Machine learning for boosting nanomedicine research

Important results from the project

During the project, most objectives were achieved, improving particle detection sensitivity beyond the current state-of-the-art. However, low signal-to-noise in scattering images prevented performance enhancement to the expected level. Objectives achieved: High-quality labelled data generated for nanoparticles. Simulated data generation for model evaluation. Exploration and optimization of various denoising and particle detection approaches. Model performance comparison with existing methods. Integrating U-Net based detection algorithm in Nanolyze’s routine.

Expected long term effects

We were able to improve particle detection beyond our current particle detection sensitivity. However, machine-learning approaches require more computational resources and can be slow in comparison to the classical methods. Nevertheless, the improved particle detection based on machine learning allows us to detect smaller nanoparticles and will help us expand adoption of our technology in pharmaceutical quality control. We believe that the knowledge gained from this project will not only improve our technology but also contribute to the advancement of nanomedicine research.

Approach and implementation

We believe the project was designed and implemented well with the right resources and personnel. In addition to Nanolyze, the project team included experts from Chalmers Industriteknik and AstraZeneca, who contributed their expertise in machine learning and pharmaceutical applications, respectively. The project included activities like data collection, algorithm development, and performance verification and we were able to execute the project smoothly and on time.