Research projects in healthcare

Project 1: eHospital – Integrated hospital management software

Project 2: ePACS – Customization for an African University Hospital

Teaching hospitals have additional research and teaching responsibilities that can potentially affect the functionality requirements of a PACS system associated with diagnostic and imaging processes. The aim of this applied research project is to adapt open-source PACS software for use in the training of young internists specialized in medical imaging.

We are using the software from the Medical Physics Department of the Liège University Hospital in Belgium to provide a PACS service that has the advantage of being low in cost. We are customizing it to provide a teaching and testing environment for radiology interns at Donka Hospital.

Project 3: Tracking data, SMS and GIS interactions, patient follow-ups and viewer/dashboard for Donka

Challenge

Deploy open-source software to track infectious diseases in Guinea (Donka National Hospital).

Project 4: QnGene – Genetic format for HPC

Project involving an extension of the Adam format to include genotyping, clinical data and genetic analysis for precision medicine, on a large scale using machine learning.

Challenge

Analysis of big data from clinical studies is currently performed by bioinformatics scientists using script pipelines (often written in Python language) that are run on relational database technology. This approach is increasingly problematic because of the huge volume of genetic data to be processed. The AMPLab at the University of California, Berkeley published the Adam format, which leverages big data technologies to accelerate this process (see article here). Our challenge is to enable researchers to extend the data format themselves to include clinical, pharmaceutical, environmental and demographic data in order to run predictive models for precision medicine using large-scale machine learning algorithms (see article here).

What we did

This project begins with a meeting with the AMPLab team in Berkeley for knowledge transfer on the Adam project. At the start of the project, we modified the Adam format to accept genetic data from clinical trials, such as Advance and CDK Gene (see Simon Grondin's report, in French). We developed scripts to dynamically adjust the data schema by the researcher as well to load huge amounts of data quickly and efficiently at low cost (all the data are localized; for more information, see Fodil Belghait's article and watch this video).

We worked on validating/testing version 1.0 of this precision medicine platform using a case study involving the use of three machine learning algorithms aimed at identifying factors that could allow a predictive score to be discovered for type 2 diabetes.

Project 5: GenomeViewer – Somatic Vizualizer

Project involving the design of a query visualizer (exceeding the capacity of existing open-source software (LocusZoom and GWAS pipeline) using an interactive SNP. This tool allows researchers to explore public data in real time. Read this article presented at the ACM Digital Health conference in London.

Challenge

Health researchers would like to interact with genetic data, but this is made difficult due to the growing volumes of data. At present, they submit queries to bioinformatics scientists who use the SQL language to design queries that are run on relational databases (in the background) for each case, a process that is inefficient. It would be ideal to have access to visualization software that integrates all the available data during the discovery stage and that would allow researchers to interact easily with these huge quantities of heterogeneous private and public data, such as COSMIC, ENSEMBL, UCSC, Cancer Gene Census, Tumorscape, OMIM, cBioCancer, Mitelman, gnomAD and ExAc.

What we did

We designed and developed a genetic visualization software prototype called GenomeViewer. To do so, we analyzed the process and software used to perform genetic queries at Dr. Pavel Hamet’s laboratory. The aim of the fourth version of the prototype is to easily compare patient data with the data from 1,000 publicly available genomes. In its final version, GenomeViewer will enable cancer researchers to fully and quickly visualize somatic mutations and compare them against several public databases.

Beatriz produced a disposable prototype, the first version called “GOAT,” which then underwent several reengineering stages. The second version of the prototype was produced by Cédric Urvoy (read his report, in French), and the third by Victor Dupuy (read his report, in French), in order to establish a maintainable internal software architecture (see GOAT publication). The prototype was renamed “GenomeViewer” and the functional target was modified for the visualization of somatic mutations. The prototype upgrade was performed by a group of students as part of a final year project (in French) to replace the graphical presentation limited through the use of Bokeh Server technology by AmCharts and replace the current SQL BD technology with NoSQL-Spark to significantly increase query execution speed. The latest version of the prototype (fourth version) contains only data from the 1,000 genomes and is now running on an AWS cluster. This version allows .vcf files to be loaded. Data loading run time decreased from 6 hours to 30 minutes. In addition, interactive back-end queries decreased from 3 minutes to 5 seconds using the Berkeley Adam format on AWS-EMR. See their master’s thesis here (in French).

Many thanks to Amazon Web Services for free access to its instances in this project.

Project 6: PACIQ – Software for monitoring continuous quality improvement at Quebec healthcare facilities

Development of prototype software for monitoring the assessment of quality criteria based on the applicable standards of Accreditation Canada, BOMA BEST , Planetree and the Réseau québécois des établissements promoteurs de santé that regulate Quebec healthcare facilities.

Challenge

It is currently difficult to obtain accurate information from all healthcare professionals within an institution on the level of compliance, based on several quality aspects. In addition, the elements that require the institution’s compliance are often repeated from one standard to another, resulting in a great degree of duplication when entering, analyzing and generating preliminary reports for the compliance process. Developed with the assistance of the former director of CHU Sainte-Justice’s Direction de la qualité, this presentation identifies the needs that have been expressed (in French).

Project 7: CorlabNow

A real-time dashboard for monitoring blood test processing to ensure the quality of the Jewish General Hospital’s laboratory in Montreal. The software displays in real time the status of blood test processing wait times, namely, service levels.

The collected samples are sent to the Diagnostic Medicine Department, after which an order is entered for analysis. The filled vial basket is sent to the processing station. The vial is labelled and then processed by the diagnostic equipment.

Challenge

Human and material errors that sometimes occur during the process create additional work for the MPA technologist. There is no specific performance indicator or way to know if there is a backlog, or if additional staff is needed to meet the required service levels. In addition, some tests must be given priority (such as a troponin test for a heart attack in the emergency department).

What we did

Design of a platform for entering information, in real time, and presentation of indicators (KPIs) that allow test backlogs to be anticipated, to have a clear idea of the workload, and to enable analysis of historical data to be done and thus allow accurate planning of laboratory behaviour.

(See BI fair presentation, in French)

See dashboard run in real time.

Project 8: Mobile App - FixMyShoulder/FixMyKnee

I have experienced a number of shoulder problems. I was treated by George Demirakos, a renowned Montreal physiotherapist. He now publishes books on physiotherapy.

This project involves the development of a mobile application (iOS) for physiotherapist George Demirakos.

Challenge

Develop a mobile app available on the App Store that provides free and paid content related to the books published by George Demirakos on shoulder and knee physiotherapy.

What we did

We tested several technologies. An initial software framework was developed, known as CMS, that allows data to be updated for two applications (Android and iOS) simultaneously. Two identical mobile app prototypes were then developed, one for iOS and the other for Android.

• Julie Vincent: first iteration (Android and web platform); see report (in French).
• Mathieu Crochet: second iteration (Android and CMS platform); see report (in French).
• The most recent work involves redesigning the application on Ionic by Quentin Muret for iOS.