Search Results

This device is indicated to acquire and display cross-sectional volumes of the whole the head. The Aquilion Precision has the capability to provide volume sets. These volume sets can be used to perform specialized studies, using indicated software/hardware, by a trained and qualified physician.

FIRST is an iterative reconstruction algorithm intended to reduce exposure dose and improve high contrast spatial resolution for abdomen, pelvis, chest, cardiac, extremities and head applications.

AiCE is a noise reduction algorithm that improves image quality and reduces image noise by employing Deep Convolutional Neural Network methods for abdomen, pelvis, lung, and cardiac applications.

Aquilion Precision (TSX-304A/4) V10.10 with AiCE is an ultra-high resolution whole body multislice helical CT scanner, consisting of a gantry, couch and a console used for data processing and display. Aquilion Precision incorporates a 160-row, 0.25 mm detector, a 5.7- MHU large-capacity tube, and 0.35 s scanning, enabling wide-range scanning with short scan times to capture cross sectional volume data sets used to perform specialized studies, using indicated software, c by a trained and qualified physician. In addition, the subject device incorporates the latest reconstruction technology, AiCE (Advanced intelligent Clear-IQ Engine), intended to reduce image noise and improve image quality by utilizing Deep Convolutional Network methods to 1024x1024 HR/SHR images. These methods can more fully explore the statistical properties of the signal and noise. By learning to differentiate structure from noise, the algorithm produces fast, high quality CT reconstruction.

Here's an analysis of the acceptance criteria and supporting studies for the Aquilion Precision (TSX-304A/4) V10.10 with AiCE, based on the provided document:

1. Table of Acceptance Criteria and Reported Device Performance

The FDA clearance document doesn't explicitly lay out "acceptance criteria" in a typical pass/fail table format with specific numerical targets. Instead, it describes performance studies and their outcomes. The central theme is achieving substantial equivalence to the predicate device, often by demonstrating equivalent or improved performance in key areas.

• 16.5 lp/cm for Body in HR mode
• 10.5 lp/cm for Cardiac in HR mode
• 15 lp/cm for Lung in HR mode
• >10 lp/cm across all AiCE available body regions. (All compared to hybrid iterative reconstruction with conventional scanning at the same dose). |
  | Dose Neutral/Equivalent LCD & HC Spatial Resolution (Body) | Dose neutral relative to AIDR normal resolution mode, equivalent LCD with AIDR normal resolution, and improved high contrast spatial resolution. | Study comparing Aquilion Precision HR (High Resolution) mode with AiCE Body Sharp to Aquilion Precision NR (Normal Resolution) mode with AIDR Body determined:
• HR mode with AiCE is dose neutral relative to AIDR normal resolution mode.
• HR mode with AiCE has equivalent Low Contrast Detectability relative to normal resolution mode.
• AiCE has equivalent Low Contrast Detectability with 10 lp/cm more high contrast spatial resolution for body. |
  | Lung Cancer Screening (Image Quality) | Equivalent or improved performance relative to predicate device (Aquilion ONE GENESIS with AIDR). | CT image quality metrics (CT Number Accuracy, Uniformity, SSP, MTF-Edge, SD, NPS) were performed utilizing phantoms relevant to Lung Cancer Screening. Results demonstrated equivalent or improved performance for the Aquilion Precision in HR and SHR mode reconstructed with AiCE DLR, relative to the Aquilion ONE GENESIS with AIDR, for Lung Cancer Screening. |
  | Clinical Image Quality | Reconstructed images must be of diagnostic quality. | Representative body, cardiac, and low-dose chest images, as well as images of diseased patients with low contrast and small lesions, were reviewed by an American Board Certified Radiologist. It was confirmed that the reconstructed images using the subject device were of diagnostic quality. |

2. Sample Size Used for the Test Set and Data Provenance

• Quantitative Spatial Resolution Improvement, Dose Neutral/Equivalent LCD & HC Spatial Resolution, Lung Cancer Screening Image Quality: These studies utilized phantoms, not human patient data. Therefore, the concept of sample size and data provenance (country, retrospective/prospective) related to human data does not apply in the conventional sense. The "sample size" would relate to the number of phantom scans and settings tested. The document does not specify the exact number of phantom scans or specific phantom models.
• Clinical Image Quality: "Representative body, cardiac and low dose chest images, as well as images of diseased patients with low contrast and small lesions" were used. The exact number of images/cases (sample size) is not specified. The provenance of this data (country of origin, retrospective or prospective) is not explicitly stated.

3. Number of Experts Used to Establish the Ground Truth for the Test Set and Their Qualifications

• Noise Texture Reader Study, Quantitative Spatial Resolution Improvement, Dose Neutral/Equivalent LCD & HC Spatial Resolution, Lung Cancer Screening Image Quality: These studies primarily involved quantitative technical metrics derived from phantom scans. Expert assessment of "ground truth" in terms of clinical diagnosis or lesion identification was not the primary method for these bench tests.
• Clinical Image Quality: "An American Board Certified Radiologist" reviewed the images to confirm diagnostic quality. This indicates one expert. The specific years of experience or sub-specialization are not stated beyond "American Board Certified."

4. Adjudication Method for the Test Set

• Quantitative Bench/Phantom Studies: These studies rely on objective, quantitative measurements (e.g., lp/cm, CNR, SD, NPS, kurtosis values). Adjudication by multiple human readers is not applicable for these metrics.
• Clinical Image Quality: The document states that the images were "reviewed by an American Board Certified Radiologist" and "it was confirmed that the reconstructed images using the subject device were of diagnostic quality." This implies a single-reader review, without explicit mention of an adjudication process (e.g., 2+1, 3+1 consensus).

5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study Was Done

• No, an MRMC comparative effectiveness study is not explicitly mentioned or described. The document focuses on demonstrating substantial equivalence through technical performance metrics and a single-reader clinical review for diagnostic quality. There is no information provided about human readers improving with or without AI assistance, or any effect size.

6. If a Standalone (i.e., algorithm only without human-in-the-loop performance) Was Done

• Yes, the majority of the performance testing described (Image Quality Evaluation, Noise Texture Reader Study, Quantitative Spatial Resolution Improvement, Dose Neutral/Equivalent LCD and High Contrast Spatial Resolution Improvement, Lung Cancer Screening Image Quality) are standalone algorithm performance evaluations. These tests assess the objective image quality characteristics produced by the AiCE algorithm (and FIRST) primarily through phantom studies. The algorithm's output (reconstructed images) is then measured or analyzed.

7. The Type of Ground Truth Used

• Quantitative Bench/Phantom Studies: The "ground truth" for these studies is typically based on the known physical properties of the phantoms and the objective measurements derived from them. For example, for spatial resolution, the known line pair patterns in a phantom serve as ground truth for resolution capabilities. For CT number accuracy, the known attenuation properties of materials in the phantom serve as ground truth.
• Noise Texture Reader Study: The ground truth for perceptual "naturalness" of noise texture is more subjective but relies on expert consensus in the field regarding what constitutes "natural" CT noise (often referencing FBP). The study used quantitative measures (NPS and kurtosis) to support the claim without explicitly defining a human "ground truth" consensus process.
• Clinical Image Quality: The "ground truth" for diagnostic quality in this context is the expert opinion of the American Board Certified Radiologist reviewing the images.

8. The Sample Size for the Training Set

• The document states that the "subject device has been retrained to fit for the updated/novel indications." However, the sample size for the training set for the AiCE algorithm is not specified in this document.

9. How the Ground Truth for the Training Set Was Established

• The document states that AiCE "improves image quality and reduces image noise by employing Deep Convolutional Neural Network methods." Deep learning models like AiCE rely on extensive training data. While the document mentions "learning to differentiate structure from noise," it does not explicitly describe how the ground truth for this training data was established. Typically, this would involve large datasets of images (scans) with corresponding "ground truth" images or labels (e.g., ideal noise-free images, images with known anatomical structures, or images reviewed by experts). This information is not provided in the supplied text.

Ask a specific question about this device