Analysing morphological patterns of blood vessels for detection of Alzheimer's disease
The physiological consequences of Alzheimer's disease (AD) concern the development of amyloid plaques and neurofibrillary tangles. Development of amyloid plaques in the brain is caused by Amyloid Beta that forms part of an amyloid precursor protein. In a normal brain, these protein fragments ar...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Conference Proceeding |
| Language: | eng |
| Subjects: | |
| Online Access: | Get full text |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The physiological consequences of Alzheimer's disease (AD) concern the development of amyloid plaques and neurofibrillary tangles. Development of amyloid plaques in the brain is caused by Amyloid Beta that forms part of an amyloid precursor protein. In a normal brain, these protein fragments are broken down and eliminated but with AD, these fragments accumulate to form hard insoluble plaques. Our techniques are based on the image analysis of brain tissue and study the branching structures of the blood vessels (which is novel itself), on the analysis of tortuosity and density. These are known to have links with the onset of AD. The branching structures are detected by evidence gathering approaches and described by their structure. This allows recognition to be achieved: the structure of those samples derived from patients with AD differs from that for normal subjects. This also occurs for the tortuosity and to a lesser extent the density. The descriptions can be classified using machine learning techniques, as such achieving an automated process from image to recognition. We analyse the structure of the blood vessels in a database of images collected from control, age-matched and patients with severe AD. The database comprises five subjects of each of the three types, imaged in controlled conditions and from the Brain Tissue Resource of Newcastle UK. We show that by automated image analysis we now appear able to discriminate between brain tissue samples from patients presenting AD and from the normal samples. The branching structure is the description that is most suited to classification purposes. On this initial dataset we can achieve 100% correct classification from a combination of these descriptions and around 90% correct classification from the branches and their paths. We are thus confident in the correct referral of patients for further investigation when this new technique is translated for clinical use. |
|---|---|
| ISSN: | 1082-3654 2577-0829 |