Digital Health and Technology in Hip and Knee Surgery
UCSF’s Arthroplasty group, within the Department of Orthopaedic Surgery, participates in and designs research studies that are on the cutting edge of digital health and technology. Surgeon and Professor, Stefano Bini, MD, has spearheaded this effort within Arthroplasty with various research projects involving artificial intelligence, mobile applications, and wearable activity trackers such as the Apple Watch and Fitbit.
The research performed in this area is innovative and can lead to how medical providers communicate with, track, treat, and follow up with patients both preoperatively and postoperatively. We invite medical students, residents, and those with the knowledge and expertise in artificial intelligence and digital health technology to join our research team as we strive to be the leader in Orthopaedic digital health and technology.
For more information about our research opportunities, please contact the clinical research coordinator Michael Henne (ArthroplastyResearch@ucsf.edu).
A validation study using gait analysis to test the accuracy of wearable sensor data in postsurgical patients
Site Investigator: Stefano Bini, MD
Wearable sensors are becoming increasingly accurate in their evaluation of motion in three-dimensional space. At UCSF, we are building the knowhow and ability to capture that data and evaluate it in the clinical, post-operative setting. The hope is that patient generated motion data will prove to be predictive of future outcomes and the potential risk of complications such as falls or pain. For this type of qualitative application, data trends can be sufficient and data consistency is most important. However, for quantitative research or individualized clinical decision making, data accuracy is crucial, particularly where few data points are generated.
Wearable sensors for the most part have been designed to capture data from healthy people with normal gait. It is not clear how effectively or accurately they capture data in post-operative patients with generally slower, abnormal gait. Testing these results against established reference standards such as gait analysis data is a necessary next step. Further, it is possible that an unknown relationship exists between wearable data results and patient outcomes and this too has not been evaluated. If a strong association can be demonstrated between wearable sensor data and clinical outcomes, wearables could quickly become the reference point for objectively evaluating clinical results with greater accuracy than ever before.
Sponsor: Center for Disruptive Musculoskeletal Innovations (CDMI)
The Oaks: A Mobile Exercise and Education Program for People with Osteoarthritis of the Knee
Site Investigator: Stefano Bini, MD
Blue Marble Health, is a digital health company that is developing a mobile app called "The Oaks" with Small Business Innovation Research (SBIR) funding. UCSF is one of 4 sites that have been subcontracted to participate in this multicenter study, which will develop and test the mobile app called “The Oaks,” a perioperative, digital, education, assessment and guided exercise software application for adults with osteoarthritis of the knee (KOA) awaiting total knee replacement (TKR) surgery. This product is relevant to the mission of the National Institute on Aging (NIA) because it addresses the call for “Development of technologies to assist in the improvement of physical function and mobility in older persons prior to (prehabilitaiton) or following (rehabilitation) elective/ planed surgery.” The envisioned minimum viable product (MVP) is an elegantly simple mobile application that is designed to specifically measure the Centers for Medicare/Medicaid’s (CMS) Comprehensive Care for Joint Replacement (CJR) required assessments and physical performance assessments prior to and following TKR, provide educational content typically contained in pre-operative TKR classes and track recovery of function for 30 days following TKR surgery. This product was conceived in response to four main drivers: 1) the growing number of adults aging with KOA needing TKR, 2) the need for developing improved patient care perioperatively for TKR surgeries, 3) the finding that attending educational classes and performing strengthening exercises prior to TKR improves outcomes and lowers post surgical medical utilization potentially saving billions of dollars and 4) the CMS CJR rule requiring healthcare systems to reduce the cost of care delivery for TKR surgeries, within a diminishing bundled payment. This application is very timely and has great potential for widespread adoption, as it will be designed based on end user feedback, from all stakeholder groups including people with KOA and the medical care team.
Sponsor: Blue Marble Health
Development and Application of Computer Vision and Machine Learning Solutions for the Analysis of Musculoskeletal Images
Site Co-Investigators: Valentina Pedoia, PhD and Stefano Bini, MD
Machine learning showed promising results in several medical imaging tasks, specifically in deep convolutional neural networks. These computational models are composed of multiple processing layers and learn representations of data with multiple levels of abstraction. Deep neural networks utilize the tendency that many natural patterns are compositional hierarchies; higher-level features are composed of lower-level features. In medical images, from local combinations of intensity and edges to local sequences, patterns assemble into parts, and parts subsequently form objects (cartilage, bone, cartilage lesions, bone marrow edema like lesions). Deep convolutional neural network technology in combination with advanced computer vision and multidimensional big data analytics techniques will be developed in this study for the analysis of MRI images to enable a fully automated multi tissue image segmentation, voxel based relaxometry analysis and data-driven feature extraction for abnormalities detection and classification using MRI clinical grading as gold standard and with the ultimate goal of predicting OA progression and final outcomes. This project will utilize data extracted from the Osteoarthritis Initiative Dataset as well as from the UCSF clinical archives including demographic and clinical data in the electronic health records; by using data extracted from the Picture Archiving and Communications Systems (PACS); with the ultimate goal of clinical translation of the proposed technical solutions. This project will have the ultimate goal of characterizing the degeneration of the joint and establish the platform to study the ability of these techniques to predict future progression.
Using Natural Language Processing to Code Orthopedic Operative Reports
Principal Investigator: Thomas Vail, MD
Implementation of electronic medical records (EMRs) have led to a large amount of stored clinical data. A large amount of this data, such as clinical notes and unstructured chart review data, is not easily accessible or coded and therefore has not been used for its full potential in research and QI yet. Orthopedic surgery has been particularly slow at collecting this data, and only a few orthopedic data registries exist in the United States. Artificial Intelligence and Natural Language Processing (NLP) could be used as a method to help create larger registries for orthopaedic surgery, without increased time-burden by creating structured data from operative notes. If so, this could provide a method for physicians to evaluate their practice patterns, and provide a method to create more meaningful national orthopaedic data registries for population level analysis.
Our objectives are to investigate whether NLP can be used to code notes, from pre-operative notes to operative reports to post-op notes, for total joint arthroplasty (TJA) and abstract useful clinical data that surgeons can use for further analysis
American Joint Replacement Registry
UCSF’s Arthroplasty group, within the Department of Orthopaedic Surgery, submits data to the American Joint Replacement Registry (AJRR). AJRR's mission is to improve orthopaedic care through the collection, analysis, and reporting of actionable data. They seek to be the National Registry for orthopaedics through comprehensive data derived from information submitted by hospitals, physicians, and patients related to total joint arthroplasty as well as through technology, resulting in optimal patient outcomes.
Cognitive Computing Support Can Improve Physician Test Scores on Standardized Medical Exams
Principal Investigator: Stefano Bini, MD
CloudMedx is a company that leverages the latest clinical algorithms, machine learning technology, advanced natural language processing, and proprietary clinical contextual ontology to improve patient journeys. The study aims to use this artificial intelligence (AI) to help improve Resident test taking skills. A pilot study conducted in 2018 at UCSF showed that this AI outperformed Residents in test taking. Now, we will be studying whether the Resident test scores improve when they take a test comprised of 2,000 randomly selected questions with support from the AI. Half of the 20 Residents will take the test by themselves and the other half will take the test with support from the AI. The AI will also take the test by itself, without any human input.
The Use of Virtual Reality Training Modules to Improve Complex Surgical Technique Recall and Task Execution in Surgical Trainees
Principal Investigator: Jeff Barry, MD
Training of surgical skills and technique is an essential component of any surgical residency curriculum. Unfortunately, there are few environments or opportunities for trainees to hone skills outside of the high-stakes environment of the operating room on real patients. Given the conflict that can arise between needing to train the next generation of surgeons and providing the highest quality care for the individual patient at hand, there is a unmet need for high-fidelity, low-risk, low-cost surgical training curriculum. We propose a plan to begin validating the feasibility and effectiveness of virtual reality (VR) based surgical skills training module for orthopaedic surgery residents. This project will look to determine if VR can be utilized as a superior alternative to the traditional out of the operative room learning methods currently in place when it comes to learning a complicated, multi-step orthopaedic surgery procedure: a partial unicompartmental knee replacement (UKA). Should this prove successful we hope to expand implementation within the orthopaedic residency and beyond for trainee benefit for multiple procedures and subspecialties.
Prospective Database of Lower Extremity Motion Using a Low-cost Motion Analysis System
Principal Investigator: Jeffrey Lotz, PhD
This study is being done to look at a new way (the Microsoft Kinect system) to measure leg motion in people with different injuries or disorders. We think that this minimally invasive technology will be able to determine differences in motion of the lower extremity in patients without injury or in those who are at risk of injury. We also think that this will help to monitor return to activities and rehabilitation in patients after intervention (such as surgery or physical therapy).
- Bini SA, and Mahajan J. Clinical Outcomes of Remote Asynchronous Telerehabilitation are Equivalent to Traditional Therapy following Total Knee Arthroplasty: A Randomized Control Study. Journal of Telemedicine and Telecare. 2017; 2(23):239-247
- Bini SA. Artificial Intelligence, Machine Learning, Deep Learning, and Cognitive Computing: What Do These Terms Mean and How Will They Impact Health Care? J Arthroplasty. 2018 Feb 27. PMID: 29656964