The research & innovation department of RADar, the learning and development center in AZ Delta, incorporates new technologies and data-driven medicine into healthcare operations. RADar focuses on data-driven healthcare research projects, using cutting-edge technology to gather, store & analyze data and gain insights into healthcare trends and patient outcomes.

Our mission is to support physicians and healthcare providers in preventing, detecting, and treating disease in a personalized and precise way. We believe that by leveraging cutting-edge technology and data-driven insights, we can provide optimal and affordable health to all.

The RADar research & innovation team is composed of a multidisciplinary team of experts in data science, AI technology, healthcare & technology, collaborating closely to pursue impactful research outcomes.

RADar research projects aim to leverage the power of data to transform the healthcare industry and improve patient care. Spearheads are artificial intelligence (A.I.), data science, mobile health (remote mHealth care), nutrition&health and business intelligence & planning.

Advanced Data-Aided Medicine Part Lung Cancer

• Objective: Developing and validating an AI model that supports physicians in their decision process for treating lung cancer patients. This AI model needs to predict the probability/evolution of the outcomes, based on clinical data and simulated lung cancer treatment plans. The outcome probabilities can be evaluated through different treatment plans to identify the optimal treatment plan for a patient.

• Methodology: Automatic unlocking, collection & transformation of lung cancer datapoints to OMOP common data model so that data is readily available for further research & analysis. Training and validating supervised machine learning models with a limited feature set as input to predict lung cancer patient outcomes. Constructing a digital patient by training an AI model fed with all data available in OMOP common data model. Validating a digital patient & optimal feature selection to enhance AI model performance via unsupervised learning techniques

• Impact: Careful analysis of the collected data using artificial intelligence tools might support physicians & lung cancer patients in their treatment decision process.

• Partners: UGent IDLab

• RADar project research lead: Louise Berteloot

• Principle investigator: Dr. Ingel Demedts

• Timeline: 2021-2024

• Funding: Vlaio O&O

• Ethics: NCT05783024 Advanced Data-Aided Medicine Part Lung Cancer- ClinicalTrials.gov

• Publication: abstract in Journal of Thoracic Oncology

• Status: research ongoing

Advanced Data-Aided Medicine Part Coronary Artery Disease

• Objective: Developing and validating an AI-driven clinical decision support tool for the optimization of treatment plans for patients with coronary artery disease. The AI-model should predict clinically relevant outcomes, such as mortality, rehospitalizations and patient-reported outcome measures (PROMs), using multiple data sources, such as hospital records, ECG recordings, laboratory values…

• Methodology: Transformation of electronic health record data to the OMOP Common Data Model, which is built on a foundation of standardized vocabularies and a simple patient-centric data model. This is done by in-house development of open-source ETL-tools, shared in the OHDSI-community. Data is extracted using our cohort extraction tool with custom clinical phenotypes.

Augmenting state-of-art deep learning models for ECG classification with structured data extracted from our OMOP dataset. Limitations in dataset size are overcome by first optimizing the model architecture on related tasks with larger datasets, followed by pre-training the model on more generic tasks.

• Impact: This clinical decision support might improve standardization of treatment plans across similar patients and guide decisions for clinically more complex cases.

• Partners: UGent IDLab

• RADar project research lead: Ir. Stijn Dupulthys, MD

• Principle investigator: Karl Dujardin, MD

• Timeline: 2021-2024

• Funding: VLAIO O&O

• Publication: https://academic.oup.com/europace/advance-article/doi/10.1093/europace/euad354/7469373

• Status: research ongoing

Prostate Cancer Support Tool

• Objective: ProCaTS aims to demonstrate the feasibility of a data/A.I.-driven decision support tool for personalized and optimal treatment choice in clinical practice in prostate cancer. The project envisions developing an explainable A.I. and combining it with a dashboard tailored to clinicians' needs. This will increase the adoption of artificial intelligence (A.I.) in clinical practice

• Methodology: Unlocking and linking data from underlying silos in an automated way, with 90% accuracy and completeness and based on the OMOP CDM. Training explainable ML models, which can be used to predict patient outcomes. Develop a decision support tool, which objectively informs physicians.

• Impact: Acceptance of physicians to use the tool, measured by tracking their actions in the software and aiming for >50% of patients to be consulted this way. Partners anticipate further development of the solution and a pathway to be planned to have the solution medically approved so that it can be rolled out to other hospitals

• Partners: Siemens Healthineers, ML6

• RADar project research lead: dr. Nathalie Mertens

• Principle investigator: Dr. Wim Van Haute

• Timeline: 2022-2023

• Funding: projectbudget 2021 Flanders.healthTech Vlaio, MEDVIA

• Status: research ongoing

Breast Cancer Support

• Objective: The BreaCS project aims to develop a quasi-real-time A.I. based clinical decision support system (CDSS) for breast cancer treatment selection. The CDSS will combine radiology, pathology and clinical data into a unique smart A.I. solution.

• Methodology: This project aims to explore the potential of a clinical decision support system, develop solutions to practical problems of using a CDSS, and evaluate the performance multicentric. 3 AI models will be developed to predict at pre-operative timing 1 the pathological tumor size, 2 the need to remove axillary glands and 3 chemotherapy treatment effects

• Impact: In Belgium, 1 in 7 women will be diagnosed with breast cancer during lifetime. This project will evaluate an A.I. based clinical decision support system (CDSS) multicentric. If further clinical validation and certification steps are finalized, this could mean that medical interventions could be avoided, treatments would be more effective & personalized.

• Partners: AZ Groeninge, GZA, Endare

• RADar project research lead: Sandra Steyaert

• Principle investigator: Dr. Barbara Bussels

• Timeline: 2023-2024

• Funding: projectbudget 2022 Flanders.healthTech Vlaio, MEDVIA

• Status: data collection & research started

• Objective: Develop a PNH screening tool on the AZ Delta Electronic Health Record (EHR) system to enhance identification of patients with high PNH risk factors and to reduce the delay in PNH diagnosis.

• Methodology: Developing an algorithm to identify a high-risk cohort of potential PNH patients who need treatment from all registered patients, with maximum ability to find relevant cases. Secondly, this high-risk cohort will be manually reviewed by clinicians for final screening.

• Impact: Enhancing early detection of PNH diagnosis since Paroxysmal Nocturnal Hemoglobinuria (PNH) is a life-threathening hematological disorder. Prevalence is estimated between 1-5 per million people, often manifested by cardiovascular, gastrointestinal, neurological or haematological symptoms. Referral is therefore typically to several specialists, resulting in PNH underdiagnosis.

• RADar project research lead: ir Kim Denturck

• Priniciple investigator: Dr. Dries Deeren

• Timeline: 2021-2023

• Funding: Alexion, Pharma Belgium provides financial support

• Status: finalizing research report & preparing research publication. Considering next steps: finetuning the algorithm and converting the system to a prospective PNH monitor. Additionally, a probability estimator ranking patients in decreasing order of potential PNH risk can be developed.

• Objective: Developing an AI model to detect Epidermal Growth Factor Receptor (EGFR) mutation from Hematoxylin and Eosin (H&E) slides.

• Methodology: Unsupervised tumor segmentation, feature extraction with pretrained ResNet-50 and attention-based multiple instance learning for EGFR mutation classification.

• Impact: For therapeutic decision making in NSCLC, molecular profiling is crucial, as a number of gene alterations are eligible for targeted therapy. However, the molecular testing is expensive, time consuming and requires enough tumor tissue for testing. AI-based detection of gene alterations e.g. pathogenic/likely pathogenic EGFR mutations in H&E stained WSIs of NSCLCs could make a fast estimation of the odds of an actionable target being present and could ideally even predict the type of mutation (L858R, exon19 deletion or other).

• RADar project research lead: Louise Berteloot

• Principle investigator: Dr. Franceska Dedeurwaerdere

• Timeline: 2023-2024

• Funding: AstraZeneca

• Status: ongoing research

Phoniatrics Artificial Intelligence Detection tool

• Objective: Through PhonAID, our objective is to transfer specialized knowledge from secondary care to primary care, enabling early detection of voice disorders. Achieving this goal involves implementing an application equipped with an AI model.

• Methodology: Combining audio recordings of speech tasks with clinical data and questionnaires as input to an AI model to classify categories of voice disorders. A population of individuals with a healthy voice will constitute a separate category to confirm the absence of a voice disorder. The AI model will be implemented in a user-friendly application for patients and general practitioners to use. The patients will enter clinical data, answer questionnaires and perform several speech tasks on which the AI model will make a prediction.

• Impact: The voice serves as a crucial means of communication, and its impairment can lead to social challenges and hinder professional performance. Additionally, a voice disorder may indicate an underlying lung or neck tumor. Early detection is strongly advised to preserve quality of life and enhance patient outcomes.

• Partners: EIORL, Goomyx, ELG De Piramide

• RADar project research lead: Louise Berteloot

• Principle investigator: Dr. Lieve Delsupehe

• Timeline: 2024-2025

• Funding: projectbudget 2023 Medvia

• Status: starting up

• Objective: Development of a ML model to predict disease progression and treatment outcome for patients with Crohn’s disease or ulcerative colitis. The goal is to improve care in combination with a monitoring tool in the form of a patient population dashboard.

• Methodology: Extracting relevant information from the UR-Care database and from the Electronic Health Record (EHR) system of AZ Delta to visualize patient and patient population characteristics in a dashboard. The project also explores the potential of combining the dashboard with a clinical decision support system by training explainable ML models to predict disease outcomes.

• Impact: Analysis of the ML models in combination with the dashboard might support physicians in finding the best suited treatment options for specific patients.

• RADar project research lead: Ir. Hanne Vanluchene

• Principle investigator: Dr. Filip Baert

• Timeline: 2024-2025

• Funding: Takeda Belgium NV

• Status: Research ongoing

RADar wants to share its expertise and knowledge to other hospital and healthcare partner in an open source/open science philosophy. RADar invites researchers and other developers to further explore its available source code and provide feedback.

Rabbit-in-a-Blender RiaB An ETL data transformation tool to transform electronic health record data into observational medical outcomes partnership (OMOP) Common Data Model (CDM) tables.

Lib_airflow Custom Airflow operators and hooks, that help to upload raw hospital data to Google Cloud environment. The source code supports developers to automate workflows or data pipelines via the creation of advanced task scheduling.

DeltaMSI Microsatellite instability scoring screening tool using AI to score regions and samples. DeltaMSI has 3 modes: training, prediction and evaluation.

AZDelta-OMOP-CDM The provided source code holds the OMOP CDM folder structure containing the necessary ETL queries, USAGI csv files and custom concept csv files of the AZ Delta hospital. The folder structure is used in combination with the RiaB ETL tool

Explainable-Decision-Support-Lung-Cancer This repository contains the code and trained pipelines from the paper "Development of an explainable clinical decision support tool for advanced lung cancer patients" -paper under peer review submission stage. The folder pipelines contains the pipelines (preprocessing + voting classifier) of which the results were mentioned in the paper. If one uses these pipelines to run inference on own databases, please provide feedback & let us know your results.

Check our latest news and project updates via RADar Learning & innovation Centre AZ Delta: Overview | LinkedIn

We would love to hear your feedback on our RADar-AZDelta github.com code and welcome any suggestions for improvements. 

📢 RADar, AZ Delta, invites you to EmP2024, the second edition of Engineer meets Physician. Explore the dynamic world of healthcare data and AI with renowned (inter)national speakers. Embrace the opportunity to collaborate, learn, and drive positive patient outcomes!

🗓 Save the Date: May 28-29, 2024 | Virtual and Live Attendance Available

❓ Why Attend?

➡️ Unlock invaluable insights from industry leaders

➡️ Foster collaboration between forward-thinking engineers & physicians

➡️ Stay at the forefront of healthcare innovation

➡️ Expand your professional network and connections

🌍 Check out our website for more details, including the full program:

🐦 Early Bird Specials available until March 1st - Secure your spot now! Register here: https://lnkd.in/e7nQybWZ

📧 For sponsorship inquiries, contact emp2024@azdelta.be.

Check our latest news and project updates via RADar Learning & innovation Centre AZ Delta: Overview | LinkedIn

If you're interested in collaborating or partnering with us, or want to join our team in our quest to transform healthcare and make a difference in people’s lives, we would be delighted to hear from you. We also offer exciting career internships, and traineeships. Feel free to get in touch with us at radar@azdelta.be.