AI-Powered Robotic Surgery

Remotely controlling robotic surgery systems over networks to allow surgeons to operate robots from distant locations. The method involves associating input signals from a surgeon console with timestamps and transmitting them to the robot. At the robot side, the timestamps are used to generate control signals at the right times to match the original inputs. This ensures synchronized motion of the robot's instruments despite network delays. The timestamps are also used to correct for out-of-order inputs.

Source

Robotic electrospinning of resorbable bifurcated vascular grafts with customizable shapes using a flexible electrically conductive internal collector and dynamically positioned mandrels. The collector extends through the mandrel lumen and bends as it rotates to coat the mandrel with densely tangled fibers. The collector and mandrel shapes are patient-specific to match bifurcated arteries. The method uses electrospinning control and robotic positioning to create resorbable bifurcated grafts for vascular applications.

Source

Registration of medical imaging and robotic devices to improve guidance and control of interventional therapies. The registration aligns robotic data with imaging data like ultrasound and fluoroscopy. This allows accurate visualization and manipulation of robotic tools inside the body. It involves calculating the robotic tool's pose relative to the imaging field, then using that alignment to drive the robotic tool's motion. This registration helps overcome challenges of accurately controlling robotic catheters and other devices in complex 3D anatomies.

Source

Automated system to enhance fluoroscopic registration for accurate and safe placement of pedicle screws, interbody spacers, and vertebral augmentation in spine surgery. The system uses image guidance to determine optimal screw diameter, length, and trajectory based on preoperative scans. It aims to improve accuracy and reduce risks by providing automated assistance for screw placement compared to manual determination. The system can be used with open or percutaneous techniques.

Source

Monitoring and adjusting a rod reduction process during spine surgery to prevent breakage or loosening of pedicle screws and the rod. A force/torque sensor on the reduction tool measures forces during rod insertion. A computerized model analyzes screw quality, rod geometry, patient alignment, etc. to determine force thresholds. If monitored forces exceed thresholds, the system pauses or adjusts the reduction plan to prevent screw/rod breakage. This reduces the risk of complications during rod insertion.

Source

Generating haptic feedback for robotic surgery to provide surgeons with tactile sensations through the robotic interface. It involves training machine learning models to predict force levels based on video images of tool-tissue interactions. The models learn correlations between visual appearances and forces. During surgery, the models analyze video frames to estimate the applied force. The estimated force is then provided as haptic feedback to the surgeon through the robotic interface.

Source

Using AI to predict nerve locations during spine surgery to reduce risks of nerve injury. The method involves training a neural network using labeled MRI scans to associate nerve positions with surrounding anatomy. It then applies this model to unsegmented CT scans to predict nerve locations during surgery. This provides a visualization of nerve paths without needing MRI. The AI can also account for patient position changes during surgery to update nerve predictions.

Source

Automatic disengagement of surgical tools from robotic arms when the user returns the hand controllers to their docks, without needing a foot pedal. A recurrent neural network classifier detects if the controller motions are docking or teleoperation. When docking is detected, the robot disengages the tools to prevent unintended movements. The classifier is trained on labeled time-series data of controller motions.

Source

Using machine learning to analyze surgical videos and derive insights like identifying deviations from surgical planes, analyzing surgical competency, and generating statistical reports with links to video evidence. The system receives video frames from multiple surgeries, categorizes them based on events, aggregates stats within categories, and presents frames from categories. It also predicts future events in ongoing surgeries, suggests video reviews, and assigns surgeons to surgeries based on skill level and time estimates.

Source

Automatically detecting surgical states, instrument locations, and motion profiles from video during surgery using machine learning. The method involves training machine learning models to predict surgical states and detect instruments in video frames based on input data. It uses shared features between state and instrument detection to improve both tasks. The models can also leverage temporal information from state prediction to boost instrument detection. This allows real-time augmentation of surgical video with instrument overlays and motion profiles.

Source

Artificial intelligence (AI) system for real-time surgical guidance during procedures like joint replacements, fracture reductions, and spine surgeries. The system analyzes intraoperative images to calculate surgical decision risks and provides automated guidance to optimize implant placement, fracture reduction, and alignment. It uses AI models trained on surgical data to identify anatomical landmarks, register images, and predict risks. The system displays visual guidance to surgeons, dynamically updating as the procedure progresses.

Source

Using AI to enhance robotic surgery by leveraging image recognition to identify tissue types during surgery and allowing the robot to autonomously perform actions based on the identified tissue. The system involves obtaining images of the surgical site, identifying the tissue types using AI, and if the identified tissue matches a targeted type, having the robot remove it. This allows the robot to perform specific actions without manual guidance once the AI tissue recognition is confident. The AI can also help with tasks like incision placement, guide wire manipulation, and imaging output projection to improve accuracy and repeatability.

Source

Tracking the position of a robotic surgical instrument using ultrasound (US) instead of optical markers. A US transducer is attached to the instrument and generates US images of the anatomy. The instrument's pose is determined by matching anatomical features in the US images to a template of the instrument. This allows accurate tracking even if the US transducer loses contact with the patient. The system switches to optical tracking if the US transducer lifts off. It also uses kinematics if the US transducer stops outputting. This provides redundancy and continuous tracking during instrument motion.

Source

Robotic surgical system with enhanced tissue suturing guidance using visual feedback and machine learning. The system uses an imaging device to capture the position and orientation of the surgical needle inside the body during suturing. It estimates the needle tip location and trajectory from the images. This data is used to refine the robot's control signals to improve needle placement and avoid collisions. The system can also augment the video feed to highlight the needle and show its path. It can also check for tissue contact to guide the robot's motion.

Source

Multi-modal position tracking system for accurately tracking objects in extended time periods for applications like surgery. The system uses hierarchical tracking sub-systems with parent, child, grandchild, etc. systems that record object positions relative to their virtual spaces. A mapping between the virtual spaces allows translating positions across the hierarchy. This provides multi-modal position estimation and verification by observing objects from multiple sub-systems and correcting them if needed. It enables accurate tracking over extended time for surgical applications by compensating for drift between sub-systems.

Source

Using machine learning and computer vision to automatically detect critical anatomical structures and surgical instruments during live video of a surgical procedure. The technique involves training machine learning models to predict the presence and location of instruments and structures in the video frame using spatio-temporal information. The models can also predict the overall surgical state. The instrument and structure detection models share extracted features with the state prediction models to improve confidence. This allows real-time augmented visualization of the surgical view. The technique improves surgical safety and workflow by providing action guidance and alerts.

Source

Computer-assisted surgery system that can recommend deviations from standard surgical flows based on intraoperative video analysis. The system uses a processing unit and memory to monitor a surgical procedure via video feed and detect conditions requiring deviations outside the standard flow. It trains a machine learning model using video and tracking data to perform tracking without the separate tracking device. This allows the system to provide intraoperative recommendations for deviations based on video analysis alone.

Source

Real-time augmented surgical guidance system using AI to provide predictive visualizations of critical structure locations and tissue viability during surgery. The system uses machine learning models trained on medical datasets to anticipate critical structures and tissue viability based on imaging and physiological data. It can also provide real-time guidance and decision support during surgical procedures. The AI analyzes imaging data from multiple modalities and extracts features to classify structures and assess viability. The system generates enhanced views with guidance and metrics to assist surgeons. It can also autonomously track deformable tissues during maneuvers. The AI is trained on whole surgery datasets with anatomical and physiological data.

Source

Using computer vision to improve safety and reliability of surgical procedures by controlling the operation of surgical tools based on recognized objects in the video feed. The system trains a computer vision model to recognize surgical tools from live video. During surgeries, the model is used to detect surgical tools in the video feed. When a tool is detected, it enables control of the tool to perform its function. This ensures tools are only used when visible, preventing accidental activation outside the field of view.

Source

Optimizing surgery by using simulation and AI to find the best tools, procedures, and entry points for minimally invasive surgeries. The method involves generating genes representing surgical procedures, simulating them in virtual bodies, evaluating optimality, and applying genetic algorithms to derive improved procedures and instrument configurations. This provides optimized surgical cue sheets and robot designs that improve efficiency and convenience in actual surgery.

Source