CRUK) is a charity funded by the public, it needs to ensure it makes the most of its funded research: sharing research data is elemental to this. Making the most of the data generated through CRUK grants could help accelerate progress towards the charity’s aim in its research strategy, to see three quarters of people surviving cancer by 2034. Jamie explained that his post – Research Funding Manager (Data) – has been created as a reflection of data sharing being increasingly important for CRUK.

The policy

Jamie started talking about the key principles of CRUK data sharing policy by presenting the main issues around research data sharing and explaining the CRUK’s position in relation to them:

  • What needs to be shared? All research data, including unpublished data, source code, databases etc, if it is feasible and safe to do so. CRUK is especially keen to ensure that data underpinning publications is made available for sharing.
  • Metadata: Researchers should adhere to community standards/minimum information guidelines where these exist.
  • Discoverability: Groups should be proactive in communicating the contents of their datasets and showcasing the data available for sharing

Jamie explained that CRUK really wants to increase the discoverability of data. For example, clinical trials units should ideally provide information on their websites about the data they generate and clear information about how it can be accessed.

  • Modes of sharing: Via community or generalist repositories, under the auspices of the PI or a combination of methods

Jamie explained that not all data can be/should be made openly available. Due to ethical considerations sometimes access to data will have to be restricted. Jamie explained that as long as restrictions are justified, it is entirely appropriate to use them. However, if access to data is restricted, the conditions on which access will be granted should be considered at the project outset, and these conditions will have to be clearly outlined in metadata descriptions to ensure fair governance of access.

  • Timeframes: Limited period of exclusive use permitted where justified

Jamie suggested adhering to community standards when thinking about any periods of exclusive use of generated research data. In some communities research data is made accessible at the time of publication. Other communities will expect data release at the time of generation (especially in collaborative genomics projects). Jamie further explained that particularly in cases where new data can affect policy development, it is key that research data is released as soon as possible.

  • Preservation: Data to be retained for at least 5 years after grant end
  • Acknowledgement: Secondary users of data should credit original researcher and CRUK
  • Costs: Appropriately justified costs can be included in grant proposals

As of late 2015, financial support for data management and sharing can be requested as a running cost in grant applications. Jamie explained that there are no particular guidelines in place explaining eligible and non-eligible costs and that the most important aspect is whether the costs are well justified or not, and reasonable in the context of the research envisaged.

Jamie stressed that the key point of the CRUK policy is to facilitate data sharing and to engage with the research community, recognising the challenges of data sharing for different projects and the need to work through these collaboratively, rather than enforce the policy in a top-down fashion.

Policy implementation

Subsequently, the presentation discussed ways in which CRUK policy is implemented. Jamie explained that the main tool for the policy implementation is the new requirement for data management plans as compulsory part of grant applications.

Two of the three main response mode committees: Science Committee and Clinical Research Committee have a two-step process of writing a data management plan. During the grant application stage researchers need to write a short, free-form description about how they plan to adhere to CRUK’s policy on data sharing. Only if the grant is accepted, the beneficiary will be asked to write a more detailed data management plan, in consultation with CRUK representatives.

This approach serves two purposes as it:

  • ensures that all applicants are aware of CRUK’s expectations on data sharing (they all need to write a short paragraph about data sharing)
  • saves researchers’ time: only those applicants who were successful will have to provide a detailed data management plan, and it allows the CRUK office to engage with successful applicants on data sharing challenges and opportunities

In contrast, applicants for the other main CRUK response mode committee, the Population Research Committee, all fill out a detailed data management and sharing plan at application stage because of the critical importance of sharing data from cohort and epidemiological studies.

Outlooks for the future

Similarly to the Wellcome Trust, CRUK realised that cultural change is needed for sharing to become the normality. CRUK have initiated many national and international partnerships to help the reward of data sharing.

One of them is a collaboration with the YODA (Yale Open Data Access) project aiming to develop metrics to monitor and evaluate data sharing. Other areas of collaborative work include collaboration with other funders on development of guidelines on ethics of data management and sharing, platforms for data preservation and discoverability, procedures for working with population and clinical data. Jamie stressed that the key thing for CRUK is to work closely with researchers and research managers – to understand the challenges and work through these collaboratively, and consider exciting new initiatives to move the data sharing field forwards.

Links

Published 5 February 2016
Written by Dr Marta Teperek, verified by David Carr and Jamie Enoch
Creative Commons License

Software Licensing and Open Access

Uncategorized, , , , , , , , , ,

As part of the Office of Scholarly Communication Open Access Week celebrations, we are uploading a blog a day written by members of the team. Wednesday is a piece by Dr Marta Teperek reporting on the Software Licensing Workshop held on 14 September 2015 at Cambridge.

Uncertainties about sharing and licensing of software

If the questions that the Research Data Service Team have been asked during data sharing information sessions with over 1000 researchers at the University of Cambridge are any indicator, then there is a great deal of confusion about sharing source code.

There have been a wide range of questions during the discussions in these sessions, and the Research Data Service Team has recorded these. We are systematically ensuring that the information we are providing to our research community is valid and accurate. To address the questions about source code we decided to call in expert help. Shoaib Sufi and Neil Chue Hong* from the Software Sustainability Institute agreed to lead a workshop on Software Licensing in September, at the Computer Lab in Cambridge. Shoaib’s slides are here, and Neil’s slides on Open Access policies and software sharing are here.

Malcolm Grimshaw and Chris Arnot from Cambridge Enterprise also came to the workshop to answer questions about Cambridge-specific guidance on software commercialisation.

We had over 50 researchers and several research data managers from other UK universities attending the Software Licensing workshop. The main questions we were trying to resolve was: Are researchers expected to share source code they used in their projects? And if so, under what conditions?

Is software considered as ‘research data’ and does it need to be shared?

The starting question in the discussion was whether software needed to be shared. Most public funders now require that research data underpinning publications is made available. What is the definition of research data? According to the EPSRC research data “is defined as recorded factual material commonly retained by and accepted in the scientific community as necessary to validate research findings”. Therefore, if software is needed to validate findings described in a publication, researchers are expected to make it available as widely as possible. There are some exceptions to this rule. For example, if there is an intention to commercialise the software there might not be a need to share it, but the default assumption is that the software should be shared.

The importance of putting a licence on software

It is important that before any software is shared, the creator considers what they would like others to be able to do with it. The way to indicate the intended reuse of the software is to place a licence on it. This governs the permission being granted to others with regards to source code by the copyright holder(s). A licence determines whether the person who wants to get hold of software is allowed to use, copy, resell, change, or distribute it. Additionally, a licence should also determine who is liable if something goes wrong with the software.

Therefore, a licence not only protects the intellectual property, but also helps others to use the software effectively. If people who are potentially interested in a given piece of software do not know what they are allowed to do with it, it is possible they will search for alternative solutions. As a consequence, researchers could lose important collaborators, buyers, or simply decrease the citation rate that could have been gained from people using and citing software in their publications.

Who owns the copyright?

The most difficult question when it comes to software licensing is determining who owns the copyright – who is allowed to license the software used in research? If this is software created by a particular researcher then it is likely that s/he will be the copyright owner. At the University of Cambridge researchers are the primary owners of intellectual property. This is however a very generous right – typically employers do not allow their employees to retain copyright ownership. Therefore, the issue of copyright ownership might get very complicated for researchers involved in multi-institutional collaborations. Additionally, sometimes funders of research will retain copyright ownership of research outputs.

Consequences of licensing

An additional complication with licensing software is that most licences cannot be revoked. Once something has been licensed to someone under a certain licence, it is not possible to take it back and change the licence. Moreover, if there is one licence for a set of software, it might not be possible to license a patch to the software under a different licence. The issue of licence compatibility sparked a lot of questions during the workshop, with no easy answers available. The overall conclusion was that whenever possible, mixing of licences should be avoided. If use of various licences is necessary, researchers are recommended to get advice from the Legal Services Office.

Good practice for software management

So what are the key recommendations for good practice for software management? Before the start of a research project, researchers should think about who the collaborators and funders are, and what the employer’s expectations are with regards to intellectual property. This will help to determine who will own the copyright over the software. Funders’ and institutional policies for research data sharing should be consulted for expectations about software sharing With this information it is possible to prepare a data management plan for the grant application.

During the project researchers need to ensure that their software is hosted in an appropriate code repository – for example, GitHub or Bitbucket. It is important to create (and keep updating!) metadata describing any generated data and software.

Finally, when writing a paper, researchers need to deposit all releases of data/software relevant to the publication in a suitable repository. It is best to choose a repository which provides persistent links e.g. Zenodo (which has a GitHub integration), or the University of Cambridge data repository (Apollo). It is important to ensure that software is licensed under an appropriate licence – in line with what others should be allowed to do with the software, and in agreement with any obligations there might be with any other third parties (for example, funders of the research). If there is a need to restrict the access to the software, metadata description should give reasons for this restriction and conditions that need to be met for the access to be granted.

Valuable resources to help make right decisions

Both Neil and Shoaib agreed that proper management and licensing of software might be sometimes complicated. Therefore, they recommended various resources and tools to provide guidance for researchers:

The workshop was organised in collaboration with Stephen Eglen from the Department of Applied Mathematics and Theoretical Physics (University of Cambridge) who chaired the meeting, and with Andrea Kells from the Computer Lab (University of Cambridge) who hosted the workshop.

The Research Data Service is also providing various other opportunities for our research community to pose questions directly of the funding bodies. We invited Ben Ryan from the EPSRC to come to speak to a group of researchers in May and the resulting validated FAQs are now published on our research data management website. Similarly, researchers met with Michael Ball from the BBSRC in August.

These opportunities are being embraced by our research community.

*About the speakers

Shoaib Sufi – Community Lead at the Software Sustainability Institute

Shoaib leads the Institute’s community engagement activities and strategies. Graduating in Computer Science from the University of Manchester in 1997, he has worked in the commercial sector as a systems programmer and then as software developer, metadata architect and eventually a project manager at the Science and Facilities Technologies Council (STFC).

Neil Chue Hong – Director at the Software Sustainability Institute

Neil is the founding Director of the Software Sustainability Institute. Graduating with an MPhys in Computational Physics from the University of Edinburgh, he began his career at EPCC, becoming Project Manager there in 2003. During this time he led the Data Access and Integration projects (OGSA-DAI and DAIT), and collaborated in many e-Science projects, including the EU FP6 NextGRID project.

Published 21 October 2015
Written by Dr Marta Teperek
Creative Commons License