An innovative artificial intelligence-based method to compress complex models into explainable, model-agnostic and reduced decision support systems with application to healthcare (NEAR).

BACKGROUND AND OBJECTIVE: In everyday clinical practice, medical decision is currently based on clinical guidelines which are often static and rigid, and do not account for population variability, while individualized, patient-oriented decision and/or treatment are the paradigm change necessary to enter into the era of precision medicine. Most of the limitations of a guideline-based system could be overcome through the adoption of Clinical Decision Support Systems (CDSSs) based on Artificial Intelligence (AI) algorithms. However, the black-box nature of AI algorithms has hampered a large adoption of AI-based CDSSs in clinical practice. In this study, an innovative AI-based method to compress AI-based prediction models into explainable, model-agnostic, and reduced decision support systems (NEAR) with application to healthcare is presented and validated. METHODS: NEAR is based on the Shapley Additive Explanations framework and can be applied to complex input models to obtain the contributions of each input feature to the output. Technically, the simplified NEAR models approximate contributions from input features using a custom library and merge them to determine the final output. Finally, NEAR estimates the confidence error associated with the single input feature contributing to the final score, making the result more interpretable. Here, NEAR is evaluated on a clinical real-world use case, the mortality prediction in patients who experienced Acute Coronary Syndrome (ACS), applying three different Machine Learning/Deep Learning models as implementation examples. RESULTS: NEAR, when applied to the ACS use case, exhibits performances like the ones of the AI-based model from which it is derived, as in the case of the Adaptive Boosting classifier, whose Area Under the Curve is not statistically different from the NEAR one, even the model's simplification. Moreover, NEAR comes with intrinsic explainability and modularity, as it can be tested on the developed web application platform (https://neardashboard.pythonanywhere.com/). CONCLUSIONS: An explainable and reliable CDSS tailored to single-patient analysis has been developed. The proposed AI-based system has the potential to be used alongside the clinical guidelines currently employed in the medical setting making them more personalized and dynamic and assisting doctors in taking their everyday clinical decisions.

Say Goodbye to the 'Paper on Screen', Rethinking Presentation of and Interaction with Medical Information.

Traditionally, Electronic Medical Records (EMR) have been designed to mimic paper records. Organizing and presenting medical information along the lines that evolved for non-digital records over the decades, reduced change management for medical users, but failed to make use of the potential of organizing digital data. We proposed a method to create clinical dashboards to increase the usability of information in the medical records. Official clinical guidelines were studied by a working group, including dashboard target users. Necessary clinical concepts contained in the medical records were identified according to the clinical context and finally, dedicated technical tools with standard terminologies were used to represent categories of information. We used this method to generate and implement a dashboard for sepsis. The dashboard was found to be appropriate and easy to use by the target users.

Digital Connecting for Health, an Open Platform Based on Data Integration and Standards to Adopt Digital and Telehealth Solutions in the Healthcare Ecosystem.

The paper presents a collaborative approach employed to identify and examine the obstacles faced by telehealth solutions. The study involved the active participation of health start-ups, telehealth providers, and healthcare professionals delivering telehealth services. By harnessing the collective expertise and diverse perspectives of these stakeholders, the research led to develop an open platform, entitled Digital Connecting for Health, that has the potential to overcome the challenges impeding the widespread adoption and effectiveness of digital health services including telehealth in delivery of care. The developed platform shed light on various obstacles faced by telehealth solutions and provide valuable infrastructures for enhancing the implementation and efficacy of various digital health solutions, including telehealth applications, from various providers.

Towards a Clinically Meaningful Model to Structure the Development of Interoperable Order Sets, Applicable to the Point of Care in Any EMR.

Standardized order sets are a pragmatic type of clinical decision support that can improve adherence to clinical guidelines with a list of recommended orders related to a specific clinical context. We developed a structure facilitating the creation of order sets and making them interoperable, to increase their usability. Various orders contained in electronic medical records in different hospitals were identified and included in different categories of orderable items. Clear definitions were provided for each category. A mapping to FHIR resources was performed to relate these clinically meaningful categories to FHIR standards to assure interoperability. We used this structure to implement the relevant user interface in the Clinical Knowledge Platform. The use of standard medical terminologies and the integration of clinical information models like FHIR resources are key factors for creating reusable decision support systems. The content authors should be provided with a clinically meaningful system to use in a non-ambiguous context.

Clinical Knowledge Platform (CKP): A Collaborative Ecosystem to Share Interoperable Clinical Forms, Viewers, and Order Sets with Various EMRs.

A large number of Electronic Medical Records (EMR) are currently available with a variety of features and architectures. Existing studies and frameworks presented some solutions to overcome the problem of specification and application of clinical guidelines toward the automation of their use at the point of care. However, they could not yet support thoroughly the dynamic use of medical knowledge in EMRs according to the clinical contexts and provide local application of international recommendations. This study presents the development of the Clinical Knowledge Platform (CKP): a collaborative interoperable environment to create, use, and share sets of information elements that we entitled Clinical Use Contexts (CUCs). A CUC could include medical forms, patient dashboards, and order sets that are usable in various EMRs. For this purpose, we have identified and developed three basic requirements: an interoperable, inter-mapped dictionary of concepts leaning on standard terminologies, the possibility to define relevant clinical contexts, and an interface for collaborative content production via communities of professionals. Community members work together to create and/or modify, CUCs based on different clinical contexts. These CUCs will then be uploaded to be used in clinical applications in various EMRs. With this method, each CUC is, on the one hand, specific to a clinical context and on the other hand, could be adapted to the local practice conditions and constraints. Once a CUC has been developed, it could be shared with other potential users that can consume it directly or modify it according to their needs.

Brain SegNet: 3D local refinement network for brain lesion segmentation.

MR images (MRIs) accurate segmentation of brain lesions is important for improving cancer diagnosis, surgical planning, and prediction of outcome. However, manual and accurate segmentation of brain lesions from 3D MRIs is highly expensive, time-consuming, and prone to user biases. We present an efficient yet conceptually simple brain segmentation network (referred as Brain SegNet), which is a 3D residual framework for automatic voxel-wise segmentation of brain lesion. Our model is able to directly predict dense voxel segmentation of brain tumor or ischemic stroke regions in 3D brain MRIs. The proposed 3D segmentation network can run at about 0.5s per MRIs - about 50 times faster than previous approaches Med Image Anal 43: 98-111, 2018, Med Image Anal 36:61-78, 2017. Our model is evaluated on the BRATS 2015 benchmark for brain tumor segmentation, where it obtains state-of-the-art results, by surpassing recently published results reported in Med Image Anal 43: 98-111, 2018, Med Image Anal 36:61-78, 2017. We further applied the proposed Brain SegNet for ischemic stroke lesion outcome prediction, with impressive results achieved on the Ischemic Stroke Lesion Segmentation (ISLES) 2017 database.