Research Projects

Accurate preoperative planning is central to the success of complex orthopedic surgeries such as Total Hip Replacement (THR) and Open Reduction and Internal Fixation (ORIF) of acetabular fractures. These procedures require precise anatomical understanding, correct implant sizing, and optimal placement to ensure stability and long-term function. Yet, current planning methods are either manual (based on printed X-rays and surgeon estimations) or reliant on expensive semi-automated proprietary software, accessible only to elite institutions. This creates disparities in surgical outcomes and increases the likelihood of complications, especially in high-volume or resource-constrained clinical settings. The need for a scalable, precise, and accessible planning tool is urgent. This project aims to develop and validate a fully automated AI-driven recommendation platform to help surgeons transform CT/MRI scans (DICOM files) into actionable surgical plans and patient-specific implant templates. The platform will integrate components such as (a) Automated 3D reconstruction of the pelvic and acetabular region (b) Segmentation and labeling of bone fragments as per standard medical definitions (c) Detection and classification of acetabular deformities or fractures (e) Optimal design of implant and fixation plate in ORIF (f) Automatic determination of acetabular socket and femoral head implant sizes tailored to patient-specific anatomy in THR (g) Computation of key parameters “inclination” and “anteversion” angles in THR (h) Generation of STL files for 3D prototyping for surgical validation and rehearsal. We hypothesize that a pipeline combining deep learning–based image segmentation, followed by fracture detection and classification for ORIF, and acetabular implant size estimation using anatomical geometry and ML techniques for THR, can generate clinically acceptable surgical plans that surpass traditional manual methods in terms of speed, accuracy, and consistency. The model will be trained and validated on a diverse dataset of anonymized CT/MRI scans labeled by experienced surgeons. The main development and validation phases involve (a) Data Collection and Labeling: Acquire and label clinical CT/MRI scans by medical experts (b) Model Development: Train deep learning models (Neural Networks) for segmentation and fracture classification and model the fixation plate contour design based on fracture patterns and anatomical curvature as an optimization problem and solve it using state-of-art optimization techniques to ensure anatomical conformity and mechanical stability (c) Develop geometry-based algorithms for cup sizing and orientation (d) Integration and Interface: Develop a unified software interface for clinicians with STL export and visualization tools. (e) Validation: Compare platform output with expert manual plans using quantitative metrics (e.g., segmentation accuracy, fit quality, implant alignment) (f) Pilot Testing: Conduct clinical pilot with orthopedic surgeons for usability feedback and outcome assessment. If successful, this project will result in a scalable, clinically viable, and cost-effective platform that transforms how complex orthopedic surgeries are planned. From a fundamental standpoint, it will advance the integration of AI and optimization in surgical design. From an application perspective, it will reduce planning and intraoperative time, improve consistency and accuracy in implant placement, lower complication and revision rates, and enable broader access to high-quality surgical planning across diverse clinical settings including government, smaller, and independent hospitals or clinics that perform orthopedic surgeries. It also offers potential for generalization to other complex procedures such as spinal, craniofacial, or tumor-related surgeries.