DR. GAIT III - A SYSTEM FOR GAIT ANALYSIS

Kathy A. Johnson1, Rebecca Denning2, Philip J. Smith3, Jack W. Smith1, & Sheldon R. Simon4 1Dept. of Pathology, 2Cognitive Science Center, 3Dept. of IWS Eng., 4Dept. of Orthopaedic Surgery The Ohio State University, Columbus, OH U.S.A.

ABSTRACT

In this paper we describe a multimedia system for gait analysis supporting two distinct functions - tutoring and report generation. Dr. Gait III provides a means to view and electronically annotate all the information commonly present for gait analysis: medical history, physical examination, time/distance data, joint angle graphs, moments, powers, force plates, EMGs, video, and stick figures. It provides annotation functions for each type of data and also contains decision support tools to automatically annotate some types of data. For tutoring, the annotation functions are combined with an audio recording system to record questions, hints, and answers. A suite of teaching cases is being developed for distribution with the program. The system may also be used as an authoring system for designing tutoring sessions for new cases. For report generation, the same annotation functions are used. In addition the system provides a way to tag individual pieces of data for inclusion in a report. The report is then composed in a word processor where further editing can take place.

BACKGROUND

Gait analysis is the process of determining the cause of abnormalities in a patient's walking pattern so as to treat those abnormalities and improve the patient's functionality (1). A gait analysis session consists of several steps. First, the patient's physical characteristics such as leg length and weight are measured. Next, several range of motion parameters are measured such as the flexion/extension of the hip, knee, and ankle. Reflective markers are then affixed to the patient in various locations and a camera system records the trajectory of these markers in 3-D space while the patient walks. Surface or wire electrodes are also used to determine the muscle activity of the muscles of interest. This information is generally collected by computer and resides in electronic forms. In addition, most labs also record video of the patient walking.

PURPOSE OF THE SYSTEM

Many areas of rehabilitation engineering use data from a variety of sources and media types. In gait analysis, a study consists of numerical data, textual data, graphical data, and video. A study typically consists of a set of printed reports from several sources as well as video of the patient. This data is then used to generate the gait analysis report and is also often used for teaching students to perform gait analysis. Such tasks become easier when all the information is available in one place. A computer is ideal for this purpose as it can display and organize all the different kinds of quantitative information, the textual information, and the video. Looking at different pieces of information is as easy as selecting a different item from a menu. For both report generation and tutoring purposes, one of the main functions needed is to draw a person's attention to pertinent pieces of information. We have developed a set of annotation tools for each type of information (text, graph, video) which add an appropriate type of visual change to an item in order to highlight that item. These annotations will be discussed further below. Another benefit of moving the data to a computer-based form is that additional support functions can be added. For example Dr. Gait III can be instructed to automatically annotate regions of the graphs that are significantly different from normal. This function makes report generation quicker because it automates a step that the analyst would do anyway. Furthermore, for tutoring purposes it can serve as an example for how to do the gait analysis steps.

DESIGN

Dr. Gait III merges and expansion of two previous programs: GAIT (Gait Analysis Interpretation Tool) (2), and QUAWDS (QUalitative Analysis of Walking DisorderS) (3). GAIT is a tutoring system that pioneered the interface and tutoring functions that are now in Dr. Gait III. QUAWDS is a knowledge-based system for gait analysis of CP patients. It provides the basis for the decision-aid tools that are embedded in Dr. Gait III. QUAWDS' roots go even further back to the systems DR. GAIT-1 and DR. GAIT-2 (4).

Dr. Gait III uses data directly from the gait analysis laboratory. The data that can be viewed on-line includes: medical history, physical exam, time and distance data, joint angle graphs, moment graphs, power graphs, force plate graphs, EMGs, Quicktime video, and animated stick figures. The stick figures are generated from actual marker data by the program.

Data Annotation

Each screen of data may be annotated in a manner appropriate to the type of data. Text screens provide the capability to change the style (bold, italics, etc.) and color of the text. The colors that are made available have been chosen for saturation and depth such that they are distinctly identifiable when printed on a grayscale printer. Screens with graphs provide two means of annotation - a colored highlight of a region and colored arrows that may be placed anywhere on the screen. Figure 1 shows an example of a red highlight and a green arrow on the sagittal joint angle graph for knee flexion. The user may place as many different highlights and arrows as they wish on a screen.

Figure 1: Right Knee Sagittal

JAG Video screens also provide two means of annotation for any frame of information colored arrows and colored vectors that can be drawn on the video. In each case the annotations remain with a single frame of video and other annotations can be added to subsequent or previous frames without affecting other frames. Figure 2 shows an example of video annotation.

Figure 2: One frame of side stick video

Decision Support

In addition to allowing the user to annotate items, we also want to provide decision support tools to aid report generation and tutoring. QUAWDS, provides a basis for these support tools. QUAWDS broke the entire task of gait analysis into several steps: finding determination, muscle fault generation, muscle fault rating, and generation of explanatory coverage of muscle faults (3). The first of these, finding determination, has been incorporated into Dr. Gait III for the joint angle graphs, physical exam and EMGs. For these data items, the system can automatically annotate the items that are abnormal.

In the case of the joint angle graphs, we have two methods available for determining abnormality - by standard deviation or by prediction region (5). Each of these methods highlights the abnormal regions of the graphs.

For the physical examination, Dr. Gait III uses a graduated system to determine whether the item is severely out of range or mildly out of range. Those that are severe are marked in red and those that are mild are marked in orange.

For the EMGs, the system uses a user-defined threshold to determine activity for 6 phases (weight acceptance, single limb stance - first half, single limb stance - second half, weight release, swing - first half, swing - second half). If the activity in a phase is all below the threshold and there is a normal indicating that there should be activity, then that region of the graph is highlighted to indicate disphasic activity. Similarly, if there is activity in a phase outside the threshold and the normal indicates that there should not be activity, then the region is also highlighted to show disphasic activity. Figure 3 shows an example of an EMG where the disphasic activity has been highlighted.

Right Anterior Tibialis Figure 3: EMG showing disphasic activity

These various abnormality findings are then available for further processing by the decision support system. We are currently working to add the remaining subtasks from QUAWDS as decision support tools: muscle fault generation, muscle fault rating, and muscle fault explanatory coverage.

Tutoring

A teacher can annotate any of the several screens of data with the kinds of visual cues that were described above such as highlighted text and colored arrows. For each screen, the teacher also records up to 9 sets of audio comments, each consisting of a question, a hint and an answer. The student then reviews the case, listening to the questions (and hints if available) and records his own answer to each question. The student is blocked from listening to the teacher's answer until he has recorded his own audio answer. These student answers are available for later review by the teacher, but the student gets immediate feedback about the answer by listening to the pre-recorded answer. We are currently developing a suite of teaching case for which the tutoring aspects have already been authored. In addition, the system may be used to author new cases for which no tutoring has been recorded.

Report Generation

Reports may be generated by annotating the data items with the program and indicating which items should be included in the report. Once the person is done marking items, the items are exported to a word processor where additional editing can take place and a hardcopy report can be printed. Currently the items that may be included in the report include: text screens (either an individual item or the entire screen), graphs, and single video frames. Markings made on the graphs and video frames are included in the report. In addition, we are working to incorporate report generation tools related to the decision aid tools. For example, a graph that has been highlighted by the system can be put into the report along with text describing the highlight such as "The flexion/ extension of the right knee is decreased from 25 percent of swing to 50 percent of single limb stance." (This is the text that would be generated for the graph in Figure 1.)

EVALUATION

The system has reached the beta software stage. We are beginning evaluations of both the effectiveness of the tutoring system as well as the ease of use, ease of readability, and time-saving capabilities of the report generation system and the correctness of the decision support tools.

DISCUSSION

We believe that using a computer to display and annotate data can greatly enhance the performance of the people working with that data. For reports, this includes both the person generating the report and the person reading the report. By enabling the author of the report to arrange the information on a case-by-case basis, the report can be more concise and better organized. Putting all the data on the computer also allows the report author easier access to all the information necessary to write the report.

Furthermore, providing decision support tools can increase the speed of report generation and also serve as a check on the person's knowledge by providing information that they may not have been aware of.

For tutoring, the system provides a way for the teacher to talk to the student without actually being there. The student's attention is drawn to the appropriate items visually while the student listens to the teacher's question. The student gets immediate feedback on his answer by listening to the recorded answer, and the teacher still has the opportunity to assess student progress by listening the students' recorded answers.

As the system is still under development, it is our intention to continue to expand and refine the tutoring, report generation, and decision support aspects of the system. We will continue to evaluate the various aspects of the system and make improvements where indicated.

REFERENCES

1. Inman, V.T., Ralston, H.J., & Todd, F. (1981) Human Walking, Williams & Wilkins, Baltimore.

2. Nippa, J. (1992) Development of a Computer-Based Multi-Media Tutoring System For Teaching Gait Analysis. M.S. Thesis, The Ohio State University.

3. Weintraub, M.A., Bylander, T., & Simon, S.R. (1990) QUAWDS: a composite diagnostic system for gait analysis, Computer Methods and Programs in Biomedicine, 32, pp. 91-106.

4. Hirsch, D., Simon, S.R., Bylander, T. Weintraub, M.A., & Szolovits, P. (1989) Using Causal Reasoning in Gait Analysis, Appl. Artif. Intell. 3(2-3), pp. 253-272.

5. Sutherland, D., Kaufman, K., Ramm, K., Ambrosini, D. (1992) Clinical Use of Prediction Regions for Motion Data, Developmental Medicine and Child Neurology, 34(9) Suppl#66, pp. 26-27. ACKNOWLEDGEMENTS This work is funded by a grant from NIDRR (H133E30009).

Kathy A. Johnson, Ph.D. The Ohio State University Division of Medical Informatics 2015 Neil Ave., Rm. 395 Columbus, OH 43210 johnson.32@osu.edu http://www.medinfo.ohio-state.edu/~kj 614-292-9358 (Phone) 614-292-7342 (Fax)

DR. GAIT III

Dr. Gait III - A System for Gait Analysis