Loughborough University
Leicestershire, UK
LE11 3TU
+44 (0)1509 263171
Loughborough University

Loughborough University Institutional Repository

Please use this identifier to cite or link to this item: https://dspace.lboro.ac.uk/2134/8492

Title: Intelligent computing applications based on eye gaze: their role in mammographic interpretation training
Authors: Chen, Yan
Keywords: Mammography
Visual search
Eye tracking
Human-Computer Interaction
Performance evaluation
Breast screening
Breast screening in China
Issue Date: 2011
Publisher: © Yan Chen
Abstract: Early breast cancer in women is best identified through high quality mammographic screening. This is achieved by well trained health professionals and appropriate imaging. Traditionally this has used X-ray film but is rapidly changing to utilise digital imaging with the resultant mammograms visually examined on high resolution clinical workstations. These digital images can also be viewed on a range of display devices, such as standard computer monitors or PDAs. In this thesis the potential of using such non-clinical workstation display devices for training purposes in breast screening has been investigated. The research introduces and reviews breast screening both in the UK and internationally where it concentrates upon China which is beginning screening. Various imaging technologies used to examine the breast are described, concentrating upon the move from using X-ray film to digital mammograms. Training in screening in the UK is detailed and it is argued that there is a need to extend this. Initially, a national survey of all UK mammography screeners within the National Health Breast Screening Programme (NHSBSP) was undertaken. This highlighted the current main difficulties of mammographic (film) interpretation training being tied to the device for inspecting these images. The screeners perceived the need for future digital imaging training that could be outside the breast screening centre; namely 3W training (Whatever training required, Whenever and Wherever). This is largely because the clinical workstations would logistically not be available for training purposes due to the daily screening demand. Whilst these workstations must be used for screening and diagnostic purposes to allow visualisation of very small detail in the images, it is argued here that training to identify such features can be undertaken on other devices where there is not the time constraints that exist during breast screening. A series of small pilot studies were then undertaken, trialling experienced radiologists with potential displays (PDAs and laptops) for mammographic image examination. These studies demonstrated that even on a PDA small mammographic features could be identified, albeit with difficulty, even with a very limited HCI manipulation tool. For training purposes the laptop, studied here with no HCI tool, was supported. Such promising results of display acceptability led to an investigation of mammographic inspection on displays of various sizes and resolutions. This study employed radiography students, potentially eventual screeners, who were eye tracked as they examined images on various sized displays. This showed that it could be possible to use a small PDA to deliver training. A detailed study then investigated whether aspects of an expert radiologist s visual inspection behaviour could be used to develop various training approaches. Four approaches were developed and examined using naïve observers who were eye tracked as they were trained and tested. The approaches were found to be all feasible to implement but of variable usefulness for delivering mammographic interpretation training; this was confirmed by opinions from a focus group of screeners. On the basis of the previous studies, over a period of eight months, a large scale study involving 15 film readers from major breast screening centres was conducted where they examined series of digital mammograms on a clinical workstation, monitor and an iPhone. Overall results on individuals performance, image manipulation behaviour and visual search data indicated that a standard monitor could be employed successfully as an alternative for the digital workstation to deliver on-demand mammographic interpretation training using the full mammographic case images. The small iPhone, elicited poor performance, and was therefore judged not suitable for delivering training with the software employed here. However, future software developments may well overcome its shortcomings. The potential to implement training in China was examined by studying the current skill level of some practicing radiologists and an examination of how they responded to the developed training approaches. Results suggest that such an approach would be also applicable in other countries with different levels of screening skills. On-going further work is also discussed: the improvement of performance evaluation in mammography; new visual research on other breast imaging modalities and using visual search with computer aided detection to assist mammographic interpretation training.
Description: A Doctoral Thesis. Submitted in partial fulfillment of the requirements for the award of Doctor of Philosophy of Loughborough University.
URI: https://dspace.lboro.ac.uk/2134/8492
Appears in Collections:PhD Theses (Computer Science)

Files associated with this item:

File SizeFormat
PhDThesis-YanChen-2011.pdf53.41 MBAdobe PDFView/Open


SFX Query

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.