User Centered Design for Medical Visualization

Medical Visualization has attracted a lot of attention in the last two decades. It provides valuable assistance to the practice of medical professionals, including diagnosis, treatment planning, training and revision. Due to its strongly practical oriented nature, a considerable number of examples have emerged which successfully apply medical visualization to support clinical practice, such as HipOp Project (Testi et al, 2004, Viceconti et al, 2004), Collaborators in Radiological Interventional Virtual Environments (Vidal et al 2004). Blood Flow Visualization (Pivkin et al 2005), etc. And the research interest in this area shows no sign of abating.

Medical Visualization is an interdisciplinary area which involves a large number of diverse issues across computer and medical sciences. The primary goal of Medical Visualization is to provide high quality and fidelity display of human anatomy and structure through the rendering of medical data captured by modern equipments, such as CT, MRI, PET and Ultrasound etc. Recently, great attention has been paid towards the visualization of complex 3D structures involved in human anatomy. This can be done either by extracting high quality surface models from 3D medical data, which is normally in volumetric form (Bertram, 2004, Livnat, 2004), or by carrying out direct rendering on the volumetric data (Schulze & Rice, 2004, Mroz et al, 2000).

Surface based visualization requires high quality extraction of surface model from volumetric dataset. It is able to generate surface models that allow numerous manipulation with clinical meanings, such as shape measurement and analysis, model cutting for virtual surgery, etc. However a critical challenge here is how to obtain surface model with sufficient medical accuracy. This issue also relates to image segmentation, which allows us to obtain region of interest from 2D and 3D images.

While the surface based visualization is still a quite important approach in medical visualization, it is generally recognized that volumetric rendering allows us to see a great deal of fine details within medical data and also significantly improves the quality of 3D medical imaging(Dong et al, 2001). One of the most critical issues in volumetric rendering is to allow users to easily choose the most relevant information to visualize by constructing an adequate and interactive transfer function, (Kniss et al, 2002).

However, just providing quality images is not the ultimate goal of using computer to assist medical practice. To be clinically meaningful, useful information needs to be identified by carrying out analysis towards medical images. To this end, great efforts have been made in Medical Image Analysis, which involves a large number of activities including lesion detection(Ye et al, 2007, Wei et al, 2007, Sluimer et al, 2006), medical image registration(Krueger et al, 2007, Napel et al, 2004, Hajnal et al, 2001), segmentation(Aldasoro& Bhalerao, 2007, Noble & Boukerroui, 2007, Manay & Yezzi, 2003) and classification(Petersen et al, 2002, Baese, 2003, Zhang, 2000), organ shape measurement , analysis and motion tracking, in order to seek clinical useful information. Many of these practices require significant prior knowledge about human anatomy. Hence, the techniques and knowledge from human anatomy, image processing and analysis jointly constitute vital input for Medical Visualization.

Currently, vast majority of existing medical visualization techniques polarize their efforts on data modeling and rendering. Their target is to generate high quality images with a great deal of fine structure details, such as blood vessels, muscle fibers, etc. Such images convey significant visual information for clinicians. However, many of these works have put too much emphasis on producing decent images, while neglecting the human factors involved during a visualization process.

Previous experiences and examples have strongly suggested that simply using graphic techniques to display medical data may not provide adequate support for clinicians. As a new research trend, the interest in human factors within the medical visualization research community has been increasing over the last few years. Instead of simply generating high quality photo-realistic images, researchers have started to pay more attention to the human perspective. In fact, there is a growing body of evidence suggesting the strong need to study human factors as a basis for Medical Visualization design. Moreover, human’s perspective, thinking and interaction with images significantly affect their understanding of the information presented visually. As the ultimate goal of Medical Visualization is to accurately deliver clinical information for medical professionals, an effective visualization process should include users as an integral part of the course of action.

Therefore, this book attempts to reflect such a trend by collecting a number of different research works and surveys in medical visualization with particular emphasis on User-Centered Design (UCD). In general terms, UCD is an approach to provide rationale and justification to the design of a product from the perspective of people who will use it. There is an international standard for UCD methodologies(ISO13407, 1999), which outlines the general process of a development life cycle focusing on user-centered activities. This cycles includes: Specifying the context of use, which identifies potential users of the product; Specifying requirements, which defines and documents detailed user requirements; Creating design solutions, which covers from initial concept to complete design for the product; and Evaluating designs, which provides product assessment through usability testing.

A good practice of USD for Medical Visualization requires careful design from the perspective of users in order to meet their business objectives. Such a design has to be user-centered and task oriented. Whether a medical visualization system or technique is intended for surgeons, or for medical students should make significant difference with respect to the design. As a medical visualization system clearly involves knowledge from both computer and medical science, which are two clearly distinguished disciplines, making user interface as consistent as possible can help users in their learning and minimize the time required for their training. Unavoidably, a medical visualization system may involve computing languages & instructions & terminologies in its interface, therefore, for the purpose of facilitating users, it is important to keep the dialogue between the users and system in a natural sequence. No redundant information should be presented to the users, as irrelevant information adds unnecessary complexity. In particular, terminologies need to be defined clearly and their meaning should remain consistent. This helps to reduce unnecessary mental effort from the users and consequently allows them to concentrate on their tasks. Many human computer interaction research have shown that complicated interaction task can significantly frustrate users and distract them from their real task, which potentially lead to low efficiency and more errors. Also, providing adequate feedback to the user interaction is of great importance. Due to the limitation of computing facilities and increasing large scale of dataset, many user-wanted jobs, e.g., rendering a large medical dataset, require considerable time to fulfill. In this case, to make the users confident on the actions that they have taken, proper feedback from the system, such as progress bar, need to be given to indicate the time remaining for the job. Such a feedback should be given at a proper level, as irrelevant diagnostic or status information can give rise to unnecessary confusion. Also, since a modern medical visualization system often involves a large number of functionality, providing adequate navigation mechanisms allows the users to familiarize themselves with the system. A proper navigation mechanism concerns many issues on computer interface, such as giving proper and consist window titles; offering location indicators. Other mechanisms such as providing a navigation map should be considered. In between different view windows, such as the rendering window to present the 3D picture of a dataset, and the statistics window to illustrate the data distribution, clear and easy routes need to be defined to allow the users to switch between them for their specific tasks. In addition, providing sufficient flexibility and tolerance is of great user interests. Often users choose system functions by mistake, for example, start to render a wrongly-selected large dataset, which involves large computing resources. They should be given clearly marked exit to leave the unwanted state, e.g. immediate canceling of the dataset rendering.

Apart from the multidisciplinary issues already involved, an user centered medical visualization further concerns a wide range of other issues in Human Computer Interaction and Interface, Design, Usability, etc. This book endeavors to explore some of these issues by looking into the UCD in medical visualization from a number of selected perspectives.