LabVis: Simple and Intuitive Visualization of Medical Laboratory Results


Adam Bodnar and Dmitry Nekrasovski


April 21, 2004





This paper presents LabVis, an interactive visualization tool for displaying and interacting with medical laboratory results. The motivation for LabVis is presented, along with relevant work in time-based medical information visualization and medical lab results visualization.  We discuss the types of tasks that LabVis is designed for, describe the visualization techniques used by the tool, and show its operation on a real-world dataset of thyroid function tests.  We then describe the results of an informal evaluation and provide directions for future work.




The rapid and accurate diagnosis of a medial condition is the primary goal of any physician.  A critical factor when making a diagnosis is the type and quantity of information the physician has about the patient.  The higher the quality of this information, the more informed the physician can be about possible medical conditions and corresponding plans of treatment.


The area of laboratory medicine is the medical specialty that deals with all aspects of performing and interpreting laboratory tests.  While performing these tests is almost entirely within the domain of the clinical laboratory, the interpretation is left to the physician.  For an experienced physician, the process of diagnosis is challenging, but for medical students, the process of learning to select and interpret appropriate laboratory tests can be enormous [9].


Our motivation for exploring this area lies in the realization that current methods for displaying and interacting with results are static, error prone, and out-dated, and do not take advantage of recent innovations in visualization and computer technology, which could lead to higher quality patient information, and thus faster and more accurate diagnosis.


In this paper we present LabVis, an interactive visualization tool for displaying and interacting with laboratory test results.  LabVis is designed to visualize thyroid function test results over time for the purpose of assisting with diagnosis and aiding in the monitoring of effects of treatment of known thyroid disease.  In this manner, LabVis can act as a tool in the education of medical students in the process of learning to select and interpret laboratory tests.  


Related Work


Visualization for medical applications has had a long history in the field of medical imaging. However, medical imaging has been traditionally concerned with problems of image acquisition, such as using tomography scans, and the graphics techniques required to visualize acquired images, such as anti-aliasing and volume rendering, rather than visually presenting medical data in an intuitive and easy to navigate format. Proposals have been made for using information visualization to both augment existing medical imaging applications with new capabilities and tackle problems that lie outside the scope of medical imaging, and a number of medical information visualization systems have been proposed; however, the field is still in its infancy [1].


A number of medical information visualization systems have dealt with visualizing time-based data. One of the first such was the Time Line Browser [2],  which visualizes events such as clinical lab results and patient states on a timeline. This concept was further developed in Lifelines [8], a system that provides the user with a visualization of patient histories grouped by category such as consultations, conditions, hospitalizations, medications. This view is augmented with temporal zoom, details (such as patient lab reports) on demand, and an ability to dynamically rearrange the events and intervals displayed. Another approach taken by researchers has been to augment existing time-based information representations used by clinicians with additional visualizations. For instance, [7] proposes extensions such as colour coding to convey information about the reliability of time-based data displayed on anesthesia monitors.


The information visualization approach taken by LabVis was particularly influenced by a system for visualizing laboratory medical laboratory data on handheld devices described in [6]. This research proposes a number of innovative visualization techniques to support rapid patient assessment, diagnosis, and long-term patient management.  Rapid patient assessment was achieved through highlighting of abnormal test results and linking them to their corresponding anatomical systems. Through an informal user test, it was found that the highlighting of abnormal test results was useful in rapid patient assessment, but the linking of the results to their corresponding anatomical system was not required.  For this reason, LabVis supports the highlighting of abnormal test results without the visual clutter of anatomical displays.  Diagnosis was achieved though a combination of hierarchical grouping of laboratory results with an elision technique to support the selective viewing of test groups. Similarly, LabVis groups laboratory results hierarchy, with primary tests and diagnosis taking precedence over secondary tests, and supports selective visualization of laboratory tests.  Long-term patient management was achieved through the display of an Excel like spreadsheet which highlighted abnormal results.  With LabVis, this is achieved by displaying time-series plot of patient tests, where abnormal results are highlighted and easily browsable. 


Problem Description

Current methods of reporting and interacting with medical laboratory results consist of a static document, containing both text and numbers, which is printed out on paper or displayed monochromatically on a computer screen (see figure 1).

Figure 1 – Current format for test results

In order to make a diagnosis, a physician will typically have to compare the test results of several analytes (chemicals and hormones whose levels can aid in diagnosing an illness) to their corresponding reference ranges.  This process can be tedious, time consuming, and error prone.  While this process may be sufficient for some physicians, it does not take advantage of advances in both computer and visualization technology, which could lead to increases in efficiency, and help in the process of teaching medical interns the process of diagnosis [9].  With LabVis, we aim to present the physician with a clear picture of patients’ laboratory results, which will allow the physician to rapidly assess the state of the patient, and aid in the process of diagnosis.

LabVis is part of a larger project, which aims to support the interpretation of laboratory tests in real time.  Lead by Dr. Wes Schreiber of the Department of Pathology and Laboratory Medicine at Vancouver General Hospital, this project aims to aid in the education of medical students and residents in the ordering and interpretation of laboratory tests.


The task is to display a series of interactive, effective visualizations that illustrate the results of laboratory tests.  The purpose of the visualizations is to support the physician in the process of diagnosis, and act as a teaching tool for new medical students and interns, for whom the challenge of interpreting appropriate laboratory tests can be enormous [9].

For the purposes of this project, the visualization involves the results of a thyroid function test, which is commonly used to diagnose disorders of the thyroid gland and also to monitor the effects of treatment of known thyroid disease [5]. The common analytes measured by a thyroid function test are free thyroxine (FT4) and thyroid stimulating hormone (TSH). The reference ranges (ranges of test values which are considered normal) for TSH and FT4 are 12-21 nmol/L and 0.3-4.0 miU/L, respectively [5]. In addition, optional supplementary analytes may also be measured when FT4 and TSH levels are outside the normal bounds. These include thyroxine (T4), free triiodothyronine (FT3), and anti-thyroid peroxidase antibodies (anti-TPO) [4].


The target data set is a set of approximately 1500 thyroid function tests, all of which include FT4 and TSH results. Some of the tests also contain measurements of T4, FT3, and anti-TPO. Most patients whose tests are included in the dataset have been tested multiple (up to 15) times.

Scenarios of Use


LabVis is intended to be used in a variety of ways to help medical personnel and students view and interpret thyroid function test results. For clinicians, the most common goal is likely to be the efficient interpretation of data for a single patient, either for diagnostic analysis or for monitoring the patient’s history over time. However, it is also possible to use LabVis to visualize a dataset containing multiple patient records, for example by clinical researchers  investigating outlier test results. Possible scenarios illustrating how the visualizations presented by LabVis could be used for both these purposes are presented below.

Scenario 1: Interpreting a specific test result

1.      A clinician activates LabVis and is prompted to enter a patient record or browse all available patient records.

2.      The clinician enters P1. The application displays a visualization of patient P1’s data.

3.      The clinician inspects the top (time-series) visualization and notices that, while the patient’s second-most recent test data is normal, his/her most recent TSH sample doesn't fall into the normal range.

4.      The clinician selects the most recent set of tests in the top visualization. This set of tests is then highlighted in red in both the top visualization and the bottom-right (scatterplot) visualization.

5.      The scatterplot visualization displays a possible diagnosis, T3 toxicosis, beside the selected point. To confirm this diagnosis, the clinician requires a measurement of the FT3 analyte.

6.      The clinician refers to the bottom-left (tabular) visualization, observes that the highlighted test results do not include a FT3 measurement, and orders an FT3 test to be performed on the patient.

Scenario 2: Investigating an outlier test result

1.      A clinical researcher enters LabVis and is prompted to enter a patient record or browse all available patient records.

2.      The researcher chooses to browse all patient records. The application displays a visualization of all available data.

3.      The researcher notices that one of the data points on the scatterplot visualization shows an excessively high value for FT4.

4.      The researcher selects the data point, refers to the tabular visualization to determine which patient it belongs to, and makes a note to investigate the reasons for the abnormal result.


Information Visualization Techniques


LabVis is made up of three related visualizations, which are linked and aimed at providing the physician with both an overview of patient test results through time and specific information for one or more tests.  This section details the various visualization and interaction techniques used by LabVis.




Visualization 1 – Time-Series Overview of FT4 and TSH


This visualization displays a time-series based overview of FT4 and TSH laboratory test results for a specific patient (see figure 2). It is used to monitor the effects of treatment of known thyroid disease [5].  Reference range highlighting is used to indicate the area where normal results should lie.


Figure 2 - Time series overview of FT4 and TSH with normal range highlighting


Visualization 2 – FT4 vs. TSH Scatterplot


This visualization displays a scatterplot of FT4 vs. TSH (see figure 3).  It is used to diagnose specific disorders of the thyroid gland [5] for the test results chosen by the physician based on the primary time-series overview.  Reference range highlighting is used to indicate the area where normal results should lie.  Points not contained within the bounds of the highlighted range are considered abnormal and a possible diagnosis is provided for them when they are selected by the user.


Figure 3 - FT4 vs. TSH scatterplot with normal range highlighting


Visualization 3 – Tabular Laboratory Test and Diagnosis Details


This visualization displays the details of the patient record, the numerical test results, as well as supplementary laboratory test data (see figure 4).  It is used to provide the physician with the specific details about each test result, and also provide a possible diagnosis for those results that lie outside the normal range.


Figure 4 – Laboratory test and diagnosis details


Together, these visualization panels make up the LabVis interface (see figure 5)


Figure 5 - LabVis interface with linked highlighting of points enabled

Visualization Details


Color and Transparency


LabVis takes advantage of the way color can be perceived preattentively by the human perceptual system, and aims to rapidly draw the attention of the physician to test results that lie outside the normal range. With the aid of an experienced graphic designer, we decided on a color scheme that aims to maximize contrast between normal and abnormal results, while also supporting linked highlighting of results and display of test result date labels.  Our color scheme also passes the common tests for color blindness (deuteranope and protanope) available from VisCheck.  Table 1 provides the specific details regarding the colors that are used in LabVis.



Visualization Object





Default Test Result Point





Selected Test Result Point





Normal Range Highlighting





Visualization Panel Background





Test Result Date Label





Table 1 - Color and transparency details for LabVis interface features


While occlusion is possible with LabVis, it is generally rare due to the small number of test result points typically displayed for a single patient. Label occlusion is dealt with by providing labeling for the selected data points only. Transparency is used to deal with occlusion of test result points, both by other test result points, and by the reference range highlighting. 

Information Density


The empirical findings of Somervell [10] suggest that low information density visualizations are preferable to high information density visualizations in an ‘attention-limited’ environment where the visualization is not the primary focus of the user. Thus, we designed LabVis as a secondary display to enable effective visualization of laboratory test results while utilizing a small information density.  In this manner, the physician can maintain his primary focus on the state of the patient and/or on interpreting other patient information, while referring to LabVis for assistance with the diagnosis and the patient’s history.

Linked Highlighting and Labeling


Each of the three LabVis visualization panels is tightly coupled, so that a change made in one panel is immediately reflected in the other panels.  Linked highlighting of data points is used to connect the test results in one window to each of the other windows. Linked highlighting is augmented by labeling of highlighted data with information appropriate for the particular visualization (test date for the time-series visualization and possible diagnosis for the TSH vs. FT4 scatterplot visualization).

Spatial Layout


The organization of points within each visualization panel is critical to the usability of LabVis.  LabVis uses a standard layout algorithm, which has been modified to work with our dataset, to position points relative to the minimum and maximum possible values for each analyte.  Using this technique, normal range highlighting remains constant in position from one patient to the next, and points that are similar in value cluster together, thus enabling the physician to quickly assess which points lie outside the normal range.




LabVis was implemented using the Java Swing API and The InfoVis Toolkit [3] developed by Jean-Daniel Fekete. The InfoVis Toolkit is designed to support the creation, extension and integration of advanced 2D information visualization components into interactive Java Swing applications, and provides a rich set of data structures and visual components for this purpose.


LabVis was designed using a Model-View-Controller (MVC) architecture which relies on and/or extends several modified InfoVis Toolkit packages, specifically the table and column packages for representing data, the visualization packages for representing visualizations, and the panel packages for rendering and displaying visualizations as Swing components. LabVis is implemented as both a standalone Java application and as a Java applet for demonstration purposes. The latter is available at


While LabVis is currently implemented to visualize thyroid function test results, it is designed so that other types of time-series medical laboratory test results could be visualized.  This would require minimal changes to the source code and interface.




We conducted an informal evaluation of the LabVis prototype with our lead user, Dr. Wes Schreiber of the Department of Pathology and Laboratory Medicine at Vancouver General Hospital. The prototype was also evaluated with three other users who did not possess prior knowledge of the medical lab visualization domain. These users were explained the nature of the prototype and the dataset, and a cognitive walkthrough was performed with them by one of the authors through the two scenarios described in the Scenario of Use section. Users were then asked to comment on the perceived clarity and usability of the visualization and suggest possible improvements. Despite lack of domain expertise (except for Dr. Schreiber), all users found the system to be intuitive and to clearly communicate information sought in the scenarios. One user suggested that the colour scheme of the application provided insufficient contrast between the data points and the reference ranges.


The evaluation feedback on the prototype received from Dr. Schreiber was highly positive. He stated that LabVis provided a significant improvement on existing paper-based visualizations of thyroid test data by enabling the user to monitor trends in test results over time and relate them through linked highlighting, while retaining the ability to perform diagnostics using numerical test data. Dr. Schreiber suggested that the prototype could be improved by separating the two scatterplot visualizations and the tabular visualization (e.g. by means of tab-activated views) to reduce clutter and increase screen real estate available to each visualization, as well as by adding reference range data and highlighting of normal results to the tabular visualization. Overall, Dr. Schreiber indicated that he viewed the LabVis project as a promising contribution to his goal of using technology to improve the education of medical students and residents in the ordering and interpretation of laboratory tests.


Lessons Learned


Through the course of developing and evaluating LabVis, we learned several things:



  • An understanding of target user preferences and tasks is essential to the development of medical information visualization applications.


  • Collaborating with experts from disciplines outside Computer Science, such as Medicine and Graphic Design, can be highly beneficial in the development of visualization interfaces


  • Drawing the attention of the user to important properties of the data isn’t a trivial task, and requires a balance of color, spatial positioning, and information density.


  • Building an application on top of an existing toolkit can aid in rapid development, but can limit creativity and lead to frustration when implementation problems arise.


  • The medical profession uses a lot of complex terms and acronyms to talk about relatively simple concepts.


Conclusions and Future Work


An important challenge in the development of visualization tools for medical laboratory results is to display sufficient information in a format that supports the physician in making a rapid and accurate diagnosis, without overwhelming or confusing the physician with information or visualizations not directly related to patient care.  In this paper we have presented LabVis, a visualization tool for interacting and displaying medical laboratory results which meets this challenge.  LabVis utilizes several information visualization techniques, which aim to support physicians, medical students, and clinical researchers in the processes of analysis of test results, diagnosis, and patient history monitoring.  LabVis was developed using a task centered design and evaluation.  Through developing a set of possible scenarios and evaluating them with users, we have illustrated the utility of LabVis, and shown how it improves on existing techniques for dealing with laboratory results.


Based on the results of our informal user study, we have shown that LabVis achieves the goals set out by [1] in that it:


  1. Allows the physician to explore the available data at several levels of abstraction


  1. Provides the physician with a greater sense of engagement with the data


  1. Gives the physician a deeper understanding of the data


  1. Encourages the discovery of details and relations which would normally be difficult to notice


  1. Supports the recognition of patterns by exploiting the visual recognition capabilities of the physician



The results presented motivate several future research directions for medical laboratory test visualization tools.  We would like to incorporate the valuable feedback gained from our informal evaluation into a new version of LabVis, and to evaluate its usability and utility in a formal user study with actual physicians and medical students. Various other improvements could also be made to the visualizations used in LabVis.  Scaling the analyte axes logarithmically rather than linearly could lead to less occlusion of data points and faster interpretation of results.  A supplementary visualization could be created to allow a clinical researcher to view the entire dataset of all patient test results.  This visualization might be used to provide a focus + context view of patient test results and allow clinical researchers to explore patterns of patient care. Finally, we are also interested in bringing LabVis to handheld device platforms, such as Palm or PocketPC, to enable rapid visualization and interpretation of lab test data in a portable environment. 





1.      Chittaro, L. Information Visualization and its Application to Medicine. In Artificial Intelligence in Medicine, vol. 22(2), 2001, pp. 81-88.

2.      Cousins SB, Kahn MG. The visual display of temporal information. In Artificial Intelligence in Medicine, vol. 12(3), 1991, pp.341-357.

3.      Fekete, J-D., The InfoVis Toolkit, Research Report RR-4818, INRIA Futurs, May 2003. Available from

4.      Great Smokies Diagnostic Laboratory, GSDL Comprehensive Thyroid Assessment, Asheville, NC, USA, 2002.

5.      Jones, G. Thyroid Function Tests. The Pathology Service of St. Vincent's Hospital, Sydney, Australia, 2000.

6.      Lam, H. Medical Reports Display on Small Screens. Master’s thesis, Department of Computer Science, Simon Fraser University, Vancouver, Canada, 2004.

7.      Lowe A, Jones RW, Harrison MJ.  The graphical presentation of decision support information in an intelligent anaesthesia monitor. In Artificial Intelligence in Medicine, vol. 22(2), 2001, pp. 171-183.

8.      Plaisant, C., Milash, B., Rose, A., Widoff, S., Shneiderman, B. LifeLines: Visualizing Personal Histories. In Proceedings of ACM SIGCHI 1996, pp. 221-227.

9.      Schreiber, W. Interpretation of Laboratory Tests in Real Time with Real Patients. Hand-Out, Department of Pathology and Laboratory Medicine, Vancouver General Hospital, Vancouver, Canada, 2004.

10.  Somervell, J., McCrickard, D. S., North, C., and Shukla M. An Evaluation of Information Visualization in Attention-Limited Environments in Proceedings of the IEEE TCVG Symposium of Visualization (VisSym '02), Barcelona, Spain, May 2002