Detection of Foreign Particles in Medical Solution Using Modified Affinity Propagation

A new detection method based on modified affinity propagation is proposed for detecting foreign particles in medical solution. The captured sequential images are used to generate moving trajectories of possible foreign particles. The external disturbances and background are almost removed by using a series of simple and effective image processing. The extracted potential targets, including foreign particles and possible residual noises, are partitioned into a number of valid clusters by modified affinity propagation clustering, in which the new similarity measure can better reflect the intrinsic link of linear-structured datasets and the temporal constraint enforces the temporal continuity of data points. The foreign particles can be easily recognized by analyzing the cluster structure according to the continuity and smoothness of moving targets trajectory. This algorithm combines modified affinity propagation method with the moving characteristics of different particles, and can achieve accurate partitions efficiently even though the clusters are badly adjacent. The experimental results indicate that the proposed algorithm can detect foreign particles effectively with high detection speed and accuracy.