Rory Stuart Clark

Radiologists are specialist doctors who examine visual medical information such as x-rays and CT scans to produce a diagnostic opinion for the treatment of a patient. Their work is extensively reliant on modern digital systems to generate scans, transfer and manipulate them for analysis, and communicate the findings to the relevant clinician. This means that radiology is often at the forefront of medical digitisation and implementation. There is, however, still a lack of holistic examination of precisely how – from an HCI standpoint -the radiologist interacts with the tools and devices at their disposal when constructing diagnostic reports and communicating them to others. This presents difficulties for the design, evaluation and implementation of new and novel capabilities. This thesis examines the space in which a radiologist works from an ethnomethodological standpoint. We conduct studies in the form of ethnomethodologically informed ethnographies that examined the radiologist’s interactions with digital and physical systems, allowing the reconstruction, analysis and critique of the design of the devices in-use from a holistic standpoint, with particular emphasis on the interoperability and usability from an outsider’s perspective. These ethnographies utilise interactive and non-invasive observational techniques, allowing observers to gain an unbiased perspective on the interactional process without being ignorant to the medical and systemic influences in place. Supplementary to our ethnographies, we conducted a pair of focus groups with clinicians before and after examining the user’s perspective on the issues they face and how they believe they can be addressed, thus allowing us to situate ourselves in the context and environment that we would be investigating and gather feedback on our findings and conclusions throughout. This work provides a much-needed reconstruction and presentation of the ways in which radiologists interact with and perceive digital devices when constructing and communicating their diagnostic opinion. We contribute an ethnomethodological analysis of the reporting process to allow for digital designers and clinical leaders to properly account for the real-world aspects of interaction that affect design and implementation of devices. Through collaborative and participant-driven methods, we also contribute codified and contextualised end-user clinical perspectives on how existing devices can be improved upon and how the design process can incorporate their domain expertise. Finally, we offer a reflective methodological contribution through exploration and analysis of the strategies when conducting research in a healthcare space as an Early Career Researcher in HCI, allowing researchers to improve the quality of future research in this domain.

This paper presents an exploration of how UK radiologists interact with VUI systems and peripherals when authoring diagnostic reports. We conducted a laboratory study with 10 practicing clinical radiologists to investigate the ways in which radiologists utilise speech-based technology to construct, edit and proof their work by having them report on real-world anonymised medical studies on camera. A sample of the participants also participated in interviews in which their data was collaboratively analysed and examined to offer deeper insight into the realism and generalisability of our findings and conclusions. We conclude that better training should be given to radiologists on how VUI systems work, and further investigation should be carried out on the best ways to interact with Speech To Text systems in safety critical environments.

This paper investigates patterns of communication and interaction in radiology departments by utilising ethnographic and ethnomethodological techniques. We conducted 12 sessions of observations with accompanying interviews in the practitioners natural working environment, and used this to construct a model of how a report is authored from an outsider perspective. These observations revealed that practitioners have a distinct way of interacting with speech recognition systems that is unlike traditional interaction with Voice User Interfaces, that there are a myriad of ways of communicating a result to peers that are dependent on context and tacit knowledge, and that reports are rarely completed without some form of cognitive interruption in the authoring stage. We finish by recommending a codified framework for guidance on communicating results that builds on evidence-based findings tailored specifically to radiology practitioners.