Combining digital histology with magnetic resonance imaging in amyotrophic lateral sclerosis

Thesis

There is an increasing demand for non-invasive biomarkers that can reliably indicate the presence and extent of neurodegeneration in its early stages. Various magnetic resonance imaging (MRI) methods have shown great potential to indicate the event of upper motor neuron involvement in the progression of amyotrophic lateral sclerosis (ALS) at the population level, but almost none of these have been systematically validated against traditional histological markers at the individual level. Validation relies on a precise and preferably automatic method to align a large number of MRI and histology images of the same tissue, which poses unique challenges compared to more conventional MRI-to-MRI registration.

The thesis summarises the development and evaluation of a novel image registration framework (Tensor Image Registration Library) that is designed to achieve automated registration between sparsely sampled histology sections and whole-brain post-mortem MR volumes derived from an ALS study cohort. A multistage automated pipeline is developed, in which two types of histology sections are registered to an MRI volume via one or two photographic intermediates. The research chapters explain the rationale behind each stage, describe the algorithm, and report the experiments that were carried out to verify the accuracy and robustness of the implementation.

The proposed method uniquely integrates part-to-whole matching and 2D-to-3D registration, allowing conventional small histology samples to be registered without serial sectioning. Furthermore, the orthogonal deformations of a slice are estimated to account for the imperfections of a free-hand cut, which are shown to be as large as several millimetres. The pipeline achieves sub-millimetre accuracy in each stage and equally accurate alignment between histology and MRI when the stages are combined. This registration approach alleviates the need for specialist cutting or stain automation hardware, and is therefore suitable for integration into a wide range of studies that utilise routine neuropathology practices.

Authors: Huszar, IN

• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: biomedical imaging; multimodal; histology; MRI; software; tensor image registration library; amyotrophic lateral sclerosis; post-mortem; neuroscience; image registration; magnetic resonance imaging; TIRL
• Subjects: Neurosciences; Software; Biomedical engineering; Magnetic resonance imaging; Imaging