Automated classification reveals morphological factors associated with dementia

Authors:
David Cornforth;Herbert Jelinek
Affiliations:
University of New South Wales, Australian Defence Force Academy, Canberra, ACT 2600, Australia;Charles Sturt University, PO Box 789, Albury, NSW 2640, Australia
Venue:
Applied Soft Computing
Year:
2008

Citing 5
Cited 4

Induction of Decision Trees

Machine Learning
The Power of Decision Tables

ECML '95 Proceedings of the 8th European Conference on Machine Learning
A Comparison of Machine Learning Approaches for the Automated Classification of Dementia

AI '02 Proceedings of the 15th Australian Joint Conference on Artificial Intelligence: Advances in Artificial Intelligence
Data Mining

Data Mining
Automated classification of dementia subtypes from post-mortem cortex images

AI'05 Proceedings of the 18th Australian Joint conference on Advances in Artificial Intelligence

Detection of CAN by ensemble classifiers based on ripple down rules

PKAW'12 Proceedings of the 12th Pacific Rim conference on Knowledge Management and Acquisition for Intelligent Systems
An approach for Ewing test selection to support the clinical assessment of cardiac autonomic neuropathy

Artificial Intelligence in Medicine
Improving classifications for cardiac autonomic neuropathy using multi-level ensemble classifiers and feature selection based on random forest

AusDM '12 Proceedings of the Tenth Australasian Data Mining Conference - Volume 134
Visualisation in biomedicine as a means of data evaluation

Journal of Visualization

Quantified Score

Hi-index	0.00

Visualization

Abstract

Dementia is believed to be associated with changes in the physical structure of brain tissue, particularly in the pattern of small blood vessels. This study investigates one of the current research questions in the understanding of dementia, that is, whether there are differentiating factors in the structure of blood vessels of the cortex associated with different dementia subtypes and controls. Our approach is to use automated classification techniques to build predictive models based on fractal and non-fractal morphological descriptors, in order to label images of post mortem brain tissue with the appropriate pathology. Our goal is not to provide automated diagnosis, but to confirm or deny the presence of a relationship between morphological features and disease. The use of a variety of machine learning methods allows the exploration of the complex relationships that may exist. This study also addresses the choice of suitable features and the role of fractal analysis in medical image processing. The results suggest that there are differentiating factors, but these are difficult to detect, and vary between different areas of the cortex. Features derived from multi-fractal analysis showed more promise in this application than the non-fractal features we studied.