Artificial Intelligence Applied to Three-Dimensional (3D) Rendered Models from MR Imaging: Optimizing Preoperative Planning & Surgical Outcomes in Patients with Uterine Fibroids
New Technology, Therapies, eHealth & mHealth
Symptomatic uterine fibroids are a common and complex gynecologic condition. Medical imaging is the first line of investigation for diagnosis and treatment planning for uterine fibroids. Our team has demonstrated the value of creating 3D printed models from medical images to improve a surgeon’s appreciation for patient-specific anatomy. To minimize the barrier to time and expertise of creating 3D models, we are developing algorithms to faciliate the 3D segmentation process. 3D models for complex surgical cases will be created from MRI data and visualised by surgeons using virtual reality (VR) technology. This will improve a surgeon’s pre-operative plan, optimize surgical decision making, and intra-operative performance, thus improving surgical and patient outcomes.
Uterine fibroids are a common gynecological disease prevalent in up to 80% of pre-menopausal females, with half experiencing significant morbidities such as heavy menstrual bleeding, pelvic pain, and infertility. Surgical intervention is commonly employed to manage these symptoms but can present with risks such as major hemorrhage, requiring blood transfusion, adhesions, ureteric/bladder injury, and infection. To optimize surgical outcomes, a thorough understanding of the patient’s anatomy is required. Recent technological advances have facilitated the clinical use of 3D rendered models from medical imaging to better visualize complex anatomy and improve preoperative planning. A major barrier to clinical uptake in gynecology has been the time and technical expertise required to create models of uterine anatomy.
Phase 1 of this project proposes an innovative approach of adapting advanced analytical and machine learning techniques to reduce processing time and develop a more automatic process for creating 3D gynecological models from medical imaging. We have segmented 54 MRI cases of uterine fibroids to date. We are currently developing pipelines for our machine learning approach and plan to complete its first iteration and have preliminary results by end of June 2023.
Phase 2 of our project will apply the analytical methods developed in Phase 1 to create 3D models of complex uterine anatomy. The 3D models will be visualized using The Ottawa Hospital’s virtual reality (VR) technology. We will evaluate the impact of using 3D VR models on preoperative planning and surgical outcomes using a randomized case-control experimental design. We are currently preparing our REB application for Phase 2 and expect first patient to be enrolled by end of August 2023.
Our team are leaders in advanced 3D visualisation of complex patient specific anatomy. 3D rendered models of uterine fibroids from MRI data to improve a surgeon’s understanding of patient-specific anatomy, optimize surgical decision making, and intra-operative performance, thus improving surgical and patient outcomes.
Methodologies developed will contribute to future research for (1) other complex pathologies that may benefit from a 3D model for pre-surgical planning; and (2) to evaluate the role of 3D models for both surgical trainee and patient education. Our vision is to use 3D models as standard of care for all surgical specialties at TOH which will improve physician performance, the patient experience, and decrease impact on our healthcare system.
Data collection and processing on-going.
Data from this project has been used for a BSc Honour’s thesis project to evalaute the agreement between 3D rendered models, planar measures on MRI and planar measures on ultrasound of uterine and fibroid volume estimates. This work is submitted for presentation at:
– Association for Gynecologic Laparoscopy (AAGL) Global Congress on Minimally Invasive Gynecologic Surgery
– Canadian Association for Gynecologic Excellence (CanSAGE) Annual Conference