Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.725709
Title: Automatic interpretation of clock drawings for computerised assessment of dementia
Author: Harbi, Zainab
ISNI:       0000 0004 6424 9585
Awarding Body: Cardiff University
Current Institution: Cardiff University
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
The clock drawing test (CDT) is a standard neurological test for detection of cognitive impairment. A computerised version of the test has potential to improve test accessibility and accuracy. CDT sketch interpretation is one of the first stages in the analysis of the computerised test. It produces a set of recognised digits and symbols together with their positions on the clock face. Subsequently, these are used in the test scoring. This is a challenging problem because the average CDT taker has a high likelihood of cognitive impairment, and writing is one of the first functional activities to be affected. Current interpretation systems perform less well on this kind of data due to its unintelligibility. In this thesis, a novel automatic interpretation system for CDT sketch is proposed and developed. The proposed interpretation system and all the related algorithms developed in this thesis are evaluated using a CDT data set collected for this study. This data consist of two sets, the first set consisting of 65 drawings made by healthy people, and the second consisting of 100 drawings reproduced from drawings of dementia patients. This thesis has four main contributions. The first is a conceptual model of the proposed CDT sketch interpretation system based on integrating prior knowledge of the expected CDT sketch structure and human reasoning into the drawing interpretation system. The second is a novel CDT sketch segmentation algorithm based on supervised machine learning and a new set of temporal and spatial features automatically extracted from the CDT data. The evaluation of the proposed method shows that it outperforms the current state-of-the-art method for CDT drawing segmentation. The third contribution is a new v handwritten digit recognition algorithm based on a set of static and dynamic features extracted from handwritten data. The algorithm combines two classifiers, fuzzy k-nearest neighbour’s classifier with a Convolutional Neural Network (CNN), which take advantage both of static and dynamic data representation. The proposed digit recognition algorithm is shown to outperform each classifier individually in terms of recognition accuracy. The final contribution of this study is the probabilistic Situational Bayesian Network (SBN), which is a new hierarchical probabilistic model for addressing the problem of fusing diverse data sources, such as CDT sketches created by healthy volunteers and dementia patients, in a probabilistic Bayesian network. The evaluation of the proposed SBN-based CDT sketch interpretation system on CDT data shows highly promising results, with 100% recognition accuracy for heathy CDT drawings and 97.15% for dementia data. To conclude, the proposed automatic CDT sketch interpretation system shows high accuracy in terms of recognising different sketch objects and thus paves the way for further research in dementia and clinical computer-assisted diagnosis of dementia.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.725709  DOI: Not available
Share: