Christian Lovis

Activités de recherche

La miniaturisation des outils informatiques et l’augmentation de leur autonomie alimente une nouvelle révolution, le « pervasive computing ». L’informatique est ubiquitaire, elle devient invisible. Montres intelligentes, smartphones, tablettes, innombrables capteurs, nous utilisons ces outils au quotidien, ils sont devenus indispensables. Le Cloud computing, en permettant de déporter la puissance de calcul et le stockage, permet à ces outils mobiles d’avoir des capacités presque illimitées.

Tous ces systèmes contribuent à produire des quantités croissantes de données. Les données médicales, issues des dossiers patients informatisés et de la génomique peuvent potentiellement être combinées à des données personnelles issues de biocapteurs, à des informations sur l’environnement, sur le comportement. Les enjeux sont immenses. Des défis technologiques, liés aux volumes de données par exemple, mais surtout des défis scientifiques, comme la construction de la capacité à interpréter ces données, des défis éthiques, liés à la protection de la sphère privée, et finalement des défis sociétaux sur l’utilisation des résultats.

Les paradigmes d’interactions entre les données, l’information et la connaissance sont un autre sujet d’étude de l’équipe. Alors que la complexité de l’information est grandissante, que les outils sont ubiquitaires, les paradigmes d’interaction n’ont pas vraiment changé. Les systèmes sont essentiellement individuels, les interactions largement basées sur le modèle de clic, même lorsque la souris est remplacée par le doigt. L’exploration de nouvelles méthodes, comme la 3D, les interfaces sans contact, la réalité augmentée avec des lunettes ou encore les environnements multi-utilisateurs, le traitement de la dimensionnalité, sont des activités importantes et qui soutiennent plusieurs de nos projets.

C’est par des cycles de construction, d’évaluation et l’analyse d’impact et de correction que progressent ces systèmes. L’Evalab complète donc naturellement nos activités de recherche et développement en nous apportant toute l’expertise et les compétences requises à l’évaluation, et notamment en termes cognitifs, mais aussi d’efficience et de sécurité. En outre, l’Evalab s’ouvre lentement aux procédures de certifications et de marquage pour l’industrie.

L'activité académique prend son envol, en particulier avec plusieurs études et publications en cours et en préparation, dans le domaine de la prescription informatisée, des outils d'aide à la décision et d'analyse de l'impact des interfaces homme-machine.

Graph representation of Clinical Information System (CIS) Data

The data stored in the CIS are evolving constantly. From the early version of the CIS containing mostly unstructured information composed mainly of documents, its content got quickly more complex with more and more numerous information. In order to be able to deal with this evolution, the adopted solution was to rely on an attribute-relation-value triple that avoid too frequent refactoring of the database. However, if this representation on a relational database avoids frequent refactoring, it makes the extraction of data more and more resource demanding.

In order to ease the persistence and the evolution of the data of the CIS, we want to change our storage technology from a relational model to a graph oriented representation. This work will requires getting familiar with the graph database technology, to identify what is the most adapted framework and associated tool and to develop a prototype.

Domain: database, data representation, big data

Event based representation of Electronic Health Record (EHR)

The electronic health record (EHR) is essential for clinical activities but also for research, public health and invoicing. The EHR or the University Hospitals of Geneva has been constructed in a modular manner, integrating progressively the different domains of the medical activity. This has induced a presentation composed divided in several tabs, making sometime difficult to contextualize the information. Recently, efforts have been undergone to regroup information under consolidated views, inspired by cockpits. Although very valuable, these views cannot reflect the cooperative and multidisciplinary workflow of today medicine.

We propose, here to modify completely the traditional organization of the EHR in order to provide a view based on events and interactions. A review of the existing technologies capable to sustain such solution (Elgg, XOOPS,…) will be necessary as well as its implementation

Domain: Clinical information system, data representation, interface

Decision support and alerting in EHR‘s

Decision support systems (DSS) and in peculiar the positive predictive value of DSS and alerting is a major challenge in EHR’s. How to improve DSS, such as using context, semantic driven systems; and how to assess and improve alerting systems, such as selecting whom to alert, when and how building escalation, are still research problems that have not yet been solved.

Domain: Clinical information system, data representation, interface, decision support

Quality of information

Anesthesiology reports are essential documents for the good proceedings of the surgery in the hospital. The quality of the information contained in these reports must be flawless. These reports can be collected under several format, free text, structured questionnaire or electronic form.

In this work, we have previously gathered reports under the three formats for same patients in order to discover whether their content is similar. In order to evaluate the similarity of the content of the different reports, we will rely on natural language processing (NLP) techniques to automatically analyze the content and perform statistical analysis.

Domain: Natural language processing

Augmented reality glasses in healthcare

Augmented Reality is an emerging area in clinical medicine. The use of several methods to combine reality and informational source to provide contextualized information is not science fiction anymore and possess many clinical applications.

This work consists to get familiar with the goggle wearable devices and to implement a project linked to such technology. Among the possible applications, it is possible to provide clinical guidelines to caregivers through the Goggles.

The development of the application will requires getting accustomed with the development of android application to implement a new human machine interface using interaction mechanism adapted to the device.

Domain: interaction, android programming

Cooperative-collaborative interfaces

Domain: interaction, android programming

Innovative data representation

The multiplication of data available in medical context can create a cognitive overflow for the clinicians. In domain such as medical alert it has been demonstrated that too frequent alerts may be counterproductive. Displaying too much information out of a relevant context can complicate the identification of relevant information for the decision process. Following this global reflection, we are trying to develop advanced clinical data representation that facilitates the clinical decision process. One part of this initiative concerns the development of graphical representation of laboratory results. By taking into account much more contextual information, expert knowledge as well as innovative representation paradigm we are looking whether it helps clinical decision process.

Domain: interaction, user interface

 

Exploring spatio-temporal phenomena is not an easy task. Usually, people simplify the problem into smaller several problem of smaller dimension and represent these smaller problems using traditional representation technique. Geographical information system (GIS) has been demonstrated to be an efficient tool to manipulate such multidimensional data. In this work, we got inspired by GIS system but at the level of the hospital. We have develop a tool that allows to project any kind of information into a 3D model of the hospital and are able to see how these data evolve over the time. This representation has already been used to visualize how iatrogenic infection spread toward the institution and therefore taking appropriate measures to avoid their spread.

Domain: interaction, user interface

Interoperability of bio-captors and various sensing systems, including domotics

Monitoring people health through activity sensors is a growing tendency. If some platforms exist to collect the information generated by a variety of sensors, in most of the time the signals generated by these sensors are analyzed independently. The aggregation of the signals has undeniably many applications in healthcare, but their implementation isn’t a simple task due to the fast pace of evolution of the sensors market and its diversity.

In this project, we want to implement a hardware platform allowing a plug and play connection of a group of sensors related to a specific healthcare application. The architecture of the platform has been already defined during a previous project and this platform must be now feed with other captors in order to offer a sufficient range of sensors for an extended use

Domain: Sensors, Hardware, plug and play

Evaluation of bio-captors

Monitoring people health through activity sensors is a growing tendency. If some platforms exist to collect the information generated by a variety of sensors, in most of the time the signals generated by these sensors are analyzed independently. The aggregation of the signals has undeniably many applications in healthcare, but their implementation isn’t a simple task due to the fast pace of evolution of the sensors market and its diversity.

In this project, we want to implement a hardware platform allowing a plug and play connection of a group of sensors related to a specific healthcare application. The architecture of the platform has been already defined during a previous project and this platform must be now feed with other captors in order to offer a sufficient range of sensors for an extended use

Domain: Sensors, Hardware, plug and play

Communication App between care organization and citizen

SmartHUG is an HTML5 App devoted to improve global communication between HUG and the citizen that might be in contact with HUG services. It is built around a very simple concept:

a)       HUG communicates with citizens: news, medias, education, etc…

b)       Citizens communicates with HUG: social network aggregators, etc…

c)        HUG at the service of citizen: access to various services, such as outpatient facilities, phone calls, specialized services, geolocation and maps,

d)       Geneva health services: access to global resources in Geneva for various health services, such as ER wiat time, etc.

Domain: interaction, Design, HTML 5, social networks

InfoKIDS

InfoKIDS is developed with clinicians at pediatric emergency room and is devoted to ensure improved emergency experience. The project considers three phases of emergency encounter:

a)       Before emergency encounter, a mobile app to use at home to help parent’s decision and drive path to various care providing approaches, from self care, to private pediatrician, to hospital ER.

b)       During emergency encounter

c)        After emergency encounter, with informational follow up.

Domain: interaction, Design, HTML 5, social networks, interoperability, evaluation

Smartphone based activity recognition

Activity recognition is a rapidly growing field. It has many applications such as activity surveillance for a healthier lifestyle, surveillance of seniors to identify deterioration of their condition or surveillance of people with personality disorder to avoid crisis.

In this project, we attempt to identify patients’ emotional state based on their smartphone usage. This application is very important for the patients suffering from emotional disorders in order to provide them an appropriate treatment in case of crisis.

The participation to this project requires getting familiar with the mobile environment in order to collect the maximum of information (with the user agreement of course). In a second stage these data will be analyzed in order to identify whether this usage can be linked to behavior change

Domain: Mobile development

Communication of a mobile application with the SMART – FHIR HL7 Standard

The clinical information system (CIS) world is still much divided making difficult the communication between the applications manipulating clinical data.

In order to insure a minimal interoperability, standards such as HL7 have been developed to make the information understandable by the different systems. In parallel other standard such as SMART http://smartplatforms.org/) have emerged to facilitate the communication between the system. Recently, the main standard has undergone a new evolution known as HL7-FHIR (http://wiki.hl7.org/index.php?title=FHIR) and facilitates the interconnection between two systems.

In this work, we would like to implement an interface between the existing CIS and a mobile application managing the nurses’’ workflow.

Domain: health application standard

The EHR is healthcare’s Big Data. It is a repository of highly valuable information. Unfortunately, this repository is rarely used for a patient’s benefit. So, we develop “Big Data” tools of data upcycling in order to find actionable intelligence. A very few known health systems have built their own EHRs (Electronic Health Records) that delivered actual clinical decision support. This process usually takes years. Today, major EHR vendors still do not provide adequate clinical decision support.

 

HUG is comprising more than 2’000 beds. According to annual statistics, approximately 60’000 admissions and over 900’000 outpatient visits took place and over 6’000'000 of laboratory findings were accomplished. The total number of full-time employees is ≈9’000, where the number of physicians is ≈ 1’300. These are the sources of Big Data in healthcare we have to face with. We focus on the bridging of the HUG’s legacy EHR system to Clinical Research Environment. We develop approaches of how to establish a coherent data pipeline, from internal/external data providers. The objective is to embrace upcycling process from data analytics to data exploitation.

Domain: Biomedical data mining

Information technology entered relatively late the work environments of health care professionals. For too long physicians and nurses relied on paper based documentation. Geneva University Hospitals were one of the world wide precursors to introduce electronic patient records to their health care professionals.

It is well agreed on that health care technology can reorganize work processes in the hospital environment for the better. In numerous large scale studies it has been shown that the introduction of clinical information systems can improve patient security, efficiency, and satisfaction. However, when introducing a novel technology it is often unclear what impact it will have on the work of health care professionals.

Human factors engineering is the scientific field where the interaction between humans, in our case health care professionals, and technology is studied. The most crucial factor to be taken into account is the safety of the device. Like any drug, medical devices and also software need formal evaluation to limit the risks that they harm patients or health care professionals. Too often, medical errors often happen following an erroneous use of a device or software. A single laps or a misinterpretation can lead to hazardous situations.

In a broader sense, human factors is studying the aspect “ease of use” of medical devices and software. Health care professionals should be equipped with devices and software that can be easily, efficiently, and successfully used.

In this context research is required. The impact of the computerization on physicians' and nurses' work processes must be studied. The goal of a usability evaluation is to apply objective, scientific methods that are based on cognitive psychology and that integrate the work domain knowledge of health care professionals in order to evaluate and improve medical software and devices.