Final Design

Nerve Tension Devices for Sensing Stress and Strain

Stress Device

Device Overview:

The following CAD model represents the final iteration of the stress device. The breakdown of its components can be found further down, but the general decision for the final design was to maintain the "pull" mechanism as it was able to interface best with both sutured and severed nerve configurations. Overall, the dimensions were adjusted to better conform to the actual shape and size of a common surgical retractor. The final maximum dimensions (or the dimensions when the device is fully extended) are 284 mm (11.2 inches) long and 88.9 mm (3.5 inches) wide. The device maintains the general concept of attaching to the nerve with the end piece, and pulling a contact piece towards a force sensitive resistor in order to register a force reading.

The final iteration was printed from MJF Nylon 12, a common prototyping material in manufacturing as well as a common material used in medical device prototypes. The polished finish allowed for a solid frictionless movement of the translational piece, and the material came with tolerancing values that allowed for more accurate interfacing of the parts.

Key Components:

The diagram depicts a simple breakdown of the key players in the device. All orange titles represent components that were changed from the previous iteration of the device (which can be seen in the multimedia page) while the black titles represent components that have remained the same throughout the entire design process.

Hand Interface:

• The hand interface portion was designed in order to maximize the ability to pull on the device. While most surgical tools often have regular cylinder handles, some retractors incorporate a curved handle with a pull tab in the middle. This design was used as inspiration for the device's handle, which was made larger from the previous iteration in order for a hand to properly grab onto its entire surface area. Ultimately, the pull tab was beneficial in using the device, as it gave the user flexibility in terms how they would have liked to hold onto the device.

LCD Screen:

• The LCD screen is a color changing 16 x 2 LCD screen from adafruit. The form factor was designed to house the LCD screen on the front of the device as a way to properly display stress measurements. While it displays the proper force readings on the screen, it also presents the data using the color changing technology and changes color depending on the range of forces experienced by the device (for example, glowing red if the values are too high, signaling that the user should lessen the pull). This was an improvement on relying on the Arduino serial port to display the force values, and helped the device become more handheld.

Contact Piece:

• The contact piece maintained it's circular form factor, with the slight adjustment of adjusting the through holes for easier assembly. It's purpose is attach to the rod attached to the translational piece, and serve as a constant shape that comes into contact with the force sensitive resistor. This ensured the same surface area was being pressed onto the sensor, and with the help of guide rails and nuts, the piece was able to move on one axis and always remain parallel with the FSR.

Translational Body:

• The translational piece's purpose was to connect the body containing the FSR to the end piece that interfaces with the nerve. In the most frictionless manner possible, it moved in the same axis as the nerve, pulling the contact piece towards the FSR. This part did not go through any dramatic changes, as edits were only made to fillet and smoothen the part to remove and sharp edges as well as integrate a slot for easy entrance of the end piece.

End Piece:

• The end piece followed the same form factor as a surgical retractor, allowing for a gentle interfacing with the nerve. The latest iteration added a square slot piece with a whole for a threaded thumb screw for easier combination of the parts.

Nuts and Rods:

• Originally, the entire device relied on press-fit components to keep everything together, but this proved to be an issue as the tolerancing was not consistent, resulting in heavy sanding to be necessary for the parts to fit. By using nuts, screws, and rods alongside a material that came with a tolerancing handbook, the device now had more reliable dimensions to plan around and use, resulting in a last iteration that printed and fit together immediately without any need for sanding.

Arduino Uno, Power Source:

• An Arduino Uno was used to connect to the FSR and process the values to be translated into force. The Uno is a great prototyping microcontroller, and its ease of use overall familiarity to the team made it a clear choice. Originally, the device purely relied on the power from the connection to a desktop computer, but the latest iteration now uses a battery pack of 8 AA batteries in order to power the device. This configuration allows the device to be completely handheld.

Force Sensitive Resistor:

• A Pololu 1.5 inch square FSR was used as the sensor of choice. The decision was made based on it being the most realistic option in being able to test a proof of concept, as well as overall familiarity with the sensor as it was very Arduino compatible. While it worked in the realm of a proof of concept, issues regarding accuracy and sensitivity ultimately result in a future design recommendation of using a piezoelectric sensor instead.

Final Results:

Gifs of final rubber band tests, pulling on a sutured (left) and severed (right) rubber band configuration. Test done over grid lines in order to be able to analyze extension lengths for strain values.

In order to generate a force-strain curve, the final testing included pulling on sutured and severed configurations of rubber bands over gridlines (4x4 set of gridlines represented one square inch). A video was taken of the pulling in an overhead view, allowing for video playback for finding the length that the rubber bands had extended corresponding to the specific force readings (this was done by observing each "stop" that was made in increments during pulling).

Ultimately, the device showed that the sutured configuration maxed at a force reading of 2.7 N, while the severed had maxed at a reading of 2.3 N. The main takeaway of this final test was the continued success of the translational movement as well as the improved calibration of the FSR. The device was now able to detect more specific readings at lower force values, and the results showed that this window could be shrunken down to max around 3N.

(While the device has been referred to as a stress device, it currently only displays force values. This is something that has been discussed with the sponsor, and he has approved of the fact that we have provided the model to derive stress from these force readings, which can be seen in the final report. It is called a stress device for naming convention purposes.)

Design Recommendations:

A full breakdown of all design recommendations can be found in the final report page, but in terms of an improvement that would benefit the device as it is in dramatic fashion, it would be to switch to a piezoelectric sensor (or anything piezo-related for the matter). From the results, it can be seen that a smaller, more precise window is required for these forces, rather than the wide ranging 0.2-10N range of the FSR. In theory, no changes would have to be made to the form factor, just a simple swap of sensors in the circuit would reap benefits to the ability of the device.

In terms of how this could affect the future iterations of the device, a piezoelectric sensor is inherently smaller than the 1.5 inch FSR. Smaller internal components are most welcome in medical devices, as it would help reduce the dimensions of the form factor. The most ideal version of a device like this would be to replicated the exact dimensions of a surgical tool, as it helps provide extreme familiarity to surgeons as well as allow for more comfortable navigation of the device around nerve tissue.

Strain Device

Device Overview:

The following is the CAD model of what the theoretical strain measurement device would look like. While the CAD model of the strain device was essential to the overall design process, the team placed a higher priority on the actual development of the digital image correlation software via Matlab. The goal was to develop an accurate data acquisition system that could extract images from a video file, transfer those images to Matlab, and use the surface markers present in the images to digital track a nerve's displacement and calculate the strain values.

The model consists of a camera at the top in order to record the surgical field. There is a hook apparatus, upon which the nerve rests on, that is also used to align a ruler with the nerve. The surface of the nerve is marked with 2D markers, drawn with a biocompatible surgical pen.

CAD model of handheld strain measurement device (not to scale).

While the team was unable to 3D print a prototype of the strain device, an experimental setup was built in order to mimic the device and test out the custom Matlab software. The components and the justifications for each design decision are explained below in the "Key Components" section.

Key Components:

Nerve: In order to simulate the viscoelastic behavior exhibited by nerve tissue, the team first chose to use rubber bands of assorted sizes for testing purposes.

Video of first iteration rubber band testing.

3D spherical pins were pushed into the rubber band in order to serve as preliminary markers and calibrate the digital image tracking software. In addition, circular markers were drawn on the rubber band surface to test whether they would be correctly identified.

In this first round of testing the Matlab script was only able to identify the 3D spherical markers and not the 2D ones. However, the team decided that 2D markers were required for the sake of biocompatibility and minimizing the risk of damage to the nerve tissue during surgery. Furthermore, using 2D markers drawn by hand made the measurement procedure simpler and more feasible for surgeons, as they typically already receive pens to mark tissue in their surgery kits.

For the second iteration, cut O-rings were used to mimic nerves due to their cylindrical shape. The rubber bands also proved to be too elastic, and so O-rings offered a more controllable stretch.

A severed black O-ring with white circular painted markers was used for a strong contrast.

Ruler: As seen in the image above, a ruler was used in the same plane as the nerve for alignment purposes and in order to physically track the movement of the painted surface markers. The ruler readings are also meant to make the procedure more consistent and help relate physical readings to the digital strain values given in pixels. Along with pens, surgeons are also provided rulers in their surgery kits.

Video Camera: While a high-definition video camera should be used in practice, the team found that an Iphone XR camera sufficed for testing purposes and provided good quality videos for each trial (4K recordings at 60 frames per second were obtained).

The following is an image of the final experimental setup created by the team.

Complete testing mechanism consisting of an adjustable phone mount and the nerve testing platform.

Final Results:

Multiple trials were conducted by stretching the O-ring by hand while recording was in progress.

Example of video recording obtained from the second iteration testing procedure.

By manually adjusting the digital radii of the markers and the sensitivity values in Matlab, the team was able to successfully detect all of the markers and track the marker movement by using the relative displacements of the centers of each circle. The strain values were then easily calculated from these displacements.

Example of accurate marker detection (numbered 1-6 from top to bottom) and strain analysis performed by the Matlab software.

To see the full dataset, including all of the obtained displacements and strains, please see the Programs/Code tab.

Sources of Error & Future Considerations:

Markers: As mentioned previously, the team chose to use black O-rings with white circular markers for a strong contrast. However, for the sake of mimicking a real surgical field, a viscoelastic material that is of a pink or red hue would serve as a better model of a true biological nerve. Furthermore, while circular markers were used for the sake of accuracy and simplicity, future versions of the Matlab software must be able to detect markers of an arbitrary shape that are drawn by the surgeon.

Examples of arbitrary 2D surface markers that should be identifiable by the digital image correlation software.

Accuracy: While the ruler component of the device acts as a helpful visual guide, the readings (in mm) still must be calibrated to the displacement values found by Matlab (in pixels). In other words, the strain within the nerve still needs to be found relative to the ruler markings.

Sources of Error: One prominent source of error is the torsion present in the O-ring as it is being stretched. While the team attempted to keep the O-ring as straight and flat as possible, there was some experimental error as the small amount of slack in the O-ring caused it to twist as it was pulled. This twisting caused some unexpected changes in the circular markers' radii, which influenced the software's detection accuracy. The software is only able to find circles within an expected range of radii defined by the user. Therefore, the circles drawn by Matlab may not represent the true shape or position of the deformed markers. This problem could be eliminated by modifying the software to use edge detection in order to recognize the boundaries of markers rather than relying on the identification of purely circular markers.

Design Recommendations: Apart from the remaining issues surrounding the shape of the 2D markers and the calibration of the system, another improvement that could be made to the device involves the implementation of a circuit component. This circuit should be used to link the Matlab script to an Arduino microcontroller via Simulink. By doing so, future investigators will be able to obtain strain values in real time rather than exporting the video files to Matlab for post-processing.