NinesMeasure

NinesMeasure is a semi-automatic software tool designed for trained radiologists to aid in the quantitative analysis of solid pulmonary nodules from adult thoracic CT images by providing measurements of their long and short axis diameters. The device leverages artificial intelligence, specifically machine learning algorithms, to automatically segment and measure nodules indicated by the user on DICOM images, with results displayed for review within a PACS environment.

NinesMeasure is a semi-automatic software tool developed by Nines, Inc., intended for use by trained radiologists. Its primary purpose is to assist in the analysis and review of adult thoracic CT images by providing quantitative information about pulmonary nodule size. The system focuses on measuring the long and short axis diameters of solid pulmonary nodules in the axial plane, which can be tracked across single or multiple studies over time. It is designed to be a measurement aid and should not replace a full patient evaluation or be used to make or confirm a diagnosis, nor does it alter the original medical image. The methodology involves the analysis of DICOM images where a radiologist first indicates the location of a pulmonary nodule. NinesMeasure then employs artificial intelligence, specifically machine learning algorithms, to automatically perform the segmentation and subsequent measurements of these nodules. These calculated axial measurements are then displayed for review and potential editing by the radiologist, integrating with standard Picture Archiving and Communication Systems (PACS). Performance testing, conducted through a retrospective, multi-center image comparison study across various scanner manufacturers and clinical sites, demonstrated that the device met its primary endpoints for normalized error on long and short axis diameter measurements for nodules of varying sizes (3-10mm), supporting its intended function as a workflow tool for nodule measurement.

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. Given its limitation to solid pulmonary nodules, what were the specific exclusion criteria for nodule types in the validation study, and are there plans to expand NinesMeasure's capabilities to subsolid or ground-glass opacities?; How does the NinesMeasure algorithm differentiate between true nodules and nodule mimics (e.g., scars, vessels), and what is the false positive rate for automatic segmentation if a radiologist broadly indicates a region of interest?; The performance data reports normalized error; could you provide absolute error measurements (e.g., in millimeters) for different nodule size categories, and what is the system's sensitivity for detecting clinically significant changes in nodule size over time compared to manual radiologist measurements?; What specific training is recommended or required for radiologists to effectively use NinesMeasure, particularly regarding the interpretation of AI-generated measurements and the capabilities for manual review or editing of segmentations?; How was the 'ground truth' for nodule measurements established in the multi-center validation study against which NinesMeasure's AI algorithm performance was compared?