Diffeomorphic Multi-Resolution Deep Learning Registration for Applications in Breast MRI

In breast surgical planning, accurate registration of MR images across patient positions has the potential to improve the localisation of tumours during breast cancer treatment. While learning-based registration methods have recently become the state-of-the-art approach for most medical image registration tasks, these methods have yet to make inroads into breast image registration due to certain difficulties-the lack of rich texture information in breast MR images and the need for the deformations to be diffeomophic. In this work, we propose learning strategies for breast MR image registration that are amenable to diffeomorphic constraints, together with early experimental results from in-silico and in-vivo experiments. One key contribution of this work is a registration network which produces superior registration outcomes for breast images in addition to providing diffeomorphic guarantees.

Noise-specific denoising method with applications to high-frequency ultrasonic images

Denoising is of utmost importance for the visualization and processing of images featuring low signal-to-noise ratio. Total variation methods are among the most popular techniques to perform this task improving the signal-to-noise ratio while preserving coherent intensity discontinuities. In this work, a novel method, termed maximum likelihood data, is proposed, endowing the total variation formulation with the capability to deal with noise-specific models and pre-processing stages for a certain image of interest. To do this, the data fidelity term is modified by means of a maximum likelihood estimator between the original and the denoised image. To assess the improvements of the proposed method with respect to the total variation formulation, we study the denoising of high-frequency ultrasonic images on in-silico and in-vivo setups. The proposed method delivered a better contrast, preservation and localization of the structures while diminishing the intensity bias of the total variation formulation for the multiplicative noise. The enhancement of medical images through denoising helps to improve the outcome of subsequently applied image processing such as registration and segmentation procedures.