Scalable Learning: Probabilistic Graphical Models for Image Analysis

Fig. 1: Segmentation evaluation of Ceramic Matrix Composite microCT data

Scientific Achievement

  • Advancement of automated data analysis coming from experimental data facilities
  • MSMcam framework analyzes data and generates easily interpretable segmentation results using appropriate set of evaluation metrics for each dataset [1].

Significance and Impact

  • Unique approach to formulating metrics to determine the quality of segmentation coming out from various techniques
  • Framework tested on several scientific datasets

Research Results

  • Use a suite of segmentation algorithms simultaneously on the same data and observe results agreement using various quality metrics
  • Automatic approach towards overcoming the fact that imaging experiments vary significantly from one experiment to another
  • Supports the analysis and comparison of image segmentation results through suitable metrics even when a ground-truth is not available



  1. T. Perciano, D. Ushizima, H. Krishnan, D. Parkinson, N. Larson, D. M. Pelt, W. Bethel, F. Zok, and J. Sethian, “Insight into 3D micro-CT data: exploring segmentation algorithms through performance metrics,” Journal of Synchrotron Radiation, vol. 24, no. 5, pp. 1065–1077, Sep. 2017.