(443 days)
InferRead Lung CT.AI is comprised of computer assisted reading tools designed to aid the radiologist in the detection of pulmonary nodules ≥ 4mm during the review of CT examinations of the chest on an asymptomatic population ≥ 55 years old. InferRead Lung CT.AI requires that both lungs be in the field of view. InferRead Lung CT.AI provides adjunctive information and is not intended to be used without the original CT series.
InferRead Lung CT.AI uses the deep learning (DL) technology to perform nodule detection. It is a dedicated post-processing application that generates CADe marks as an overlay on original CT scans. The software can be installed in a healthcare facility or a cloud-based platform and is comprised of computer-assisted reading tools designed to aid radiologists in detecting, segmenting, measuring and localizing actionable pulmonary nodules that are 4mm or above during the review of chest CT examinations of asymptomatic populations, with enhanced capabilities for pulmonary nodule follow-up comparison and lung analysis. InferRead Lung CT.AI provides auxiliary information and is not intended to be used if the original CT series is not available.
The provided 510(k) clearance letter and summary discuss the InferRead Lung CT.AI device, its indications for use, and a comparison to predicate devices. It also details some standalone performance tests conducted to assess the newly introduced features of the device.
Here's an analysis of the acceptance criteria and study information based on the provided text:
1. Table of Acceptance Criteria and Reported Device Performance
The document details performance for newly added features rather than explicitly defined "acceptance criteria" for the overall device's primary function of nodule detection. However, it states that "predetermined testing criteria" were passed and that "validation tests indicated that as required by the risk analysis, designated individuals performed all verification and validation activities and that the results demonstrated that the predetermined acceptance criteria were met."
For the newly introduced functions, specific performance metrics are reported:
| Feature Tested | Acceptance Criteria (Implied/Expected) | Reported Device Performance |
|---|---|---|
| Nodule Registration | High accuracy in matching nodule pairs between current and prior scans. | Overall Nodule Match Rate: 0.970 (95%CI: 0.947-0.994) |
| Scan Interval Subgroup: |
- 0-6 months: 0.976 (95%CI: 0.911-1.0)
- 6-12 months: 1.000 (95%CI: N/A)
- 12-24 months: 0.938 (95%CI: 0.880-0.997) |
| Nodule Lobe Localization | High accuracy in identifying the correct lung lobe for detected nodules. | Overall Lobe Localization Accuracy Rate: 0.957 (95%CI: 0.929-0.986) |
| Lung Lobe Segmentation | High geometric similarity between automated segmentation and ground truth. | Average Dice Coefficient: 0.966 (95%CI: 0.962 to 0.969) |
2. Sample Sizes Used for the Test Set and Data Provenance
- Nodule Registration Standalone Test: 98 lung cancer screening cases with 206 nodule pairs.
- Nodule Lobe Localization Standalone Test: 94 lung cancer screening scans with 188 nodules.
- Lung Lobe Segmentation Standalone Test: 22 lung cancer screening cases with 110 lung lobes.
Data Provenance: The document does not explicitly state the country of origin for the data used in these tests, nor does it specify if the data was retrospective or prospective. It refers to "lung cancer screening cases/scans," suggesting these are clinical datasets.
3. Number of Experts Used to Establish the Ground Truth for the Test Set and Their Qualifications
The document does not explicitly state the number of experts or their qualifications who established the ground truth for the standalone performance tests.
4. Adjudication Method for the Test Set
The document does not specify the adjudication method used for establishing the ground truth for the test sets in these standalone performance evaluations.
5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study Was Done, and Effect Size
The document does not indicate that a Multi-Reader Multi-Case (MRMC) comparative effectiveness study was done to assess the improvement of human readers with AI assistance versus without AI assistance. The performance tests described are standalone evaluations of specific AI functions.
6. If a Standalone (Algorithm Only Without Human-in-the-Loop Performance) Was Done
Yes, the document explicitly describes standalone performance testing for the newly added functions: "For the newly added functions, including nodule registration, nodule localization and lung lobe segmentation, we conducted standalone performance testing." The reported results (Nodule Match Rate, Lobe Localization Accuracy Rate, Dice Coefficient) are all metrics of the algorithm's performance without human interaction.
The document also states: "Regarding the performance of the AI outputs, the nodule detection and segmentation functions were consistent with the predicate product (K192880), as verified through consistency testing." This implies that the primary nodule detection and segmentation capabilities were also assessed in a standalone manner, likely by comparing the AI's output to a ground truth.
7. The Type of Ground Truth Used
The document does not explicitly state the type of ground truth used for the standalone tests (e.g., expert consensus, pathology, outcomes data). However, for metrics like "Nodule Match Rate," "Lobe Localization Accuracy Rate," and "Dice Coefficient," the ground truth would typically be established by expert radiologists or reference standards. For Dice Coefficient in segmentation, it would likely be expert-drawn segmentations.
8. The Sample Size for the Training Set
The document does not provide any information regarding the sample size of the training set used to develop the InferRead Lung CT.AI device.
9. How the Ground Truth for the Training Set Was Established
The document does not provide information on how the ground truth for the training set was established.
§ 892.2050 Medical image management and processing system.
(a)
Identification. A medical image management and processing system is a device that provides one or more capabilities relating to the review and digital processing of medical images for the purposes of interpretation by a trained practitioner of disease detection, diagnosis, or patient management. The software components may provide advanced or complex image processing functions for image manipulation, enhancement, or quantification that are intended for use in the interpretation and analysis of medical images. Advanced image manipulation functions may include image segmentation, multimodality image registration, or 3D visualization. Complex quantitative functions may include semi-automated measurements or time-series measurements.(b)
Classification. Class II (special controls; voluntary standards—Digital Imaging and Communications in Medicine (DICOM) Std., Joint Photographic Experts Group (JPEG) Std., Society of Motion Picture and Television Engineers (SMPTE) Test Pattern).