Article
Integration of AI platforms with clinical IT: Process modelling for generic use cases
Search Medline for
Authors
Published: | August 19, 2022 |
---|
Outline
Text
Introduction: For several years, a significant increase in the use of AI methods can be observed in many industries, which includes the health sector [1]. Various systems have already been proposed to use AI applications in the healthcare domain [2], [3]. However, there is a gap that needs to be filled to introduce AI applications into the clinical environment. Our approach proposes an AI platform that acts as a complement to existing clinical source systems (e.g., HIS, PACS) and medical data integration centers (MeDICs) [4] that are currently in development. Our AI platform will not only regulate the inference and clinical decision support, but also facilitate the training process of clinical data.
Methods: To illustrate the interaction between the three mentioned system landscapes, we have created process diagrams in the form of BPMN (Business Process Model and Notation) diagrams. For this purpose, four generic use cases were identified that illustrate the entire workflow in the context of the overall process [5].
Results: These generic use cases are Data Selection, Data Annotation, Training/Testing, and Inference, with the latter being differentiated into Single, Semi-automated, and Automated Inference.
Data Selection describes how a cohort data set is queried and transferred from the MeDIC to the AI platform. The cohort data set is then stored pseudonymized in its Repository for Temporary Data. The user can subsequently evaluate the cohort data set and store the associated information in the MeDIC.
Data Annotation describes how the previously compiled cohort data set can be annotated in the AI platform with the aid of an annotation tool. In this process, the AI platform interacts with the MeDIC to retrieve the data and, after annotation, to store the corresponding metadata in the MeDIC registry and the annotation files with the annotation information in the MeDIC repository.
Training/Testing describes how the reviewed and annotated cohort data set can be retrieved from the MeDIC. Furthermore, it is illustrated how training and testing can be performed within the AI platform. Then, the trained and tested AI model is stored in the permanent« repository of the AI platform. In addition, the generated technical and medical documentation of the AI model is stored in the MeDIC.
The inference describes the prediction process for single, semi-automated, and fully automated cases. It is demonstrated how the data from the clinical source systems can be transferred to the AI platform in order to first make a prediction with the appropriate AI model and then store the log and result in the MeDIC. Finally, the result is visualized and displayed to the user.
Discussion: The resulting BPMN diagrams can be seen as the blueprint to derive more specific use cases. Moreover, these BPMN diagrams are used in a regional AI project (KI-SIGS) project [6] as a basis for the system architecture of an AI platform.
Conclusion: The BPMN diagrams illustrate the interaction between the three system landscapes. Therefore, most of the functional aspects of the AI platform are covered. Future improvements may include other aspects, such as federated learning or security and data protection aspects.
The authors declare that they have no competing interests.
The authors declare that an ethics committee vote is not required.
References
- 1.
- Bohr A, Memarzadeh K. The rise of artificial intelligence in healthcare applications. Artificial Intelligence in Healthcare. 2020;25-60. DOI: 10.1016/B978-0-12-818438-7.00002-2
- 2.
- Scherer J, Nolden M, Kleesiek J, et al. Joint Imaging Platform for Federated Clinical Data Analytics. JCO Clin Cancer Inform. 2020;4:1027-1038. DOI: 10.1200/CCI.20.00045
- 3.
- Gruendner J, Schwachhofer T, Sippl P, et al. KETOS: Clinical decision support and machine learning as a service - A training and deployment platform based on Docker, OMOP-CDM, and FHIR Web Services [published correction appears in PLoS One. 2019 Nov 13;14(11):e0225442]. PLoS One. 2019 Oct 3;14(10):e0223010. DOI: 10.1371/journal.pone.0223010
- 4.
- Haarbrandt B, Schreiweis B, Rey S, et al. HiGHmed - An Open Platform Approach to Enhance Care and Research across Institutional Boundaries. Methods Inf Med. 2018;57(S 01):e66-e81. DOI: 10.3414/ME18-02-0002
- 5.
- Amershi S, Begel A, Bird C, et al. Software Engineering for Machine Learning: A Case Study. In: 2019 IEEE/ACM 41st International Conference on Software Engineering: Software Engineering in Practice (ICSE-SEIP). 2019. p. 291-300. DOI: 10.1109/ICSE-SEIP.2019.00042
- 6.
- KI-SIGS. Space for Intelligent Health Systems. [cited 2022 Apr 11]. Available from: https://ki-sigs.de/