University Hospital of the Albert-Ludwigs-University of Freiburg: Institute of Medical Biometry and Medical Informatics

Our instituteDepartment of Medical Informatics

Visual BayesVisual Bayes - A tutoring and simulation system for the validation of diagnostic tests German

Visual Bayes is an interactive computer based training programme that presents the basic methods for the interpretation and validation of diagnostic tests in an intuitive way. Independent on mathematical and other preliminary knowledge it addresses to students of medicine and is conceived in especial as a supplement to the course of biomathematics.

(Installation see Software.)

Design of the programme

Visual Bayes is arranged in two parts. In the first part (dichotome tests), the central notions and definitions in context with diagnostic tests are trained. The student is taught how to interpret a 2x2 table, in especial how to obtain the important and significant values prevalence, sensitivity and specificity. Core of this part is the formula of Bayes, which ties these values together and allows a precise valuation, to what extent a positive or negative test result respectively indicates the existence or non-existence of a certain disease (predictive values). It constitutes the mathematical foundation of the process of logical deduction and has nowadays found a wide range of application in medical diagnostics.

In the suite, the predictive values and predictive gains calculated by Bayes' formula are demonstrated graphically in an additional chapter. Especially, it allows to test their dependence on prevalence, sensitivity and specificity, and thus passes on to the following part of the programme, in which the user has to a larger extent the possibility to define own values and try out their effects. To round up the first part, a chapter about the combination of tests is joined, which imparts to the student how to evaluate even advanced and complicated processes of decision on the ground of the theorem of Bayes.

As the comprehension of these fundamental facts is of great importance, in the first part of the programme the student is brought to train and try out his knowledge continually by questioning the previously explained subjects. Although it is possible to vary nearly every given value, the general structure is kept linear. Thereby, the logical construction and coherence is lined out, which simplifies considerably the understanding and memorising of the abstract matter in its context. Examples out of medical practice illustrate the practical use and freshen up the theory.

Reverend Thomas BayesIn the second part of the programme, this concept is expanded onto tests with a one dimensional reel test variable. Therefore, likelihood densities like the bell-shaped Gaussian distribution or a logarithmic normal distribution can be selected. In form of a histogram, the data set of a real sample is simulated, and by the choice of the cut-off point such a test can be dichotomised. The thereby resulting characteristic values sensitivity and specificity are represented by a receiver operating characteristic curve, which allows to estimate the quality of a test. Finally, the influence of the cut-off point onto mean costs is regarded by way of the cost function.

The essential targets of the second part of the programme are communicated mainly in optically suggestive form by means of graphics and diagrams. Intentionally the interest of the user is awakened, who in this part shall train his understanding by way of his own intuition. By visual means like, for example, the possibility to draw and shift own distributions, additional play-scopes are created which allow it to vary the values to a certain extent and thus to deepen ones understanding by own exploration. In addition to the thorough training of the definitions in the first part, here the second important aspect in learning theoretical facts is emphasised: The student shall develop a sense for their relations between each other and their connections to reality.


The Bayes project descends in its origins to the free University of Amsterdam and has later been developed and carried on at the University of Limburg in Maastricht in MUMPS. By the Departments of Medical Informatics of the Universities Maastricht, Freiburg and Gent, it has been rewritten as an MS-DOS version (German) in PASCAL and translated into German, French and English. These tasks have been supported by the ERASMUS-programme of the commission of the European Community.

The present once more improved and enlarged version for Windows has been developed by the department of Medical Informatics at the University Freiburg/Breisgau (Germany) with Visual Basic 4.0 (16 bit). The current English translation will shortly be reviewed by a native speaker.


Visual Bayes is freeware, and can thus be copied and distributed without registration. It is not chargeable; nevertheless, for the further development of the project a voluntary contribution is welcome under


System requirements: PC with 4 MB RAM and 16-Bit CPU, 256 colours in VGA mode, Windows 3.11 or higher.

The installation of Visual Bayes (bilingual) is possible directly from the Internet. Please choose one of the following two options:

  1. Executable version, 16 Bit [0,8 MB]. The file vbayes1.exe is self extracting and unpacks when being executed.

  2. Executable version, 32 Bit [0,8 MB]. The file vbayes2.exe is self extracting and unpacks when being executed.

  3. Complete set-up version, 16 Bit [1,9 MB] with further library files (which normally already ought to be present). The file vbayes3.exe is self extracting and unpacks when being executed, after which the set-up file setup.exe must be run. This version exists as well on Disk 1 and Disk 2.

No further steps are necessary. All installations will result in the programme file vbayes.exe. For more information, please consult the included file readme.txt.

In addition, Visual Bayes can be run within the university by using the clinical software service InfoServer (see German page).

To every passing on of the programme we join the request to inform us in brief about its employment. News about occurring problems, propositions for improvement, suggestions and experiences are of especial value for the further development of Visual Bayes, and are accepted thankfully per e-mail.

Last update 28. August 1998 by: Pius Franz