Lund University trained neural networks in MATLAB to match heart donors and recipients.
Better donor-recipient matching was needed to improve transplant survival, and earlier open-source tools crashed during multiday simulations. The team used MATLAB, Deep Learning Toolbox and Parallel Computing on a 56-processor cluster. The page reports 5-10% better five-year survival for ANN-selected patients, simulations cut from 3-4 weeks to about 5 days, and 20% higher transplant candidacy than traditional criteria.
MathWorks makes MATLAB and Simulink, used by pharma and medical device teams for modelling, machine learning and regulated device software development
This sits in the Build stage of the AI use-case lifecycle: Named examples from the foundation-model labs and platforms pharma teams use to actually construct, train and generate the model or application.
Published by MathWorks on its own site — read the original case study on mathworks.com ↗
Related use cases
- 01Genentech built a MATLAB supervisory control platform for microbial fermentations that cut algorithm rework from six months to two weeks.
- 02MedImmune automated flow cytometry analysis in MATLAB, cutting turnaround from weeks to minutes.
- 03Pfizer applies systems pharmacology modelling in MATLAB and SimBiology to reduce phase II attrition.
- 04Medviso built cardiac MR/CT analysis software in MATLAB that won FDA 510(k) clearance and CE marking.
- 05Battelle decoded brain signals with MATLAB machine learning to restore hand movement in a paralysed patient.
- 06Preceyes developed the control system of an eye-surgery robot with MATLAB and Simulink model-based design.