Accurate and Reliable Classification of Unstructured Reports on Their Diagnostic Goal Using BERT Models.

Understanding the diagnostic goal of medical reports is valuable information for understanding patient flows. This work focuses on extracting the reason for taking an MRI scan of Multiple Sclerosis (MS) patients using the attached free-form reports: Diagnosis, Progression or Monitoring. We investigate the performance of domain-dependent and general state-of-the-art language models and their alignment with domain expertise. To this end, eXplainable Artificial Intelligence (XAI) techniques are used to acquire insight into the inner workings of the model, which are verified on their trustworthiness. The verified XAI explanations are then compared with explanations from a domain expert, to indirectly determine the reliability of the model. BERTje, a Dutch Bidirectional Encoder Representations from Transformers (BERT) model, outperforms RobBERT and MedRoBERTa.nl in both accuracy and reliability. The latter model (MedRoBERTa.nl) is a domain-specific model, while BERTje is a generic model, showing that domain-specific models are not always superior. Our validation of BERTje in a small prospective study shows promising results for the potential uptake of the model in a practical setting.

Modelling of Dynamic Behaviour in Magnetic Nanoparticles.

The efficient development and utilisation of magnetic nanoparticles (MNPs) for applications in enhanced biosensing relies on the use of magnetisation dynamics, which are primarily governed by the time-dependent motion of the magnetisation due to externally applied magnetic fields. An accurate description of the physics involved is complex and not yet fully understood, especially in the frequency range where Néel and Brownian relaxation processes compete. However, even though it is well known that non-zero, non-static local fields significantly influence these magnetisation dynamics, the modelling of magnetic dynamics for MNPs often uses zero-field dynamics or a static Langevin approach. In this paper, we developed an approximation to model and evaluate its performance for MNPs exposed to a magnetic field with varying amplitude and frequency. This model was initially developed to predict superparamagnetic nanoparticle behaviour in differential magnetometry applications but it can also be applied to similar techniques such as magnetic particle imaging and frequency mixing. Our model was based upon the Fokker-Planck equations for the two relaxation mechanisms. The equations were solved through numerical approximation and they were then combined, while taking into account the particle size distribution and the respective anisotropy distribution. Our model was evaluated for Synomag®-D70, Synomag®-D50 and SHP-15, which resulted in an overall good agreement between measurement and simulation.