How wearable sensors can be used to monitor patient recovery following ACL reconstruction; The Impact of Machine Learning on Medical Diagnoses

The technical capstone and STS research paper included in this portfolio are interrelated in their themes. The technical capstone project focuses on leveraging patient EMG data to improve recovery following ACL surgery, and the STS research paper centers on the impact that machine learning solutions for patient diagnosis will have on patient outcomes. Both projects are healthcare related and involve using data driven solutions in the medical field. In addition, both projects involve machine learning. To sum up the relationship between the two projects simply, both projects are rooted in the desire to improve the quality of care provided to patients by leveraging data driven solutions.
Anterior Cruciate Ligament (ACL) reconstructions are among the most common sports medicine procedures performed in the world. Over 100,000 patients in the United States annually elect to have ACL reconstruction (ACLR) in hopes of returning to pre-injury levels of activity. In the first two years following an ACLR, patients are at their highest risk for re-injury to both the repaired and contralateral knee. The overall incidence rate of an ACLR patient having to go through a second repair in 24 months is six times greater than someone who has never had an ACL tear. Early detection of functional deficits is vital to optimize post-operative rehabilitation and to restore normal movement patterns in patients. The decision of when to return to unrestricted physical activity or competitive sports has come under much scrutiny due to the lack of evidence-based criteria that have sufficient predictive value. Current methods of detection require unconventional movements which cannot be done in the early stages of recovery for fear of damaging the newly repaired ligament. The need for a precise, objective, and whole-body approach to movement evaluation is essential for the health and safety of patients recovering from ACLR. The objective of this research was to leverage sensing technologies to monitor patients post ACLR and form an understanding of how body sensors can be used to aid medical decision-making regarding rehabilitation progressions. In our study, patient data, extracted from the sensors during several functional assessments, was used for multi-level analysis to extract features indicative of mobility and muscle activation. In the conclusion of this pilot, we have identified key features effective in determining patient health post-ACLR and implemented these into a machine learning model to estimate the efficacy of lower-body wearable sensors as a means of assessing patient recovery.
In a field as old as healthcare, machine learning is emerging as an innovation capable of revolutionizing the way patients receive and doctors administer care. However, little is widely accepted about the impact of incorporating machine learning systems on patient outcomes. This STS research paper is concerning the use of machine learning in medical diagnoses. Specifically, the research question for this STS research project is: how has the introduction of machine learning as a tool in medical diagnosis impacted patient outcomes? Actor network theory is applied as a framework to draw conclusions about machine learning and its efficacy as a tool in medical diagnosis. Results from this research should inform key stakeholders about relevant strengths and limitations for machine learning as a tool in medical diagnosis. This research is significant because it focuses on only machine learning algorithms that are applied to medical diagnoses. The research also has merit due to the diverse perspectives included in the analysis. News sources, books, and scholarly articles are used to build a well-rounded evaluation of the impact that machine learning as a tool for medical diagnosis has on patient outcomes. Ultimately, this research will help practitioners in the medical field properly integrate machine learning methods into medical diagnoses.
There was great value in completing the STS research paper and capstone simultaneously. The research conducted for the STS research paper was regarding the entirety of the medical industry, while the technical capstone was a more granular technical solution. As the technical capstone was completed, the research for the STS research paper provided wider knowledge of how technical solutions would function outside of the clinic. Had the projects been completed independently of each other, the technical capstone would likely not be as grounded in the current medical space. The STS research paper also benefitted from being completed while an actual machine learning solution was developed for the technical capstone. The limitations of statistical and machine learning solutions could be seen firsthand by completing the technical capstone, and these insights influenced the STS research paper.

School of Engineering and Applied Science
Bachelor of Science in Systems Engineering
Technical Advisor: Mehdi Boukhechba
STS Advisor: Bryn Seabrook
Technical Team Members: Kevin Cox, Sean Lynch, Alice Warner, Jane Romness, Sydney Lawrence, Drew Hamrock