A Machine Learning Method for Automated Intraoperative Assessments in Total Hip Surgery

Achieving precise positioning and alignment of anatomical structures and surgical instruments is paramount for successful outcomes in orthopedic surgery. The current standard practice heavily relies on intraoperative fluoroscopy (low dose real-time X-ray imaging) to assess the position and alignment of operative anatomy. However, these methods are qualitative and prone to errors and poor repeatability due to variabilities in positioning and orienting the imaging equipment, leading to suboptimal patient outcomes and increased risk of disabilities, often necessitating repeat surgeries. To address these challenges, Torus Biomedical has pioneered a unique solution known as ConfirMap system. This system seamlessly integrates with existing mobile fluoroscopy equipment in operating rooms and employs innovative calibration, tracking, and image processing techniques to provide extended radiographic visualization and precise measurements. Through strategic collaboration with a leading spine company in the US, a specialized version of the ConfirMap technology tailored for spinal surgery is slated to launch in the US market by Q2 2024. Currently, the company is dedicated to enhancing its technology to meet the specific demands of the orthopedic trauma field, particularly in femoral and tibial bone fracture reduction surgeries. This initiative aims to tackle critical technical challenges, thereby enhancing the viability of the ConfirMap system for orthopedic trauma applications, especially for procedures involving long bone fractures.

To ensure the effectiveness of the ConfirMap system in the ortho-trauma market, automating image analysis and measurement reporting with minimal surgical staff intervention is paramount. Our primary focus lies in procedures involving subtrochanteric and midshaft femoral fractures, where assessing the length, alignment, and rotation of the fractured bone poses significant challenges. Our project aims to leverage machine learning and computer vision techniques to develop and train a convolutional neural network (CNN) specifically tailored for the automatic segmentation of key bone landmarks based on the image format generated by the ConfirMap system. The outcome will streamline quick access to critical radiographic measurements during surgery, promising to enhance surgical efficiency and precision in ortho-trauma cases. The post-doctoral intern assigned to this project will play a pivotal role in developing a new machine learning model for the solution. As the initial step, they will establish a pipeline for creating realistic synthetic fluoroscopic images and corresponding labeling metadata from Computed Tomography (CT) images. Subsequently, they will construct, train, and optimize a deep neural network model, drawing inspiration from architecture layers proposed in academic literature. An important advancement of this research in the clinical application of the ConfirMap product will be the incorporation of multiple fluoroscopic images along with their spatial coordinates as input to the model.

Through this project the intern will have the opportunity to cultivate a comprehensive skill set encompassing research and development in medical imaging and machine learning. Collaborating with the company’s engineering team and academic co-applicants, esteemed authorities in medical image analysis and computer vision, the intern will benefit from invaluable mentorship, supervision, and guidance across all project stages. From conducting thorough literature reviews to designing and developing the neural network model from inception, the intern will gain hands-on experience in applying cutting-edge techniques to an innovative surgical imaging solution. This immersive experience will nurture the development of critical thinking, problem-solving, and collaboration skills as the intern tackles complex technical challenges. Furthermore, the intern’s participation in academic publications resulting from the project will not only advance their career development but also enhance their academic credibility in the field. Overall, this internship provides a unique and enriching opportunity for the intern to expand their knowledge, skills, and professional network within a dynamic and impactful research environment.

This project will support the company in developing a crucial technical solution to propel its product into new market opportunities. Introducing a new software feature for automatic radiographic measurements holds the potential to significantly enhance the effectiveness and versatility of the ConfirMap system for ortho-trauma applications. Upon demonstrating that the developed techniques meet the necessary accuracy and performance standards, the research outcomes will be seamlessly integrated into the ConfirMap system. Additionally, the findings of the research will be submitted to peer-reviewed scientific and clinical journals for wider dissemination and validation within the research community.