Team 5: Biomechanical Analysis of Proximal Femoral Varus Osteotomy Fixation
Dina Eloseily, Michael He, Lauren Losin, Neelesh Mupparapu, Grace Powers

Advisors: Dr. Sean Tabaie and Dr. Martha Wang

In pediatric Cerebral Palsy (CP) patients, hip dysplasia may develop due to spasticity. This often requires a corrective gap osteotomy of the proximal end of the femur to varus (<130°) in order to improve hip congruity and correct deformities. When conducting an osteotomy, the bone is stabilized using a fixation plate, but pediatric orthopedic surgeons have identified flaws in these plates. Plate complications can lead to pain and potential reoperation for CP patients, which must be minimized. The purpose of this project is to evaluate the maximum compressional and torsional load capacities of two osteotomy fixation methods, the locking cannulated blade (LCB) plate and the locking proximal femoral (LPF) plate, in order to compare their biomechanical abilities in children suffering from CP. The wider goal of this study is to gain a better understanding of which fixation method performs the best mechanically so that the optimal surgical approach for proximal femoral varus osteotomies can be taken moving forward, thus improving the clinical outcomes and prognoses for pediatric CP patients with hip dysplasias. To achieve this goal, the team was able to perform osteotomies on sample sawbones, uniformly pot both the proximal and distal ends of the sawbone in epoxy resin, and validate the compressional and torsional potting setup. 3D mechanical simulations in CAD showed that the femur experienced more stress than the potting mechanism for both compression and torsional testing, indicating that the bone rather than the potting mechanism will fail first during MTS testing. Mechanical testing with an MTS machine proved that, when the sawbone was loaded in compression, the femur failed before the potting. CAD modeling and initial MTS validation proves that the epoxy potting setup is sufficient in meeting the potting design criteria of accurately mimicking physiological loads in vivo, resistance to heavy loads, and full stabilization during potting and for experimental testing. Due to a delay in shipment of the physiologically relevant 4th generation sawbones, potting and testing was only performed on spare sawbones. In the future, the team plans to validate the torsional potting setup via mechanical testing, perform osteotomies on the proper 4th generation sawbones, perform MTS compression and torsional testing, and conduct a statistical comparison of plate strengths.

Team 8: Tele-Neurorehabilitation After Stroke: Validation of LiDAR for Human Pose Estimation
Maximillian Berenschot, Sabrina De Nicola, Lauren Hoorens Van Heyningen, Noriana Jakopin, Jennifer Song 

Advisors: Dr. Kim Stroka, Dr. Richard Macko and Dr. Charlene Hafer-Macko 

Stroke is the second leading cause of death worldwide, claiming over 130,000 lives in the United States alone each year. Survivors of stroke often suffer from a multitude of impairments, such as decrease in muscle strength, inability to balance, and an overall loss in mobility. These issues are typically treated through physical therapy in the presence of a medical professional; however, this is not accessible to many stroke patients due to high costs, difficulties with traveling, and additional restrictions as a result of the ongoing COVID-19 pandemic. A possible solution may lie in a digital neuro-telerehabilitation program that provides physical therapists with a platform to create personalized therapeutic programs for stroke patients. LiDAR, or light detecting and ranging cameras, record three dimensional data which is then used to construct a human figure. LiDAR will be validated against the gold standard Optogait system as a potential alternative for motion capture to perform movement analysis. By recording the march-in-place exercise with the LiDAR camera system and processing the data with Openpose software, the X & Y coordinates of 25 key skeletal points were generated for every frame. The coordinates were then processed through a Python script to precisely track movements and calculate cadence, flight time, and contact time of the foot. The percent error between the LiDAR system and the Optogait system was 9.9%, 19.3%, and 8.9% for flight time, contact time, and cadence, respectively. Results show that the 2-dimensional data alone is slightly outside of the acceptable error when compared to a validated system. This points to the importance of mapping the depth data to the X & Y coordinates to improve accuracy, which is in progress and is the immediate next step. Though the telemedicine platform is an external system, there are several risks to consider. It is imperative that the virtual nature of the patient-physician relationship does not diminish the quality of care that the patient receives. To solve this, future steps will focus on the development of a feedback scheme to facilitate communication and track the patient’s progress. In terms of privacy, there will be a secure connection that can only be accessed by the two consenting parties. If implemented, the LiDAR system will positively impact the lives of stroke patients by offering them an accessible route to recovery. 

Team 9: Rapid Automated Cell Expansion and HyperspectraImaging Monitoring for Improving the Production of Cell Therapy Drugs
Jess Conway, Anya Filatova, Amy Musser, Anthony Neri, Yiding Yuan

Advisors: Dr. Maurizio Cattaneo, Dr. Yang Tao, Anjana Heva, Chiao-Yi Wang, Callie Weber, Dr. William Bentley, Dr. Chen-Yu Tsao 

Personalized, cell-based therapeutics are emerging as effective and exciting treatments for some of the most severe diseases. These innovative therapies, such as adoptive cell therapy (ACT) for cancer treatment by tumor-infiltrating T-lymphocytes (TILs), tissue regeneration via induced pluripotent stem cells (iPSCs), and the use of cell-produced therapeutic proteins incorporate a patient’s genetic code into their method of attack. Industrial cell expansion via a bioreactor is a crucial step in the production of cell-based therapeutics. Constraints to this process include the transfer and isolation of cell samples to a separate facility and manual intervention with highly invasive cell monitoring probes by a trained specialist. This increases the production time, cost, and risk of contamination, raising the price per patient for these potentially life-saving therapies. The goal of this project is to increase the efficacy and therefore the availability and affordability of cell-based therapeutics by automating glucose concentration monitoring in bioreactor cell cultures. Our team combined a short-wave infrared (SWIR) hyperspectral camera with a flask bioreactor and a deep learning algorithm to non-invasively determine glucose concentration in the media. CHO cells were seeded and grown in a reusable spinner flask and then were pumped by a magnetic drive pump through an Artemis Biosystem’s VHU (viral harvesting unit), where the cells are maintained allowing the perfusate to flow through a flow cell and imaged by the hyperspectral camera and pumped back into the flask bioreactor. The hyperspectral data was fed into a supervised linear regression (ordinary least squares/OLS) ML model in Python that achieved a R2 score of 0.92062 for training data and 0.89065 for testing data. Removing an outlier at 9.45 mg/dL improved these values to 0.95978 and 0.94009 for training and testing data respectively. PLS regression achieved a R2 score of 0.95863 for training data and 0.94848 for testing data, which is about the same performance as OLS. Based on these models additional media is supplied via a peristaltic pump to achieve the determined desired glucose level. Future applications of this process can enable the industrial culture of multiple cell types and imaging via low-cost infrared spectroscopy devices, further increasing the availability of cell-based therapeutics. This non-invasive method may also be applied to other parameters of a bioreactor that need to be monitored as well, reducing the amount of attention that the cell expansion phase requires. By reducing contamination risk and increasing the efficiency of bioreactor cell expansion, personalized therapies have the potential to become a more feasible method of fighting various diseases.

Team 10: Infant transanal rectal suction biopsy device with external stabilization
Noemi Gonzalez-Martinez, Varunaa Hemanth, Anish Kakarla, Yaniv Kovich, and Tia Puskar

Advisors: Dr. Helim Aranda-Espinoza and  Dr. Louis Marmon 

Hirschsprung disease is a congenital condition where missing nerve cells in the large intestine cause issues passing stool. The lack of nerve cells in the muscles of the colon reduces stimulation in the gut which helps stool move through the colon and can cause bowel blockages. In order to diagnose the disease, an infant transanal rectal suction biopsy device is “blindly” inserting the instrument into the infant’s rectum. If the results of the biopsy diagnose the infant with Hirschsprung disease, then surgery will be required to bypass the part of the large intestines lacking nerves. Currently, the RBI2 Suction Rectal Biopsy System is what is used to conduct infant rectal biopsies but when clinicians use this device is it difficult for them to stabilize the probe. Proper biopsy location is critical to ensure that the target tissue is obtained and a proper diagnosis can be made. The current method lacks a reliable way to visualize where the piece of tissue needs to be taken, relying on a physician's vision alone. An accident could result in a tissue biopsy in the wrong adjacent areas such as the skin, anoderm, or pectinate area. In the past, incorrect biopsies have led to rectal hemorrhages and, in severe cases, death from gangrene of the buttocks. The product we are designing is an external stabilization disk for use in conjunction with an infant transanal rectal suction biopsy device to help the clinician visualize how far up the rectum the probe is. This product would be an add-on to the current biopsy device. The device should be able to move up and down in order to change the desired length entered into the rectum. Initial prototype will be designed by printing the clasp out of Polylactic acid (PLA) filament using a 3D printer, would be the best method. Rubber inserts will also be purchased to complete the prototype. The testing of the assembled device includes stabilization and mobility of the add-on product. Stabilization testing showed that as the force applied increased the displacement increased as well. From our criteria and the results we decided on making the prototype 2.5 cm in length. Using a two sample t-test, the results from the grid testing showed that there is a significant difference in stabilization when the disk is added to the probe compared to the probe itself. Our device has a positive ethical device because it benefits the infant and the clinician by giving more accurate diagnostic results. 

Team 11: Deep Learning for Automated Expert-Level Diagnosis of Sports Injuries on Knee MRI
Anagha Rama Varma, Keyona Curry, Anna Fraser, Janam Shankar

Advisors: Dr. Xioaming He and Dr. Paul Yi 

ACL tears account for approximately 50% of all knee injuries (Joseph, 2013). The diagnosis of ACL tears often relies on the assessment of an MRI image of the knee by an experienced radiologist. However, in underserved communities where such a provider may be unavailable, diagnosis may be delayed, which increases the potential for further complications. In order to address this disparity, we have developed a 2D convolutional neural network deep learning model which can diagnose the presence of an ACL tear from an MRI image input and can predict the presence of an ACL tear with 82.7% accuracy. This model was based on the Stanford MRNet model, which was modified to exclusively perform ACL tear diagnosis and trained on an expanded data set of 1947 knee MRI images. We have additionally developed a user interface to facilitate ease of use in a clinical care setting. The availability of this technology will increase access to accurate diagnosis for underserved communities and can also act as a method of diagnosis confirmation for clinicians, increasing the likelihood of accurate and prompt diagnosis of ACL tears and reducing the risk for complications.

Team 12: Designing a Strain Gauge for Laparoscopic Surgery
Lauren Becker, Justin Ciemian, Sarah Covert, Scott Fan-Loh, Keren Sneh

Advisors: Dr. Brian Blair and Dr. Louis Marmon 

Laparoscopic surgery is a type of minimally invasive surgery, specifically for abdominal procedures, in which long-handled instruments are inserted into small incisions in the abdomen so surgery can be performed without having to fully open the abdomen. Laparoscopic approaches have been shown to greatly reduce surgical complications and recovery time. However, due to the significant length of the instrument, surgeons experience a lack of tactile feedback during surgery. This absence of tactile feedback makes it difficult for surgeons to assess how much force is being applied to tissues or when suturing, which can lead to dangerous mistakes. The objective of our project is to create a device that reliably measures the force applied by the jaws of a laparoscopic instrument and displays this force to the surgeon. The successful implementation of this project would positively impact surgeons and their patients. Surgeons would be more confident during their procedures, resulting in fewer mistakes being made. In order to achieve our objectives, we chose to place our strain gauges on the internal shaft of the laparoscopic grasper where they will measure the deformation of the bar when the jaws are opened and closed. Our 3D printed prototype did not deform enough to be detected by the strain gauges, however our design using the internal shaft of the original laparoscope had successful results. Using Arduino software, we were able to measure the change in voltage between the Wheatstone bridge when the internal shaft experienced bending.