Intravital Imaging of Mouse Heart Using Pulse Oximeter Triggering

We took on the task of removing motion artifacts while imaging mice and tackled many problems from a technical standpoint. Our design accomplishes its requirements quite well, and every element is needed and provides some sort of purpose for the whole device. The set up is simple, it needs an Arduino Uno, a breadboard, basic lab components, and some good old software implementation. The device is fairly practical as there are only 3 components other than the Arduino that are needed. The MouseOx pulse oximeter output is filtered through a low-pass filter. The input is tracked using the Arduino and we wrote code that converts the analog input into a digital square wave. Moreover, to ensure that we are acquiring images properly at the right time (diastolic image acquisition), we incorporated a potentiometer that allows us to manipulate the time delay between the time systole is detected, and the trigger is output to the microscope. The simplistic design of the hardware, along with the complex Arduino code is what made this project so exciting for us. The best part is that the device only has two ports, an input and an output, which makes it easy to use and highly repeatable. Furthermore, we also wrote a user manual highlighting the step-by-step process of setting up and operating the device such that anyone can use it. It’s practical, feasible, and it works, and our team is proud we could accomplish what we did.

The previously employed method to obtain images relied on physically restraining the heart using suction. This method delivered blurry images while also damaging the delicate tissues surrounding the heart. The concept of using pulse oximeter triggering to acquire images was inspired by another study [1]. They used this system to track leukocyte activity in the internal carotid arteries of mice. We expanded the idea so there are multiple physiological applications of the device. It can be used to acquire motion-artifact free images of the mouse heart and arteries located in other parts of the mouse. We did this by refining the adjustable delay feature of the device, so there is better control over detecting diastole during mouse heartbeat.

Moreover, Kubes Laboratory did not have a system for performing 3D imaging of mouse heart. We explored the Live Data Mode of the Leica software, which allowed us to set the z-slices and acquire z-stacks over time. Through this new mode, we were able to acquire 2D and 3D images simultaneously without worrying about synchronization issues.

Prior to using our design solution, the untriggered images of the heart taken without using our device were blurry and contained motion blurs. Our design solution consisted of an arduino-based triggering system, which sent out a trigger pulse to take pictures when the heart is most stable. The triggered images of the heart were then produced with minimum motion artifacts and this feature of our design to take clear images makes our design solution effective. 

In addition, our design effectively obtains the z position information of each slice for 3D images and this saves time when aligning these images into hyperstacks during the post processing. In post processing, the frame selection feature to obtain the best out of 3 images taken in the same z position and the frame registration feature to remove the residual motion artifacts further increases the effectiveness of our design to produce motion artifact-free movies or images of the heart. 

From the biological standpoint, the motion artifact-free images produced from our design solution allows for the tracking of the immune cells and is considered effective in understanding the role of immune cells during a pathophysiological event of the heart.

We used several different methods for validation of our design. The first is using an image comparison algorithm known as Structural Similarity Index Measure (SSIM). This is based on perceived structural information within the images and the idea that pixels close together have dependencies on each other. A high SSIM score means the reference image is similar in its structural information to the other images.

The next method we used was a feature presence measure. This method involves finding a feature that is present in the reference image and measuring its dimensions. The next frames are then checked for the same feature. We can calculate the percentage of frames that share the same feature as compared to the total frames.

Finally, a visual comparison of the triggered and non-triggered images shows the stability provided by the triggering circuit as well as the post processing. 

To ensure our design was feasible, we developed our system with easy to find components while keeping the cost relatively low. The components used to process the pulse oximeter signal were purchased from a local electronics shop and the Arduino was purchased online. The enclosure that was designed to house the components was manufactured in the University of Calgary Makerspace Complex and is designed for easy use and transportation within Dr. Kubes’ lab. The total cost was under $100.00 + GST.

The tools we used for post-processing were open-source plugins for the ImageJ image processing program, which was already an industry standard tool [1]. To improve the usability of the system, a user manual is provided to assist users with setup and basic troubleshooting.