Project: Development of a semantic web module and PACS system for radiotherapeutic practice

Over the last decades the introduction of digital radiology and digital pathology as well as the 'age of -omics' (genomics, proteomics, microbiomic, etc) gave rise to massive data repositories in healthcare. In radiology in particular, patients are receiving more and more scans who are all digitally stored for multiple years. Moreover, the complexity of data has grown substantially due to the increase in resolution of e.g. CT and MRI and the introduction of hybrid imaging (e.g. the combination of PET and MRI or PET and CT), 3D and even 4D data (e.g. spatial and temporal registration with ultrasound or MRI). Altogether, this has resulted in a typical amount of radiological data in the order of tens of terabytes for an average sized hospital._x000D__x000D_This bulk of data is called 'big data' which is meant to represent both its sheer size and its huge value. Based on this data more accurate diagnosis and better treatment planning, treatment assessment and treatment personalisation can be achieved by means of clinical decision support systems. From an R&D perspective the data contains unprecedented opportunities for comparative effectiveness and cost-effectiveness studies, predictive modelling, statistical tools for trial design and biomarker and drug discovery. Altogether this data can hugely contribute to both the quality and cost reduction of healthcare. The overall annual economic value of big data for the US health care system was estimated by McKinsey at a massive $300 billion. It can reasonably be expected that 'big data' holds a similar economic value for the European healthcare system. [James Manyika et al., McKinsey Global Institute, May 2011]_x000D__x000D_There is however one major issue that hampers leveraging this value: It is currently largely impossible to optimally analyse and process the data. This primarily lies in the syntactical nature of the data structure that was designed in the early 80's and is still the absolute golden standard in radiology: DICOM. _x000D__x000D_To explain the limitation of DICOM we will use the following analogy: Let's say you would be interested in the price of the latest blackberry phone and would query a regular search engine such as Google for 'blackberry price'. The search engine would return a list of websites with both fruit and phone related content as blackberry can have both meanings. This is because Google is syntactical driven: it does not understand the meaning of the word blackberry, it just searches its database for exact matches of the words in the query. On the contrary, a semantic search engine would understand the meaning of your search query as well as the information on websites and would yield an actual answer to your question instead of a multitude of sites that potentially contain relevant information._x000D__x000D_The same holds for the DICOM standard, although it allows metadata, there are very limited possibilities to give data meaning through annotation. Querying a DICOM file repository for 'What is the average tumour shrinkage of a tumour of type T undergoing both radiotherapy X and chemotherapy Y after Z weeks?' is thus virtually impossible. By introducing semantics by including the meaning of data (e.g. 'these pixels represent a tumour'), this would become possible and the potential of the data can finally be fully leveraged. The technology enabling this is called semantic web technology and is on the rise in a multitude of sectors handling 'big data'. In healthcare however, accurate solutions are not yet commercially available._x000D__x000D_Additional difficulty in healthcare is the diversity of data silos that have been piling up over the years. In current clinical practice it is common that each department (e.g. radiology, histology, nuclear medicine, radiotherapy, etc) uses different silos from different vendors with different data formats. The ability to integrate and simultaneously query these heterogenous types of data would yield even more opportunities to leverage the value of the data._x000D__x000D_Recently SOHARD Software GmbH, an expert company in semantic software development, has presented a concept for a semantic adapter for DICOM-based systems that would allow semantic querying of multiple existing DICOM based data repositories. This resulted in an overwhelming response from both industrial and academic hospitals proving the urgent need for such a system. Together with MAASTRO, a frontrunning and renowned radiotherapy clinic and research institute, SOHARD will further develop the software and validate its value in radiotherapeutic clinical practice. The project will result in the first validated semantic web adapter and stand-alone PACS system that will enable unprecedented ways to access and analyse medical data and finally leverage its true value for patient, industry, academia and society.

Acronym	SEDI (Reference Number: 7831)
Duration	01/04/2013 - 31/03/2015
Project Topic	SOHARD Software GmbH, an expert company in semantic software development for healthcare, and MAASTRO, a renowned radiotherapy clinic and research institute, together will develop a semantic web adapter and a stand-alone PACS system that enables unprecedented ways to access and analyse medical data.
Project Results (after finalisation)	- The first implemented semantic web system in radiotherapeutic practice_x000D_- The validation of the value of semantic web technology in radiotherapeutic practice_x000D_- The first open-source ontology for radiotherapy as well as a system in place to manage it_x000D_- The first semantic adapter translating semantic queries (SPARQL) into DICOM R/Q and vice versa ready for CE approval
Network	Eurostars
Call	Eurostars Cut-Off 9

Project partner

Number	Name	Role	Country
2	MAASTRO	Partner	Netherlands
2	SOHARD Software GmbH	Coordinator	Germany