Deep learning for medical image analysis

Abstract:
Deep learning methods are a class of algorithms that learn hierarchical abstract representation of data using a complex model. A typical model is made up of multiple hidden layers of neuronal units which are based on computational model of biological neurons. These algorithms are just beginning to be explored in medical image analysis, e.g. classifying medical image pixels into lesion and normal tissue. Classification carried out on the abstract representation is expected to provide higher accuracy when compared to classification carried out on raw image pixels. 
An overview of recent developments that lead to efficient implementation of deep neural networks will be presented followed by applications of deep learning to Multiple Sclerosis segmentation and Brain Tumor segmentation.

References:
1] Yann LeCun, Yoshua Bengio & Geoffrey Hinton, Nature 521, 436–444 (28 May 2015)
2] G. Krishnamurthi et al, First Prize, Longitudinal MS challenge, IEEE International Symposium on Biomedical Imaging, New York (2015)