Search Results

cNeuro cPET aids physicians in the evaluation of patient pathologies via assessment and quantification of PET brain scans.

The software aids in the assessment of human brain PET scans enabling automated analysis through quantification of tracer uptake and comparison with the corresponding tracer uptake in normal subjects. The resulting quantification is presented using volumes of interest and voxel-based maps of the brain. cNeuro cPET allows the user to generate information relative changes in PET-FDG glucose metabolism.

cNeuro cPET additionally allows the user to generate information relative changes in PET brain amyloid load between a subject's images and a normal database, which may be the result of brain neurodegeneration.

PET co-registration and fusion display capabilities with MRI allow PET findings to be related to brain anatomy.

cNeuro cPET aids physicians in the image interpretation of PET studies conducted on patients being evaluated for cognitive impairment, or other causes of cognitive decline.

cNeuro cPET has been developed to aid clinicians in the assessment and quantification of pathologies derived from PET scans. The software enables the display, co-registration, and fusion of PET images with those from MRI. Additionally, cPET enables automated quantitative and statistical analysis of tracer uptake by registration of a volume-of-interest atlas to the PET images and by comparing voxel and region-based uptake with corresponding uptake in healthy, amyloid-negative subjects. There are two quantification pipelines, one when the patient's MRI is not available (PET-only) and one when the patient's MRI is available (PET-MR). Quantification results are presented using volumes of interest, voxel-based or 3D stereotactic surface projection maps of the brain.

The provided text describes the regulatory filing for cNeuro cPET (K231576), a device intended to aid physicians in the evaluation and quantification of PET brain scans for pathologies, particularly related to FDG glucose metabolism and amyloid load.

Here's an analysis of the acceptance criteria and the study proving the device meets these criteria, based on the provided information:

1. Table of Acceptance Criteria and Reported Device Performance

The acceptance criteria are implicitly derived from the performance evaluation conducted, demonstrating the device's accuracy, robustness, and agreement with established methods and a predicate device. While explicit "acceptance criteria" values are not presented in a table form, the performance summary details the results proving the device meets its intended use and is substantially equivalent to its predicate.

2. Sample Size and Data Provenance

• Test Set Sample Size: 2,275 subjects.
• Data Provenance: The text explicitly mentions "a large dataset from the Alzheimer's Disease Neuroimaging Initiative (ADNI)", indicating a combination of retrospective and potentially prospective data, likely from diverse geographical origins given ADNI's multi-national nature, although specific countries are not stated. It also refers to Florbetaben Phase III data. The type of tracers supported (Flutemetamol, Florbetaben, FDG) were used in the testing.

3. Number of Experts and Qualifications for Ground Truth

• Number of Experts: For the majority visual read ground truth, 5 readers were used.
• Qualifications: The text does not explicitly state the qualifications of these readers (e.g., "radiologist with 10 years of experience"). However, given the context of PET brain scan interpretation, it can be inferred they are medical professionals specializing in nuclear medicine or radiology.

4. Adjudication Method for the Test Set

• For the visual read ground truth for Florbetaben categorization: A "majority visual read" by 5 readers was used. This implies a 3 out of 5 agreement for positive/negative classification.
• Additionally, a subset of scans where there was "unanimous categorization by the readers" (all 5 readers agreed) was analyzed separately.
• For histopathology-based ground truth, no adjudication method is relevant as it's a direct pathological diagnosis.

5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study

• The document does not describe an MRMC comparative effectiveness study where human readers' performance with and without AI assistance is compared.
• The study focuses on the standalone performance of the cNeuro cPET software and its agreement with established methods (predicate, FreeSurfer) and ground truth (histopathology, expert consensus). While expert visual reads were used to establish ground truth, they were not used in a comparative "human-in-the-loop" study format.

6. Standalone Performance (Algorithm Only)

• Yes, a standalone performance study was clearly conducted. The reported performance metrics (correlation with predicate, test-retest variability, agreement with histopathology and expert consensus) are all measures of the algorithm's performance independent of human-in-the-loop assistance. The device output includes quantitative metrics (SUVR, Z-scores, Centiloids) and maps for physician interpretation, but the performance testing itself is on the software's ability to generate these outputs accurately.

7. Type of Ground Truth Used

The study utilized multiple types of ground truth:

• Expert Consensus: "Majority visual read" by 5 readers for Florbetaben scan categorization.
• Pathology: "Histopathology" for Florbetaben scan categorization.
• Reference Methods/Data:
  • Comparison with the predicate device (CortexID Suite).
  • Comparison with a FreeSurfer based PET-MR quantification method used in ADNI.
  • Validation against Centiloid Project values (a standardized method for quantitative amyloid plaque estimation).

8. Sample Size for the Training Set

• The document does not specify the sample size used for the training set. The 2,275 subjects are explicitly stated as the dataset for "testing of cNeuro cPET". Without further information, it's unclear if parts of this dataset were used for initial training or if a separate, larger training set was used. It states that the "ADNI" dataset was used for "assessment of robustness," which might imply testing, but ADNI datasets are also often used for model development.

9. How Ground Truth for Training Set Was Established

• Since the training set size is not provided, the method for establishing its ground truth is also not explicitly described. However, based on the performance evaluation, it is highly probable that similar methods (expert consensus, established reference methods, or potentially a form of "weak" or "noisy" labeling for initial training) would have been employed. The document mentions the device performs "automated analysis through quantification of tracer uptake and comparison with the corresponding tracer uptake in normal subjects," which implies the use of a "normal database" that would have required some form of ground truth or normative data establishment during development.

Ask a specific question about this device