Systems and methods for orthopedic implants
Inventors
Casey, Niall Patrick • Cordonnier, Michael J
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12251313-B2
Publication Date
2025-03-18
Expiration Date
Abstract
A system and computer-implemented method for manufacturing an orthopedic implant involves segmenting features in an image of anatomy. Anatomic elements can be isolated. Spatial relationships between the isolated anatomic elements can be manipulated. Negative space between anatomic elements is mapped before and/or after manipulating the spatial relationships. At least a portion of the negative space can be filled with a virtual implant. The virtual implant can be used to design and manufacture a physical implant.
Core Innovation
The invention describes a computer-implemented method for providing surgical assistance by receiving one or more pre-operative images of a patient and determining a patient-specific treatment plan using at least one trained machine learning model. The patient-specific treatment plan includes a virtual model of a corrected spine of the patient based on the one or more pre-operative images and a patient-specific implant design for the patient-specific treatment plan. After the patient-specific implant is implanted in the patient, the method receives one or more post-operative images and determines a treatment outcome based on the received one or more post-operative images and the planned treatment.
The invention further determines whether to retrain the at least one trained machine learning model based on the treatment outcome and, in response to determining to retrain, selects post-operative data of the patient as one or more training items. It then retrains the at least one trained machine learning model using the one or more training items and determines, at least in part using the retrained at least one trained machine learning model, at least one patient-specific treatment plan for another patient. The approach is presented as training and reusing trained machine learning models to generate patient-specific treatment plans for additional patients.
Additionally, the invention includes a process for training at least one trained machine learning model that receives one or more pre-operative digital images, determines a surgery outcome based on the received pre-operative digital images, a pre-operative surgical plan, and post-operative surgical data, and retrains the at least one trained machine learning model when the surgery outcome meets at least one training criterion. The retrained model is configured to determine a planned anatomical correction for another patient and/or design a patient-specific implant for another patient to achieve the planned anatomical correction. A system and a non-transitory computer-readable medium are described as performing these same training operations.
Claims Coverage
The independent claims are clm-00001, clm-00013, clm-00020, and clm-00026. Across these independent claims, the coverage centers on determining patient-specific treatment plans using trained machine learning models from pre-operative images and training or retraining those models based on post-operative outcomes that meet at least one training criterion, with outputs including planned anatomical correction and patient-specific implant design for another patient.
Machine-learning based patient-specific surgical assistance with corrected spine model and implant design
Receiving one or more pre-operative images of a patient; determining a patient-specific treatment plan determined at least in part using at least one trained machine learning model, including a virtual model of a corrected spine of the patient and a patient-specific implant design; receiving one or more post-operative images; determining a treatment outcome based on the received one or more post-operative images and the planned treatment; determining whether to retrain the at least one trained machine learning model based on the treatment outcome; retraining the at least one trained machine learning model using selected post-operative data; and determining, using the retrained model, at least one patient-specific treatment plan for another patient.
Surgery-outcome-triggered retraining using pre-operative images and post-operative data
Receiving one or more pre-operative digital images of a patient; determining a surgery outcome based on the received one or more pre-operative digital images, a pre-operative surgical plan, and post-operative surgical data of the patient; in response to the surgery outcome meeting at least one training criterion, inputting at least one of the pre-operative digital images as a training item into at least one trained machine learning model; and retraining using at least one of the post-operative digital images, where the retrained machine learning model is configured to determine a planned anatomical correction for another patient and/or design a patient-specific implant for another patient to achieve the planned anatomical correction.
System for training machine learning models to determine correction or implant design for another patient
A system comprising one or more processors and one or more memories storing instructions that cause the system to receive one or more pre-operative digital images, determine a surgery outcome based on the received one or more pre-operative digital images, a pre-operative surgical plan, and post-operative surgical data, and in response to the surgery outcome meeting at least one training criterion input at least one of the pre-operative digital images as a training item into at least one trained machine learning model such that the model retrained using at least one of the post-operative digital images is configured to determine a planned anatomical correction for another patient and/or design a patient-specific implant for another patient to achieve the planned anatomical correction.
Non-transitory computer-readable medium for surgery-outcome-triggered model training
A non-transitory computer-readable medium storing instructions that cause a computing system to receive one or more pre-operative digital images of a patient, determine a surgery outcome based on the received one or more pre-operative digital images, a pre-operative surgical plan, and post-operative surgical data, and in response to the surgery outcome meeting at least one training criterion input at least one of the pre-operative digital images as a training item into at least one trained machine learning model, where the model retrained using at least one of the post-operative digital images is configured to determine a planned anatomical correction for another patient and/or design a patient-specific implant for another patient to achieve the planned anatomical correction.
Overall, the independent claims consistently require pre-operative imaging and a pre-operative surgical plan, a surgery outcome or treatment outcome determined using post-operative surgical data or images, a training criterion that triggers retraining, and outputs of planned anatomical correction and/or patient-specific implant design for another patient using retrained machine learning models.
Stated Advantages
Reduction of radiation exposure (intraoperative imaging).
Improved patient-specific coverage versus stock implants (coverage/logistics and suboptimal correction with stock implants).
Support for intraoperative trialing.
Documented Applications
Surgical assistance for patients undergoing orthopedic surgery, including spinal interbody implant design using a corrected spine virtual model and post-operative outcome evaluation.
Training machine learning models based on pre-operative images, pre-operative surgical plans, post-operative surgical data, and surgery outcomes to determine planned anatomical correction and/or design patient-specific implants for other patients.
Broader surgical assistance system providing recommendations, including virtual models and simulation (VR/AR) and surgical guidance/annotations (insertion points and angles).
Interested in licensing this patent?