INTRODUCTION
In recent years, as a result of computerization and technological advances, we have witnessed an exponential increase in electronic databases in diffe- rent areas of society. In health care field it is not different, the increasing digitization of health records (clinical histories, imaging, genetics, laboratory results, among others) has exponentially expanded the medical information available electronically (West, 2013). This escalate is based on the explicit benefits of electronic health records (EHR) compared to paper records such as cost, readability, accessibility, storage, and security (Jin, 2016). Despite the well-known advantage of EHR, some recent studies have shown that EHR- based systems hardly improve the ability of healthcare providers to make better decisions (Abdullah, 2020). Their volume, heterogeneity and comple- xity may overload the user, leading to interpretation errors or overlook vital data (Abdullah, 2020) (Aigner, 2012) (Ledesma, 2019) (Martignene, 2020).
Therefore, currently, one of the main focuses of research in healthcare is related with processing and communicating large amounts of information in
© 2022. Published by AHFE Open Access. All rights reserved. 252
Radial Visualization Model in Health Care: A Survey 253
an interactive, understandable, and adequate manner to the needs of each user (Filipov, 2021). With better clinical decision-making tools, that health professionals may use, the data becomes more amenable to visualization. The refinement of the models in association with biomedical illustrators, and a more intuitive and dynamic interface can assist clinicians to treat patients instead of spending hours reading computer screens. Nonetheless, the deve- lopment of data visualization in healthcare is in its beginning (Boyd, 2017). One of the first paper on this area was published less then 3 decades ago (Plaisant, 1996). Since then, several interfaces were developed, such as Life- Lines (Plaisant, 1996), KNAVE-II (Goren-Bar, 2004), TimeLine (Bui, 2007) HARVEST (Hirsch, 2015), OutFlow (Wongsuphasawat, 2012), Care Path- way Explorer (Perer, 2015), EventFlow (Monroe, 2013) offering to the user a better comprehensive view, making it easy to detect temporal associations. Most of these interfaces used a linear model to present their data, however, over the last decade we have witnessed the development of tools for analyzing patient clinical data that replace more traditional linear graphics with radial models (Jin, 2016) (Joshi, 2012) (Zhang, 2013) (Hu, 2014) (Baytas, 2016) (Ledesma, 2016) (Senathirajah, 2017) (Kaushal, 2017) (Ha, 2019) (Abdullah, 2020).
Radial Models Importance for Data Visualization One of the most effective ways of communicating information based on large datasets is through graphical visualization. As there are different graphic models to represent data, various communication contexts require distinct types of graphics to deliver an informative and meaningful visualization for a given problem.
Among the different graphic models there is the radial model. Draper et al. (2009) refers that radial visualization is not a panacea for all information visualization problems. However, there are many situations where this type of graphics is the most appropriate way to promote the data relationship and effectively communicate the information. Radial models come in diffe- rent shapes such as circular, elliptical, or spiral form (Diehl, 2010), typically using polar coordinates instead of Cartesian coordinates, used in linear meth- ods. There are situations where radial models present the data in a more comprehensive form and therefore convey important messages that in other type of graphics wouldn’t be evident or not properly emphasized. Indeed, although linear methods are generally considered easier to read, more accu- rate and faster, there are some upsides in using radial models, namely their more natural shapes, more appealing aesthetics. This aspect is also conside- red important in certain applications since they may drive to a more positive attitude, improve the engagement, the availability to learn and explore, and better explaining historical narratives (Filipov, 2021). Time-related data is an example of a dataset that will have a good representation using a radial model. As Aigner (Aigner, 2008) stated, there are many natural processes that are cyclic/periodic and therefore will have a better visualization with radial models. Other feature where radial models are good is in addressing the display fragmentation issue. This feature mitigates the need to access different
displays to see the information, avoiding the need to retain information in memory while other information is researched (Senathirajah, 2017) (Senath- irajah, 2020). Radial models are more visualization-space efficient and easing the comprehension and interaction of the user, with the capability of provi- ding valuable insights from the data (Maças, 2018). Another aspect of this type of model that can be very powerful, when used for the right purpose, is the fact that can be easier to memorize. For instance, when the most impor- tant dimension is represented as sectors it enhances the memorization (Diehl, 2010). Radial models are particularly interesting for focusing on a particu- lar dimension (Maças, 2018), or when depicting two dimensions that are not equally important (Diehl, 2010), or even to represent periodical patterns (Maças, 2018). When there are several categories related to the same object that have a dimension in common, you can quickly apprehend information from a radial graphic, whereas in a linear graphical you would have to visit several displays, several times, to come to the same conclusion. This principle can be seen, applied to a set of clinical events (categories) over time (common dimension) for a patient (object) (Bastardo, 2021).
Radial Models Evolution in Health Care Historically the term radial view model was introduced by Hoffman et al. in 1997 (Hoffman, 1997), although radial representations date back to the beginning of the 19th century, in the field of mathematics and statistics, with pie charts. The works of William Playfairs, “The Statistical Breviary”, and Florence Nightingales are considered the first examples of the use of radial methods (Hoffman, 1997).
Drapper et al (Draper, 2009), propose a classification of radial visualizati- ons into three main design patterns: “Polar Plot”, “Space Filling” and “Ring Pattern”. The “Polar Plot” is divided into 2 subpatterns: “Tree” and “Star”. In both the center of the graph has some special meaning. In the “Tree Pat- tern”, the origin of the graph is located at or near the center of the graph. From the center, segments of lines radiate outwards, and these segments may present ramifications. This type of pattern is mainly used for visualization of hierarchical structures, as well as visualization of the relationship between disparate entities. The “Star Pattern” is like the previous, where the origin of the graph is equally in the center, from which linear segments come outwards, however, unlike the “Tree Pattern” these segments do not have ramifications. They are mainly used for ranking search results as well as visualizing the relationship between disparate entities. In the “Space Filling”, also known as “Radial Space Filing” (RSF), as in the “Polar Plot”, the center of the graph also represents the origin, however instead of the information being dispersed in lines from the center to the outside, it is arranged in the form of concentric circles (“Concentric Pattern”), spirals (“Spiral Pattern”) or spatial clusters (“Euler Pattern”). These types of patterns are used especially for visualizing sequential, periodic and time-oriented data. In the “Ring Pattern”, unlike the other radial patterns, the center of the graph is not very relevant. As the name implies, the data is presented in the form of different rings. There are 2 sub-patterns, the “Connected Ring Pattern” and the “Disconnected Ring
Radial Visualization Model in Health Care: A Survey 255
Pattern”. The purpose of both is to find points of relationship between dif- ferent sets of data. In the first, a point is placed on one of the rings and then edges are drawn between common points of different rings. In the second sub- pattern, instead of lines joining common points from different datasets, the relationship is established through colors, shapes, labels, or both. This way the confusion created by the lines in the middle of the graphic of the previous pattern is avoided. The last decade has shown a growing interest in applying this type of model in Medicine field, with the development of several tools for analyzing patients’ clinical data (Jin, 2016) (Joshi, 2012) (Zhang, 2013) (Hu, 2014) (Baytas, 2016) (Ledesma, 2016) (Senathirajah, 2017) (Kaushal, 2017) (Ha, 2019) (Abdullah, 2020). Some of these radial models were developed to analyze specific clinical conditions of patients (Joshi, 2012) (Hu, 2014) (Ha, 2019) (Abdullah, 2020), while others aim to map the general clinical situa- tion of the patient (Jin, 2016) (Zhang, 2013) (Ledesma, 2016) (Senathirajah, 2017) (Kaushal, 2017), or populations (Baytas, 2016). Joshi and Szolovist (Joshi, 2012) developed a radial “starburst” interface to assess the condition of critical patients in Intensive Care Units and provides prognostic previews of patient’s clinical conditions in real-time. To reduce the complexity of data represented by over 100-dimensional space, the model use machine learning to group similar clusters of patients characterized by eight physiological foci (general, lung, cardio, liver, kidney, hematology, acid-base, electrolytes). The radial axes capture organ severities from 1 (being normal) to 8 (being the worst). Different colored lines depict different time points during the patient’s ICU stay. The “Five W” model (Zhang, 2013) is based on the journalistic concept of describing each occurrence based on answering the five basic questions: who, what, where, when and why. In this interface, the who is the patient, the where is the patient’s body, and the when, what, and why is a reasoning chain which can be interactively sorted and brushed. Using the patient’s body as a radial map, the display captures all medical events that have occurred in the past and present, serving as a quick overview for the doctor who is evaluating the patient, always answering the five basic questions.
OmicCircos, a tool to analyze large-scale genomic data collection, can be used to generate high-quality circular plots for visualizing genomic variations, including mutation patterns, copy number variations, expression patterns, and methylation patterns (Hu, 2014). Jin et al (Jin, 2016) present an inte- ractive system to illustrate health information in an intuitive way, centering their tool on a two-dimensional image of the human body to support body- centered data layout. The user can explore the different locations of the human body to see different symptoms. By selecting these symptoms, the user can easily add, enter and update information. Furthermore, to provide a better layout of the health records, a timeline and color coding are integrated to distinguish different components.
Phenotree is a hierarchical, and interactive phenotyping interface, that permits users to participate in the phenotyping process of large-scale EHR cohorts. This visual analytic tool that allows physicians to interactively explore EHR cohorts, and generate, interpret, evaluate, and refine phenoty- pes by building and navigating a phenotype hierarchy. Specifically, given a
cohort or sub-cohort, PHENOTREE employs sparse principal component analysis to identify key clinical features that characterize the population (Baytas, 2016).
Ledesma et al. (Ledesma, 2016) developed hFigures, an open-source library for visualizing a complete, accurate and normalized graphical repre- sentation of health data. The idea is based on the concept of the hGraph, developed earlier by MITRE Corporation (Follett, 2012). The hGraph design consists of a circular space with an area defined by two circumferences. The area represents the minimum and maximum recommended values for a given measurement. The values are distributed in a circular space. A graph is for- med by joining the data points around the circular area. This polygon or graph reveals a pattern, and its shape provides a quick overview of the general situation of all the values and how they deviate from the recommendations. The hGraph shows a static overview of a person’s wellness. Disease and well- ness are processes that change over time. Thus, in addition to a static snapshot such as in hGraph, hfigures provides additional key features, including a com- parison of multiple health measurements over time. It makes an emphasis on multiple graphs, or figures, in order to provide a graphical representation of evolution of the data over time (multiple snapshots of the data at certain points in time).
MedWISER tool, consists in a “Sunburst” type visual system, in which the patient’s clinical data are represented in different rings. These rings are subdivided into the different medical categories of the patient (lab results, clinical notes, images, among others). These categories are subdivided into more specific subcategories as the rings become more exterior. There is also a color code to identify clinical elements that belong to the same category. In addition, one of the main features is that the user can select the categories he wants to view in each case and simultaneously, which avoids the problem of fragmentation, as repeated browsing and viewing other displays are not necessary (Senathirajah, 2017).
Patient Journey Visualizer (PJV) is a tool that helped visualize patient journeys, from sickness to recovery, using Parallel Coordinates, Sankey, and Sunburst charts. Parallel Coordinates are used to visualize multivariate data concerning patient journeys at the individual level. Sankey charts assist in visualizing the flow of patients between various phases of patient journeys. Sunburst charts provide a representing hierarchical relationship between dia- gnoses, procedures, and prescription medications (Kaushal, 2017). Ha et al. (Ha, 2019) developed RadVIS a 3D radial model to assist psychiatrists in evaluating multidimensional data from groups of patients with dementia. It allows the user to get a better understating of the characteristics of patient cluster and analyze the variable values of data comprising each cluster at the same time. The user can choose the number of clusters for segmenta- tion after selecting either a foggy cluster or a random cluster algorithm. A patient with dementia is represented by a single node in this visualization. RadVis also supports a multi-filtering function through parallel coordina- tes plot to assign different conditions for a more comprehensive analysis. VISA_M3R3 is a system designed to help clinical researchers to identify medications and medication combinations that can be associated with a
Radial Visualization Model in Health Care: A Survey 257
higher risk of acute Kidney injury (AKI). By integrating multiple regres- sion models, frequent itemset mining, data visualization, and human-data interaction mechanisms, VISA_M3R3 allows users to explore complex relati- onships between medications and AKI. The analytics module of VISA_M3R3 performed a single-medication analyzer that focus on finding associations between each medication and AKI. Also has a multi-medication analyzer that purposes to identify the medication combinations that are associated with AKI. (Abdullah, 2020).
DISCUSSION AND CONCLUSIONS As shown in this paper, there has been an intensive discussion around what is the best graphic way to represent big amounts of data so there is a better communication to the users. A better communication consists of capturing the most important messages for a given use in an efficient and timely way. This characteristic is mostly important in the healthcare sector where frequ- ently decisions must be made based on a big quantity of different types of data in a very short period. Radial models can play an important role in this context. Not being a “one-fits-all” solution it proves to be most suitable to many situations in health care. The different case studies use diverse graphic approaches: radial (Joshi, 2012) (Zhang, 2013) (Baytas, 2016) (Senathirajah, 2017) (Kaushal, 2017) (Ha, 2019) (Abdullah, 2020); human-body based (Jin, 2016) (Zhang, 2013) and circular plots (Hu, 2014) (Ledesma, 2016). The radial models further divide in radial (Abdullah, 2020), radial tree (Baytas, 2016), radial sunburst/starburst (Joshi, 2012) (Senathirajah, 2017) (Kaushal, 2017) and 3D radial (Ha, 2019). In the healthcare sector this kind of graphics is used for different goals. There are some used for analytic purposes, such as (Hu, 2014) (Baytas, 2016) (Abdullah, 2020). Others are used in clinical envi- ronments. Of these some are dedicated to patient’s overview (Joshi, 2012) (Zhang, 2013) (Follett, 2012) (Senathirajah, 2017) (Kaushal, 2017) (Ha, 2019), and one explores the prescription medication (Kaushal, 2017). This survey shows that the radial model is used in many situations related with healthcare, providing a variety of options that cover a wide range of appli- cations, proving that Information Visualization research plays an important role in shortening the distance between users and their own health data.
REFERENCES
Abdullah SS, Rostamzadeh N, Sedig K, Garg AX, McArthur E. (2020) Multiple Regression Analysis and Frequent Itemset Mining of Electronic Medical Records: A Visual Analytics Approach Using VISA_M3R3. Data. 5(2):33. Abdullah, S. S. (2020) Visual Analytics of Electronic Health Records with a focus on
Acute Kidney Injury. Electronic Thesis and Dissertation Repository. 7086. Aigner W., Federico, P., Gschwandtner T., Miksch S., Rind A. (2012) Challenges of Time-oriented Data in Visual Analytics for Healthcare.IEE VisWeek Workshop on Visual Analytics in Healthcare. Bastardo, R., Castro, M., Pavão, J., & Ramos, L. (2021). A timeline model for clinical
events: Empowering data. Paper presented at the Procedia Computer Science, 196 pp. 1053–1058.
Baytas M., Lin K.,Wang F., Jain A. K. and Zhou J. (2016) PhenoTree: Interactive Visual Analytics for Hierarchical Phenotyping From Large-Scale Electronic Health Records. in IEEE Transactions on Multimedia, vol. 18, no. 11, pp. 2257–2270, (2016) Boyd AD, Young CD, Amatayakul M, Dieter MG, Pawola LM. (2017) Develo-
ping Visual Thinking in the Electronic Health Record. Stud Health Technol Inform.;245:308-312. Bui, A.A.T., Aberle D.R., and Kangarloo H. (2007) , TimeLine: Visualizing Integrated Patient Records, IEEE Trans. Inf. Technol. Biomed. 11 462–473. Diehl S., Beck F. and Burch M. (2010) Uncovering strengths and weaknesses of radial
visualizations—an empirical approach. IEEE Transactions on Visualization and Computer Graphics, 16(6):935–942. Draper, G., Livnat Y., Riesenfield, R. F. (2009) A survey of radial methods for information visualization. IEEE TVCG 15 (5): 759-776 Filipov, V., Schetinger, V., Raminger V., Soursos N., Zapke, S., Miksch (2021)
Gone full circle. A radial approach to visualize event-based network in digital humanities.” Visual Information 5: 45–60. Follett J, Sonin J. hGraph (April 2012) An Open System for Visualizing Personal Health 26. Metrics. Technical report, Involution Studios, Arling- ton, Massachusetts. Accessed 03 Jan 2022. http://www.hgraph.org/docs/ hGraph_Whitepaper.pdf. Goren-Bar, Dina, Yuval Shahar, Maya Galperin-Aizenberg, David Boaz, and Gil
Tahan (2004) KNAVE II: the definition and implementation of an intelligent tool for visual-ization and exploration of time-oriented clinical data. In Proceedings of the working conference on Advanced visual interfaces, pp. 171–174. ACM Ha H, Lee J, Han H, Bae S, Son S, Hong C, Shin H, Lee K. (2019) Dementia Patient Segmentation Using EMR Data Visualization: A Design Study.” Int J Environ Res Public Health. 16(18):3438. Hirsch J.S., Tanenbaum J.S., Lipsky Gorman S., Liu C., Schmitz E., Hashorva D.,
Ervits A., Vawdrey D., Sturm M. and Elhadad N. (2015) HARVEST, a longitudinal patient record summarizer, J. Am. Med. Inform. Assoc. JAMIA. 22 263–274. Hoffman P., Grinstein G., Marx K., Grosse I. and Stanley E. (1997) DNA visual and analytic data mining. Proceedings. Visualization ‘97 (Cat. No. 97CB36155), pp. 437–441. Hu Y, Yan C, Hsu CH, Chen QR, Niu K, Komatsoulis GA, Meerzaman D. (2014)
OmicCircos: A Simple-to-Use R Package for the Circular Visualization of Multi- dimensional Omics Data. Cancer Inform. 16; 13:13-20. Jin, Y. (2016) Interactive Medical Record Visualization based on Symptom Location in a 2D Human Body. Joshi R, Szolovits P. (2012) Prognostic physiology: modeling patient severity in Inten-
sive Care Units using radial domain folding.”AMIA Annu Symp Proc.:1276-1283. Kaushal K. K. et al. (2017) “Patient Journey Visualizer: A Tool for Visualizing Patient Journeys,” 2017 International Conference on Machine Learning and Data Science (MLDS), pp. 106–113. Ledesma, A., Al-Musawi, M. & Nieminen, H. (2016) Health figures: an open-source
JavaScript library for health data visualization. BMC Med Inform Decis Mak 16, 38 Ledesma, A., Bidargaddi N., Strobel J., Schrader G., Nieminen H., Korhonen, I., Ermes, M. (2019) Health timeline: an insight-based study of a timeline visua- lization of clinical data. BMC medical informatics and decision making, 19(1): 170.
Radial Visualization Model in Health Care: A Survey 259
Maças C. and Machado P. (2018) Radial calendar of consumption. In: 22nd International Conference Information Visualization (IV): 96–102. Martignene, N., Balcaen, T., Bouzille, G., Calafiore, M., Beuscart, J.B., Lamer, A.,
Legrand, B., Ficheur, G., Chazard, E. (2020) Heimdall, a Computer Program for Electronic Health Records Data Visualization. Stud Health Technol Inform. Jun 16; 270:247-25. Monroe M., Lan R., Lee H., Plaisant C. and Shneiderman B. (2013) Temporal Event Sequence Simplification, IEEE Trans. Vis. Comput. Graph. 19: 2227–2236. Perer A., Wang F. and Hu J. (2015), Mining and exploring care pathways from
electronic medical records with visual analytics, J. Biomed. Inform. 56: 369–378. Plaisant, C., B. Milasch, A. Rose, S. WIdoff and B. Shneiderman (1996) LIfeLines: visualizing personal histories. SIGCHI Conference on Human Factores in Computing Systems Proceedings 221-227. Senathirajah Y, Kaufman D, Cato K, Borycki E, Fawcett J, Kushniruk (2020) A
Characterizing and Visualizing Display and Task Fragmentation in the Electronic Health Record: Mixed Methods Design.JMIR Hum Factors 7(4): e18484. Senathirajah Y, Wang J, Borycki E, Kushniruk A. (2017) Mapping the Electronic Health Record: A Method to Study Display Fragmentation. Stud Health Technol Inform. 245:1138-1142. W. Aigner, S. Miksch, W. Müller, H. Schumann, and C. Tominski (2008) Visual meth-
ods for analyzing time-oriented data.” IEEE Transactions on Visualization and Computer Graphics, 14(1):47–60. West, V., Borland, D., Hammond, W.E. (2013) Visualization of I and Health Related Data for Information Discovery. Workshop on Visual Analytics in Healthcare. Wongsuphasawat K., and Gotz D. (2012) Exploring Flow, Factors, and Outcomes
of Temporal Event Sequences with the Outflow Visualization, IEEE Trans. Vis. Comput. Graph. 18: 2659–2668. Zhang Z, Wang B, Ahmed F, Ramakrishnan IV, Zhao R, Viccellio A, Mueller K.
(2013) The five Ws for information visualization with application to healthcare informatics. IEEE Trans Vis Comput Graph. 19(11):1895–910.