Search

The success of modern medical imaging systems has created a data overload problem, where an ever-increasing number of examinations, generate more images per study, which all need to be evaluated by radiologists or other reporting practitioners. This operational bottleneck hasthe potentialto create fatigue and burnout due to the high mental workload that is required to keep up with the demand. The focus of this problem centres around the cognitive complexity of the radiology reporting workflow, and the associated workstation interactions involved in diagnostic report generation. There has been a significant body of work evaluating the behaviour of radiologists using controlled laboratory-based techniques, but these non-naturalistic studies fail to address the highly context dependant nature of the radiology reporting workflow. For example, the early eye-tracking work of Charmody et al; the psychometric studies by Krupinksi et al; and also the workstation interaction evaluations of Moise et al; whilst highly principled, can be all be questioned on the grounds of ecological validity and authenticity. This thesis asserts that the only way to truly understand and resolve the radiology data overload problem, is by developing a situated method for observing the reporting workflow that can evaluate the behaviours of the reporting clinicians in relation to their authentic reporting context. To this end, this study has set out to develop a new approach for observing and analysing the cognitive activities of the reporters relative to the demands of their genuine working environment, and supported through the application of a Critical Realist's perspective to naturalistic workplace observations. This goal was achieved through the development of four key project deliverables: • An in-depth exploratory study of the radiology overload problem based on an extensive literature review and situated observations of authentic reporting workflows. • A descriptive hierarchical activity modelof the reporting workflow that can be understood by both clinicians, application designers and researchers. • A generalised methodology and research protocolfor conducting situated observations of the radiology reporting workflow, using an analysis based on the process tracing of sequencesof Object Related Actions, captured with eye-tracking and multimodal recordings. • A set of case studies demonstrating the applicability of the research protocol involving 5 Radiology Consultants, 2 Radiology Registrars and one Reporting Radiographer at a single NHS Hospital within the UK. The final workflow evaluation of the case studies demonstrated that activities such as error correction, and the collection of supporting radiological information from previous studies is complex, time consuming and cognitively demanding. These types of activities are characterised by long, low utility actions that correspond to what Kahneman refers to as "Thinking Slow". Also, the participants appeared to be self-optimising their workflow via a sparse use of complex functionality and system tools. From these observations, the author recommends that any intervention that can reduce the number and the duration of the object related actions used to produce radiology reports, will reduce cognitive load, increase overall efficiency, and go some way to alleviate the data overload problem. 4 This study establishes a new set of situated techniques that are able to capture and quantify the complex dynamicactivities that make up the radiology reporting workflow. Itis hoped that the ability to distil usefuland impactful insightsfrom the user's workstation behaviours can be used as the basis for further development in the area of workflow analysis and redesign, which will ultimately improve the working lives of Radiologists and other Reporting Clinicians. Lastly, the generic nature of these techniques make them amenable for use within any type of complex sociotechnical human factors study related to the cognitive efficiency of the user.