Artificial intelligence in the analysis of time-resolved (4D) micro-computed tomography data

Artificial intelligence in the analysis of 4-dimensional micro-CT data – new insights into tablet disintegration

Today’s research relies more and more on the generation and processing of large amounts of data. The challenge of processing big data gave rise to the application of artificial intelligence. Especially in areas that exceed the capabilities of classical approaches such as the analysis medical image data, specifically tomography data, computer vision diagnostics based on AI find widespread application. The main focus thus far has been on static image classification. However, without the consideration of time-resolved data analysis further improvement of this analytical method is stymied.

This work offers a potential solution to the aforementioned problem. Our approach notably makes use of artificial intelligence for segmentation of highly heterogeneous time-resolved (4D) micro-CT data such as mini-tablets undergoing disintegration. The trained convolutional neural network handles the image segmentation quickly, consistently and outperforms humans in many aspects, demonstrating the untapped potential of using artificial intelligence coupled with time-resolved CT in pharmaceutical applications.

The neural net was trained and validated on data acquired at the Swiss Light Source high brilliance synchrotron facility at the Paul Scherrer Institute in Villingen.

We have established a method for handling and segmenting massive amounts of 4D micro-CT data, rapidly and consistently delivering accurate results thereby enabling and facilitating data analysis. This methodology can streamline existing time-resolved tomographical image analysis pipelines by introducing massively parallelized AI algorithms.

A detailed description of the methodology as well as the CT dataset will be published as part of a publication that is currently in the works.

 Download the full poster as a PDF here

Article information: S. Waldner, J. Huwyler and M. Puchkov. Artificial intelligence in the analysis of time-resolved (4D) micro-computed tomography data. Division of Pharmaceutical Technology, University of Basel.