Soheil Hor wins prestigious “MICCAI Young Scientist Award” for Best paper at MICCAI 2015 with Mehdi Moradi

Soheil Hor and Prof. Mehdi Moradi, recently proposed a fundamentally new paradigm for handling the missing data problem in machine learning based on the concept of random forest classifiers. Their contribution to the field was recognized at the Medical Image Computing and Computer Assisted Interventions (MICCAI) conference. Soheil, an ECE MSc student, received the Young Scientist Award for the paper co-authored with his supervisor, Dr. Moradi, describing the new method of treating missing data titled, “Scandent tree: a random forest learning method for incomplete multimodal datasets”.

Dr. Moradi is currently a Research Staff Member at IBM Research, Almaden Research Center in San Jose, California and Soheil works with the Engineers in Scrubs master program at UBC.

The new method called “scandent tree” enables the use of new clinical data modalities in machine learning algorithms before one accumulates a lot of data from this new modality. It gives the researchers the ability to use a small dataset to train a supervised random forest classifier, and then enrich and enhance that classifier using data from older modalities often available in imaging/data archives. Soheil used a benchmark cardiac disease dataset to show that scandent tree is far superior to the conventional methods available in machine learning literature which often rely on estimating the values of missing data points. These methods fail when the portion of the data missing the new modality is very large. Scandent tree is capable of handling this scenario.

Medical Image Computing and Computer Assisted Interventions is the highest impact international conference that covers advanced analytics algorithms for medical images and data. The Young Scientist Award is the most prestigious publication related award of the field and goes to 4-5 early career scientists with a paper of the highest quality.

Many UBC professors and students participated in MICCAI this year, presenting seven papers at the conference and Prof. Rob Rohling was recognized with a Best Reviewer Award by conference organizers.

Photo: Prof. Tim Salcudean, Prof. Purang Abolmaesumi, Saman Nouranian, Soheil Hor, Prof. Mehdi Moradi, Shekoofeh Azizi, Siavash Khallaghi and Amin Suzani 

Find out more:

Scandent tree: a random forest learning method for incomplete multimodal datasets,  by Soheil Hor, Mehdi Moradi.

Engineers in Scrubs

Medical Image Computing and Computer Assisted Interventions