Precise Image

Precise Image is an Artificial Intelligence-powered CT reconstruction software designed to reduce radiation dose while improving image quality, specifically enhancing low contrast detectability and reducing noise in adult CT imaging. Utilizing Convolutional Neural Networks, the software reconstructs X-ray transmission data to produce images with an appearance similar to traditional Filtered Back Projection, offering adjustable settings to optimize diagnostic outcomes for head, body, and vascular applications.

Precise Image is an advanced reconstruction software application designed for Philips whole-body Computed Tomography (CT) X-Ray Systems, utilizing Artificial Intelligence (AI) and Convolutional Neural Networks (CNN). Its primary purpose is to generate CT images that offer an image appearance similar to traditional Filtered Back Projection (FBP) images, while simultaneously reducing radiation dose and enhancing image quality. The software, which is trained on and compatible with the Philips Incisive CT System, aims to achieve lower noise levels and improve low contrast detectability. It provides five user-adjustable settings, allowing radiologists to tailor dose reduction and image quality according to their preferences for diagnostic imaging of the head, body, and vascular system in adult patients; it is not indicated for pediatric use. Compared to conventional methods like iDose, Precise Image is engineered for significantly lower radiation dose—claiming up to 80% dose reduction—while providing up to 85% lower noise and 60% better low contrast detectability, and an image appearance more akin to FBP at higher doses. Non-clinical testing showed improved low contrast resolution at lower dose levels (e.g., 5 mm @ 0.3% @ 5.5 mGy CTDIvol) and ensures the Noise Power Spectrum (NPS) does not shift by more than 6% when noise is reduced by at least 50%. Clinical performance was supported by a comparative image evaluation study where six board-certified radiologists assessed 55 image set pairs from Precise Image and predicate devices, evaluating diagnostic confidence, sharpness, noise, image texture, and artifacts, with all images deemed to have good image quality.

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 Precise Image AI algorithm specifically balance the trade-off between noise reduction, low contrast detectability, and image texture to mimic Filtered Back Projection images across different anatomical regions and patient sizes, particularly at significantly lower doses?; What were the specific quantitative results from the comparative image evaluation study (e.g., mean Likert scores for diagnostic confidence, sharpness, noise) when comparing Precise Image to the predicate Incisive CT, especially at the claimed 80% lower dose?; The document states Precise Image has 5 user-adjustable settings. What are these settings, how do they interact, and what training is recommended for radiologists to optimize image quality and dose reduction for specific clinical scenarios?; Given the training on the Philips Incisive CT system, how robust is the Precise Image algorithm to variations in data from different CT scanners within the Philips family, or to variations in acquisition protocols not included in the initial training dataset?; What are the computational performance implications, such as reconstruction time, of using the Precise Image AI algorithm compared to standard FBP or iterative reconstruction methods like iDose on the Philips Incisive CT system?