KB-Perry as a tool for Kirby-Bauer tests
If you ever feel like a plastic bag drifting through the wind while conducting Kirby-Bauer disc diffusion tests, we recommend you try our measurement system: KB-Perry. We designed, 3D printed, and tested KB-Perry, intending to make Kirby-Bauer measurements more precise. Most current protocols for this disc diffusion test recommend simply using a ruler to measure the zone of inhibition (1). But because a ruler can only approximate the zone inhibition sizes to the nearest tenth of a cm, this presented a problem for us. Statistical analysis revealed a significant difference in measurements for the same treatment type, and we could not confidently eliminate measurement as a source of variability. Our methodology was not precise, and we needed a standardized method to collect data. KB-Perry is here to fix that.
The hardware component of the project came about to solve a single critical issue: the data being processed must be standardized, consistent, and controlled. Given inconsistent or erratic data, the resulting information is difficult and time-consuming to parse through. This is consistent with a common principle in most statistical models: garbage input leads to garbage output.
Thus, the hardware component of the KB-Perry subproject emerged as a crucial part of the overall system, allowing for consistency when analyzing images of Kirby-Bauer tests. To tackle these issues, the hardware was designed with two main design objectives: consistency and flexibility.
Consistency was the priority when designing KB-Perry, and was the central priority guiding every design decision. This manifested itself through the creation of many slightly over-engineered, extra-durable components. Flexibility, on the other hand, took a backseat for many design decisions, typically implemented after a system was demonstrated to work. Though the KB-Perry hardware originally took a single-size-fits-all approach, we later prioritized flexibility in order to expand the reach of our system, both in our lab and in the labs of others. The full specifications of the system can be seen in the table below:
|Minimum Dimensions||Length x Width x Height, in millimeters, rounded to 3 significant figures.||102 x 100 x 135|
|Maximum Dimensions||Length x Width x Height, in millimeters, rounded to 3 significant figures.||102 x 1231 x 180|
|Minimum Phone Dimensions||Length x Width x Height, in millimeters, rounded to 3 significant figures.||903 x 70.0 x 0|
|Maximum Phone Dimensions||Length x Width x Height, in millimeters, rounded to 3 significant figures.||1804 x 93.21 x 30.04|
|Required Hardware||Any non-3D printable parts that are not included in the KB Perry system and must be produced or procured independently.||3 x 1/4"-202 3” Threaded Rod1
3 x 1/4"-202 Wing Nut
1 Based on the length of the threaded rod.
2 Default specification, design can be easily modified to fit other rods.
3 Measured from bottom to edge of camera, to ensure system is unobscured.
4 Untested, theoretical maximum, based on the inherently false assumption that the phone is perfectly uniform and balanced right on the center of mass. Use at your own risk and at your own discretion.
Our hardware consists of 4 main components: the legs, the sliders, and the two bridge pieces. Each component serves an important role in the overall function of the project. Starting from the ground up, we have the legs. The legs of the structure serve an important role in supporting the bridge by providing a foundation to build off of. The legs are large, hollow, rectangular prisms with a wide “foot” on the bottom, providing a larger area of contact and allowing for greater frictional forces, helping to prevent unwanted movement. The legs feature holes with 15 mm spacing (center-to-center), allowing for the height of the system to be adjusted in 15 mm increments by moving the sliders. At its shortest, the system has a height 12 cm above the surface, while at its largest, the system stretches to 16.5 cm above the ground, more than sufficient for the vast majority of use cases.
The sliders are equally simple to the legs: they are smaller rectangular prisms with a single hole at the bottom to interface with the legs. The sliders are designed to, as the name suggests, slide up and down the inside of the legs to allow for height adjustment to occur. This happens by lining up the hole on the bottom with a corresponding hole in the leg and inserting a peg to fix them together. The sliders come in two variations, each with a different “head,” designed to interface with the respective bridge pieces.
Finally, the bridge is the part of the system responsible for holding your phone in the correct place. The bridge consists of two main pieces: the larger left piece, serving as the primary hub for all the parts and the smaller, “passthrough” piece on the right, responsible for moving down the rods to allow for variable width control. The bridge components fit into the leg assembly through the circular extrusion on each slider’s head, which corresponds to a circular indentation in the bridge pieces.
The accessory hardware is responsible for ensuring effective interfacing between the components. There are three pieces of hardware necessary for the system to function as intended; the threaded rods, serving as strong rails for the passthrough bridge component to move on; the wing nuts, allowing for the width to be locked and unlocked as desired; and the peg, which, as mentioned earlier, allows for the variable leg height to be locked in. The threaded rods and wing nuts used in this project were procured at a local hardware store and are size 1/4"-20 due to Canada’s tendency to use imperial units for construction hardware. This sizing, however, can be easily modified in the CAD files by changing the radius of the holes on the bridge components. The peg, on the other hand, is 3D printable (though it can be independently procured if desired) and is a very basic, 0.9 cm x 4.5 cm cylinder with a head to serve as a stopper, designed to slot in and out of the holes in the leg assembly.
Figure 1. KB-Perry in action.
The parts, while designed to work without any user modifications, do come with a few challenges that must be noted before usage. To facilitate the smooth movement of parts within the system, the parts have huge margins and tolerances. This allows for the parts, as is, to be printed with any moderately calibrated 3D printer and function without excessive friction or needing to force parts in. However, these huge clearances also create a large amount of play in the system, and the connections between parts may be loose and wiggling may occur. Though this may not be a huge issue in most circumstances, it is recommended to tighten up the margins of any loose parts to create a tighter, more secure fit.
The necessary 3D models and the engineering diagrams are available for download as .stl files in the appendix
Instead of measuring the zone of inhibition - which stretches from the outermost portion of the inhibition ring, across the disc, and to the other end of the inhibition ring - with a ruler, we decided to utilize ImageJ’s measurement tool. We simply uploaded our image files, taken with our KB-Perry hardware, onto ImageJ Fiji. We then opened up the measurement tool and stretched it across the zone of inhibition 3 times. This was to account for any misshapen circumferences, as the zone of inhibition was not always perfectly circular. All three measurements were recorded along with an average. Using this method yielded much more precise results with a smaller margin of error, as 1mm was approximately equal to 3.3pixels - or in other words, a previous measurement which could have been 120 mm ± 10 mm depending on how accurately we held our ruler against the plate, would now be exactly 393 pixels ± 1 pixel.
Using KB-Perry, we conducted three new rounds of Kirby Bauer disc diffusion tests with nisin against B. subtilis in comparison to thymol, phenol, and water as a negative control. As an additional parameter, we measured the effectiveness of each treatment over time. For our August 17th round, we only had one biological replicate, but for our August 30th and September 21st rounds we decided to increase the number of data points by having two biological replicates. Each petri dish was plated with B. subtilis at time point 0:00 hours, and treatments were applied immediately and then every 2 hours for a total of 6 hours. Our results were as followed:
Figure 2. Kirby-Bauer measurement disc diffusion measurements for three rounds of data collection (August 17th, 30th, and September 21st) using KB-Perry. Nisin was compared against thymol, phenol, and water (as a negative control) for its effectiveness against B. subtilis. Errror bars represent SEM.
Notably, we had several missing measurements. Plates that had an unclear zone of inhibition formed around the disc were discarded because we knew including the measurements would introduce unreliable points. Our explanation for the unclear zones was due to a lack of lawn growth formed by B. subtilis.
With our new and more precise dataset, we asked ourselves the following questions:
To answer our first question, we ran a two-way ANOVA with time and treatment type as our variables, and compared the datasets from two different days (August 30th and September 21st). Our results showed that there was a significant difference (p< 0.001) between the two days. Because KB-Perry provided us with a more reliable method of collecting our data that was both simple and effective, we knew that any statistically significant difference between the same treatment on different days would be due to biological variability. In the future, it would be useful to keep track of the density of cells in our overnight tubes of B. subtilis to ensure a concentrated volume was added to the plates for lawn growth, along with when the tube was made to ensure the viability of the cells.
To answer our second question, we ran another two-way ANOVA with time and treatment type as our variables, and compared all treatments at each time point using the entire dataset. Our results showed that there was a significant difference (p< 0.001) between time points and a significant difference (p< 0.001) between treatment types. Further analysis such as a Tukey post-hoc test is required to find out which treatments and time points specifically had a difference.
Using KB-Perry we validated that our Kirby Bauer tests will produce more reliable zone of inhibition measurements, allowing us to make better conclusions about our data sets.