LNCS Proceedings

Mark Jenkinson

Keynote Speaker

Ben Glocker

Keynote Speaker

Best Paper Award - Giorgos Sfikas

"Bayesian Multiview Manifold Learning applied to Hippocampus Shape and Clinical Score Data - Giorgos Sfikas, Christophoros Nikou"

Sponsored by:


Our goal is to highlight the potential of using Bayesian or random field graphical models for advancing scientific research in biomedical image analysis. The technical program will consist of two keynote talks, depending on speaker availability, from prominent figures in the community, and the presentation of previously unpublished and contributive papers.

This workshop builds on the success of the 1st and 2nd BAMBI workshops. Initially introduced at MICCAI 2014, the 1st BAMBI included excellent talks by Koen Van Leemput, Mike Miller, and Ramin Zabih. A year later during MICCAI 2015 Munich, the 2nd BAMBI included thought provoking talks from John Asburner and Nikos Paragios.

We are looking for original, innovative and mathematically rigorous models and inference schemes for the analysis of medical image data. Emphasis will be placed on novel methodological approaches and interpretability with respect to the include plausible and realistic generative models, efficient inference strategies, model comparison and averaging, model uncertainty, modelling of multi- modal data, hierarchical graphical modelling and comparison to traditional/heuristic methods.

Probabilistic graphical models provide a framework for describing observed data in a theoretically principled manner, while the construction of and inference on appropriate models requires careful decisions, which necessitate further investigation. Potential applications cover the full scope of medical image analysis: segmentation, registration, classification, fusion, reconstruction, atlas construction, tractography, structural/functional modelling, and population analysis.

The main objectives of this workshop differ from some other related workshops, e.g. machine learning in medical imaging, by having a stronger mathematical focus on the foundations of probabilistic data analysis, as opposed to generic machine learning applications. Furthermore, this forum facilitate the presentation and detailed discussion of novel and speculative works, which may be outside the scope of the main conference, but are essential for the advancement of modelling and analysis of medical imaging data.

Why probabilistic models?

Probabilistic models provide a coherent framework to describe observed imaging and medical data, and as such, have an increasingly important role to play in biomedical image analysis. These approaches have recently achieved widespread popularity in machine learning, computer vision and medical imaging, partly due to modern computationally efficient inference schemes.Fundamentally, describing a model as a probabilistic graph allows the dependencies between parameters to be explicitly represented, allowing flexible, and interpretable descriptions of the relationships and constraints of the model. This elicits several further advantages: the inference of posterior distributions, multi-layered hierarchical models and objective comparison between models. Despite the demonstrable power of these methods, there are still many significant questions to consider regarding their effective exploitation for analysing biomedical images. This workshop will provide a discussion platform for the use of probabilistic modelling approaches in this context, with a particular focus on methodological innovation and model interpretability.

Important Information

Paper submission date: 24th June 2016 (extended)
Workshop Date: 21st October 2016
Workshop Location: Intercontinental Athenaeum, Theta

Why Submit?

Why should one submit to BAMBI:

- To disseminate new probabilistic models and inference schemes to the most relevant members of the MICCAI community, without requiring comprehensive validation.

- To have your work published as part of the a Springer LNCS volume.

- To have the chance of winning the €300 best paper award, sponsored by IcoMetrix.