SIS System

The SIS System is a medical image processing software designed to assist clinicians in neurological procedures, particularly Deep Brain Stimulation, by providing enhanced visualization and localization of specific deep brain structures including the STN, GPe, GPi, and the newly added Vim. It utilizes machine learning, specifically pre-trained deep learning neural networks trained with ultra-high resolution 7T MRI data as ground truth, to generate patient-specific 3D anatomical models from standard clinical MR images, aiding in stereotactic surgical planning.

The SIS System, developed by Surgical Information Sciences, Inc., is an advanced medical imaging software intended for the viewing, presentation, and documentation of medical images. It facilitates image processing, image fusion, and intraoperative functional planning, with its 3D output usable in stereotactic image-guided surgery or with other devices for further processing and localization. Specifically, the system aids medical professionals like surgeons, neurologists, and radiologists in visualizing and locating deep brain structures such as the subthalamic nuclei (STN), globus pallidus externa and interna (GPe/GPi), and the ventral intermediate nucleus (Vim) during neurological procedures, including Deep Brain Stimulation (DBS) lead placement. This version represents a modification to a previously existing system, primarily adding support for Vim visualization. The core technology of the SIS System is based on machine learning and image processing, employing pre-trained deep learning neural network models to enhance standard clinical MR images (1.5T and 3T). These models are trained using ultra-high resolution 7T Magnetic Resonance (MR) images to establish ground truth, enabling the generation of patient-specific 3D anatomical models of the targeted brain structures. The system can also identify implanted leads from post-operative CT images and correlate their positions with pre-operative brain structure models. Performance validation for the new Vim visualization feature involved comparing SIS outputs against ground truth segmentations (derived from clinical MRI, 7T MRI, and DiMANI). The validation demonstrated high accuracy, with 90% of center of mass distances for Vim below 1.83mm, 90% of mean surface distances below 0.86mm, and a mean Dice coefficient of 0.7, meeting pre-specified criteria.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How robust is the SIS System's Vim visualization to variations in input clinical MRI scanner manufacturers, field strengths (1.5T vs. 3T), and acquisition protocols, beyond the tested datasets?; What are the specific advantages and limitations of using DiMANI for ground truth segmentation of the Vim compared to direct manual segmentation on 7T MRI in the context of training and validating the deep learning models?; How does the system handle cases with significant brain anatomical abnormalities or prior surgical interventions that might affect the typical appearance or location of the targeted nuclei?; What is the computational time required for the SIS System to process a typical patient dataset and generate the 3D model of the Vim, and how does this integrate into a typical clinical workflow for surgical planning?; Are there plans to expand the SIS System's capabilities to include other subcortical structures relevant to neurological disorders or to integrate other imaging modalities beyond MRI and CT for model generation or validation?