Dexterit-E 3.6 has been a monumental effort for the team at BKIN and we are very proud of what we’ve been able to accomplish in this “landmark” release.
Dexterit-E 3.6.1 is the first general release for Dexterit-E 3.6. The release notes provide extensive detail on the enhancements. As with all releases, we strongly recommend you review the upgraded User Guides under Documentation. We’ve summarized below our key updates:
KINARM Standard Tests Changes
Elbow Stretch Test
The newest KINARM Standard Test measures passive elbow motion; it is only available for the KINARM Exoskeleton Robot. We have named the test the Elbow Stretch Test. The task is entirely passive in that the subject does nothing other than relax as the robot moves their arm. The robot attempts to moves the subject’s elbow across a known angular distance within one of two specific durations. The intent of the task is to quantitatively measure elbow stretch response in flexors and extensors. The task comes with a built in report and normative comparisons.
All KINARM Standard Task reports now have an overall “Task Score” measure for each completed task. The Task Score is a measure of how far from the “normal” a subject’s behaviour is for a given Task. The Task Score is derived from the root-mean-square (RMS) of Z-scores of all applicable parameters in a KINARM Standard Test Report. The Task Score is always >= 0. A Task Score of 0 means that the subject had exactly the mean score for each parameter. The units are equivalent to standard deviation units, e.g. a task score of 1 means that 68.4% of the healthy population will have a task score of <= 1 (i.e. same percentile as ±1 standard deviation of a normal distribution). See the Dexterit-E User Guide for more details.
A related metric, the M-score is also calculated, but is not part of the Report. The M-score is output to the CSV files alongside the Task Score. Whereas the Task Score is based on the RMS distance of parameter Z-scores, the M-score is based on the Mahalanobis distance of the same parameter Z-scores. The M-score has the capacity to detect patterns of abnormal behavior that are not represented in individual parameter Z-scores.
Improvements to Normative Modelling; Normative Databases Are No Longer Split by Sex or Handedness
Significant changes were made to BKIN’s algorithms that are used to create the models of normal behavior used to generate Z-scores and typical ranges. These changes improve the robustness of our Task Score and parameter Z-scores. We provide a summary here of the reason for the changes, but also direct you to Dexterit-E User Guide Section 16.1 for further details on the methods used. We strongly recommend that you take advantage of the improved robustness of KST analysis in Dexterit-E 3.6.1 by regenerating reports and/or exporting CSV results.
Problem being addressed: In prior versions of Dexterit-E, for task parameters that were found to have sex or handedness differences, normal data sets were split accordingly and independent normal models were generated. We have identified that when independent normal models are generated this way, there can be substantive differences when comparing large Z-scores (i.e., |Z| > 3) derived from these models (e.g. between males versus females, or left-hand versus right-hand). Please see the Dexterit-E User Guide for information.
Note that there is no impact when comparing Z-scores derived from a single normal model (e.g. comparing two male subjects, or a single subject at different time points).
Current Solution: The most immediate means to address the above problem was to stop splitting our normal data based on sex and handedness.
Long-term Solution: In the future we will investigate additional methods that allow us to account for the effects of sex and handedness without re-introducing the problem described above.
Impact of the Change: Many parameters have negligible changes between Dexterit-E 3.5 and 3.6 in the model used to produce Z-scores. However, some parameters can have noticeable effects. For example, for Z-scores of ~1.64, the mean of the standard deviation of Z-score changes from Dexterit-E 3.5 to 3.6 across all parameters was 0.16 (i.e. typically the change will be less than 0.16).
For the present release, we believe the increased robustness and confidence provided by combining the previously separated sex databases, thus effectively doubling or quadrupling the sample size, greatly outweighs the impact of the small shifts between sexes.
We have added support for MATLAB R2015a SP1. Unfortunately, R2015a SP1 requires a patch from BKIN in order to work with Dexterit-E. As part of Dexterit-E 3.6.1, this patch is automatically installed as part of the TDK.
When using R2015a SP1, models will run at 4 kHz, whereas in all previous versions of MATLAB models will still run at 2 kHz. For example, if you have custom designed a digitial filter that assumes a 2 kHz update rate, you may need to determine new filter coefficients.
R2015a SP1 is required to run KINARM Exoskeleton Labs installed in 2016 or later.
- MATLAB R2007b (xPC Target 3.3) is no longer supported as of Dexterit-E 3.6.
- Dexterit-E 3.6 will be the last major version that will support MATLAB R2010a and xPC Target 4.3.
EtherCAT (Ethernet for Control Automation Technology)
EtherCAT is the control mechanism used with the new version of the KINARM Exoskeleton Lab. It replaces the PMAC motion control card which KINARM Labs have used up until this point. EtherCAT gives us the ability to add many new features like:
- Fully digital control of the motor system
- Status indicators on the Status Bar showing emergency stop statesSupport for absolute encoders
- Direct recording of applied torques (no DAQ card is required)
- Torque limiting to avoid motor overheating without disabling the drives
- Integrated digital I/O ports (no DAQ card is required)
- Support for robot mounted calibration buttons and therefore a simplified calibration routine for the new Exoskeleton Lab
- Extensive enhancements to error reporting
EtherCAT required extensive changes to the KINARM Labs’ electronics, thus is only available with new Labs.
For more details please see the release notes.
As with all releases, we welcome your feedback at firstname.lastname@example.org !