Inspect3D Tutorial 1

This tutorial will show you how to use Inspect3D as a data cleaning, or quality assurance, tool. You will learn how to check for and correct faulty force assignments from all files at once, thereby removing any artifacts from the data. This is particularly useful for lab managers (or supervisors) who may not be familiar with the raw data, or collaborators that were not involved in the collection process.

Data

This tutorial uses overground walking data from four subjects. The subjects walked at three different speeds; slow, normal and fast. The data was analyzed using a pipeline which included an automatic gait event detection command. Link model based items (including right and left ground reaction forces) were created using a pipeline. This data is reflective of a typical outcome from batch data analysis in Visual3D. These data files can be downloaded using this link: Inspect3D Samples

Set the library path to the data directory

1. Click Load Library in the toolbar to open the Load Library dialog.

2. Click Browse and select the folder where the data is stored. (Choose .cmo file types when browsing to see the sample data files.)

3. Click Load button to import the data.

This step selects the path to the data you are using. If you intend to modify the data, e.g., by correcting invalid assignments, then your data must be saved in a directory where you have read and write access. This tutorial uses a folder located on the desktop. The data folder must contain .cmo and .cmx files for every subject, since without both file types Inspect3D will be unable to load the files. Each .cmo file has multiple .c3d files with the different walking speeds identified using Tags.

NOTE: After the CMO library path is set, files from the selected folder and sub-folders will be loaded and parsed. Inspect3D will show a summary of the loaded files and what they contain in the Load Library dialog. Examine this summary to make sure that the correct files have loaded.

Define queries and group signals

1. Open the Query dialogue to start defining your data groups.

1.1 In order to create a new query click the Add Query button and in the text box at the top left type in GRF, then click Modify Name.

1.2 At the bottom left click on the Add button. This will create a condition. On the right side of the screen at the top enter the condition name (R_GRF) into the text box.

1.3 There are three sub-categories that should be filled out; events, signals, and refinement.

Events: This tab allows the user to specify the desired event sequence to extract data from. For instance, the right gait cycle could be extracted using the event sequences RON, ROFF or RHS, RHS. For this tutorial use RON and ROFF.

Signals: This tab allows the user to perform signal selection. This is done by identifying a Type, Folder, Component, and Name. When completing this signal selection, the user should consider the events selected in the previous category. For instance, if the events refer to the left side, then the selected signal should also refer to the left side. For this tutorial use the following: TYPE - LINK_MODEL_BASED, FOLDER - ORIGINAL, NAME - R_GRF, and COMPONENT - Z.

Refinement: This category allows the user to refine their selected signal using two separate and distinct methods. The first method uses signals contained within the .cmo files, where only signals that meet the refinement will be extracted. For example the settings could be as follows: Type - Metric, Folder - Processed, Name - Height, Component - 0, Value Must Be - Less Than 1.8. It would result in participants with a height of less than 1.8m being included in the query. The second type of refinement uses Tags. Here the user can select which tags they wish to include in the query. By selecting multiple tags, all files with any of the tags assigned will be used. This effectively acts as an 'OR' operator. If the Use AND Logic checkbox is selected only files with all the selected tags. For this tutorial use the Refine using tag, leave the Use AND Logic checkbox blank and select SLOW, NORMAL, and FAST.

2. For this tutorial, complete a second condition in the same query except for the left side. Open the Events tab and double click on "LON" in the All Events list to move it to the Event Sequence list. Events can also be moved with the arrow buttons found between the lists. Double click on "LOFF" to move the second event to the Event Sequence list.

2.1

Because the purpose of this tutorial is to demonstrate how to validate force assignments, the groups defined here are relevant to force assignments. In this case, the signals selected are the right and left ground reaction forces. Users should bear in mind that the signal selection step requires that the user has some understanding of a “normal” signal shape, thus allowing them to identify abnormalities. If the desired signal does not appear in the combo boxes, then it does not exist in the .cmo files. In order for signals to appear, they must be created in Visual3D or our FileConverter (for other users).

3.

Open the Signals tab. Set the TYPE as LINK_MODEL_BASED, set FOLDER as ORIGINAL, select L_GRF from the All Signals table, and set COMPONENT to Z.

4. Open the Refinement tab. Select the Refine Using Tag checkbox. Select SLOW, NORMAL, and FAST, leaving STATIC unselected. Make sure that L_GRF is in the Condition Name text field at the top of the section and click Save. 4.1 In this example, no metric refinements are desired, however to assess the veracity of the force assignments, all dynamic walking conditions need to be selected. Prior to performing the query, a meaningful name must be assigned to the query and conditions. Because the aim of this tutorial is to validate force assignments, it is important to differentiate between right and left in the condition name, this is done here by simply including R and L in the names.

After defining a set of queries and calculating the associated groups, it is important to always examine the results visually by plotting them.

Visualizing and exploring your data

1. With the signals defined, the data contained within the .cmo files can be explored. The query names defined previously are shown in the query groups section (see 1). When the groups are clicked, the subjects contained within those groups are shown (see 2). The check mark beside subjects indicates that no data has been excluded for the subject. The line color beside the group names refers to the color of that group on the data plot.

2. Select all of the subjects either by using the Select All Subjects check box or by selecting them all in the Group Subjects list. Then select the Plot Subject Raw check box and click on the Refresh Plot button.

2.1 The initial exploration of the data starts by plotting the raw data for the subjects. This is achieved by clicking the check boxes 'Plot Subj Raw', followed by clicking 'Refresh Plot'. Check the appropriate boxes in order to plot the selected groups and workspaces only.

3. When there is missing data in your queries the y-axis will go to -999 999.0 and the data will be compressed to lines. This missing data can be removed.

3.1 Open the Options Dialogue and deselect the Show data not found checkbox.

3.2 When the graph is updated using the Refresh Plot button it should resemble the graph below.

3.3 More information on graphs with data missing can be found on the Troubleshooting page.

4. Use your cursor to select only lines on the graph that you wish to inspect.

4.1 With the data plotted, the user can click on single signal traces in order to examine individual curves without the other curves 'polluting' the graph. When a signal is clicked, text will appear describing the signal trace.

Verifying raw data and excluding traces

1.

Select any lines that may be a result of a 'bad' event.

1.1 In this example, with the data plotted, it is evident that some incorrect forces assignments exist. In order to exclude these assignments, they must be selected. For the obvious 'bad' events, the user can click and drag to selected multiple signals.

2.

Right-click and navigate to Exclude, at the very bottom, and click Exclude Trace (raw data).

3. Verify that your desired traces have been properly excluded.

3.1 When traces are excluded, two notable differences should appear in the 'Query Data' tab. Firstly, the previously selected, but now excluded data should not be visible on the graph (see 1). Secondly, an X should replace the original green check mark beside each subject where signal traces have been excluded (see 2). In this case, all subjects' data contained poor quality events. If unintentional exclusion occurred for a subject, you can right click on the subject and select 'Re-Include Data'.

4.

If there are events that need to be examined more closely before deciding to include or exclude them ues the following steps:

4.1 Select the events by clicking and dragging.

4.2 Click the Show Animation button.

4.3 This action opens an animation view of the selected event sequences. An animation of the traces only appears if there is enough data in the .cmo file to recreate the animation. All of the selected sequences are in a drop down box near the top of the animation frame. Using this view, the user can determine what (if any) issue exists in the data to merit an exclusion. As traces are excluded, they are removed the both the background graph as well as the drop-down box within the animation frame.

5. Once all of the desired exclusions have been made, the original data is ready to be updated.

Updating the original data

1.

Once the bad force assignments have been identified the user can update the original cmo data by clicking the circular blue arrow in the top toolbar.

1.1 When this button is clicked, a dialog will appear providing some exclusion options. The first options specifies whether a BAD event should be added to the excluded events sequences. This option should be selected if the intention is to Tag the bad data, re-open the original data, and make manual corrections one trace at a time. The second option actually removes the force assignment in the original data. This option should be selected if the user does not intend to revisit the original data after the modifications have been made (Note: by selecting the second option, re-performing the same queries as described here will produce different results).

2.

Select both the Add BAD Event option and the Remove Force Assignments option.

2.1 This will open each cmo file in the background, remove the assignments, and add a BAD tag to the event sequence. The process can take some time depending on the .cmo size. However, you can further explore your data (or get a coffee) as it is being performed. When the update is complete, the green check marks beside the .cmo files will return and the original data will be modified. For these files (which are typical), the process took around 30 seconds.

Recap

In this tutorial you learned how to use Inspect3D as a data cleaning tool for quality assurance purposes. Here, the process of loading data, defining queries, excluding data, and updating the original data are described.