Control Of Dissolving Cytostatics Using Computer Image Analysis

In this paper we present the concept and prototype of a visual inspection system for the control of dissolving cytostatics. It is a part of a broader system for the automatic preparation of cytostatics. The idea consists in rotating the vessel, stopping it, and analysing a sequence of images while the particles in the solution are still moving. We proposed 30 descriptors of the variability of statistical properties of image noise. We used PCA for the reduction of space dimensionality, and cross-validation to optimise the parameters of the method and to assess its performance. Our approach allows us to detect undissolved powder, even if the size of its particles is comparable with pixel size.