SIS Software

SIS Software is an application designed to enhance neurosurgical planning by using machine learning and a 7T MRI reference database to create patient-specific 3D models of the subthalamic nuclei from standard clinical MRI scans. Its methodology aims to improve visualization accuracy for STN targeting in procedures like stereotactic surgery, offering a significant improvement over standard visualization techniques.

SIS Software is a medical imaging application intended for viewing, presentation, and documentation of medical images, featuring modules for image processing, image fusion, and intraoperative functional planning. It is specifically designed to assist medical professionals, including surgeons, neurologists, and radiologists, by enhancing the visualization of the subthalamic nuclei (STN), providing a 3D output that can be utilized with stereotactic image-guided surgery systems or other devices for further processing. The software employs machine learning and image processing techniques to improve standard clinical MR images (1.5T or 3T). It leverages a reference database composed of high-resolution 7T MRI scans—where deep brain structures like the STN are more clearly discernible—to accurately identify and model these structures on a patient's lower-resolution clinical images, thereby generating a patient-specific 3D anatomical model of the brain. The performance of SIS Software was validated through rigorous testing. Developmental testing, using a leave-one-out statistical method on 10 subject datasets, demonstrated an average prediction accuracy of 1mm for the STN, which was significantly better than standard atlas methods. Pivotal validation testing was conducted on images from 34 subjects (68 STNs), using a dataset entirely independent of the one used for algorithm development. This validation assessed metrics such as center of mass distance, surface distance, and Dice coefficient. The results showed that 95% of center of mass distances and 100% of surface distances were not greater than 2.0mm, successfully meeting pre-specified accuracy criteria and significantly outperforming estimated standard-of-care visualization success rates (p<0.0001). The Dice coefficient recorded was 0.64, a value considered acceptable given the small size of the STN. These findings confirm that the SIS Software performs as intended, offering a clinically significant improvement in STN visualization for neurosurgical planning.

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 does the SIS Software's machine learning algorithm adapt to anatomical variability in the STN across different patient populations (e.g., age, pathology)?; What is the typical processing time required by the SIS Software to generate the patient-specific 3D model from clinical MRI data, and how might this impact intraoperative workflow?; Could the methodology of using a 7T MRI reference database be expanded to improve visualization of other deep brain structures beyond the STN that are poorly resolved on standard clinical MRI?; What are the specific limitations or contraindications for using SIS Software, particularly concerning image quality prerequisites for the input clinical MRI scans?; How does the performance of SIS Software compare, in terms of STN localization accuracy and clinical utility, to other advanced imaging techniques or manual segmentation methods employed by experienced neuroradiologists?