InferRead Lung CT.AI is comprised of computer assisted reading tools designed to aid the radiologist in the detection of pulmonary nodules during the review of CT examinations of the chest on an asymptomatic population. 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.

Device Description

InferRead Lung CT.AI uses the Browser/Server architecture, and is provided as Service (SaaS) via a URL. The system integrates algorithm logic and database in the same the simplicity of the system and the convenience of system maintenance. The server is able to accept chest CT images from a PACS system, Radiological Information System) or directly from a CT scanner, analyze the images and provide output annotations regarding lung nodules. Users are an existing PACS system to view the annotations. Dedicated servers can be located at hospitals and are directly connected to the hospital networks. The software consists of 4 modules which are Image reception (Docking Toolbox), Image predictive processing (DLServer), Image storage (RePACS) and Image display (NeoViewer).

AI/ML Overview

Here's a breakdown of the acceptance criteria and study details for InferRead Lung CT.AI, based on the provided document:

1. Table of Acceptance Criteria and Reported Device Performance

The document primarily focuses on a comparative effectiveness study and discusses the device's performance in comparison to unaided human reading. It doesn't explicitly list "acceptance criteria" with numerical targets in the same way a standalone performance study might. However, the objective of the clinical study serves as the de facto acceptance criteria.

Acceptance Criteria (Inferred from Study Objective)	Reported Device Performance
Significantly improve radiologists' nodule detection performance (AUC)	Increase in AUC (Aided - Unaided): 0.073 (95% CI: 0.020, 0.125). The document states this increase was "significant," indicating that the lower bound of the confidence interval (0.020) is above zero, satisfying the improvement criterion.
Without significantly increasing reading time	Decrease in reading times (Aided - Unaided): -23 seconds (95% CI: -42, -3). The document states this decrease was "significant," meaning the upper bound of the confidence interval (-3) is below zero. This indicates a reduction in reading time, thus satisfying the criterion of not increasing reading time and indeed improving it.

2. Sample Size and Data Provenance for Test Set

Sample Size (Test Set): 249 CT scans.
Data Provenance: The document does not explicitly state the country of origin. It specifies that the data included "chest CT scans from patients who underwent lung cancer screening," implying it's clinical data, and the study was "retrospective." This suggests the data was collected from existing patient records.

3. Number of Experts and Qualifications for Ground Truth

The document mentions that a "pivotal reader study" was conducted, involving "10 board-certified radiologists." These radiologists were part of the MRMC study, where their consensus or interpretations would contribute to the ground truth.
However, it does not explicitly state how many of these, or other, experts were specifically used to establish the definitive ground truth for the test set independent of the reader study itself. The ground truth for the reader study is the consensus of the readers, or a reference standard against which their performance is measured (see point 7).

4. Adjudication Method for the Test Set

The document describes a "fully crossed, multi-reader multi-case (MRMC) study." In such studies, all readers review all cases. While it doesn't explicitly state an adjudication method like "2+1" for establishing a separate ground truth, the MRMC setup inherently uses the collective performance of the expert readers (in both aided and unaided modes) to evaluate the device's impact. The ground truth for nodule presence/absence in the cases would have been established prior to the reader study.

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

Yes, a MRMC comparative effectiveness study was done.

Effect Size of Human Readers Improvement with AI vs. without AI:
- Nodule Detection Performance (AUC): The AUC increased by 0.073 (Aided - Unaided), with a 95% confidence interval of (0.020, 0.125). This indicates a statistically significant improvement in detection performance when radiologists used the InferRead Lung CT.AI device.
- Reading Time: Reading times decreased by 23 seconds (Aided - Unaided), with a 95% confidence interval of (-42, -3). This indicates a statistically significant reduction in reading time.

6. Standalone Performance Study (Algorithm Only)

Yes, a standalone performance study was done.

The document states: "Standalone performance testing which included chest CT scans from patients who underwent lung cancer screening was performed to validate detection accuracy of InferRead Lung CT.AI. Results showed that InferRead Lung CT.AI had similar nodule detection sensitivity and FP/scan compared to those of the predicate device."
This suggests comparison against the predicate device's standalone performance, which also implies a form of quantitative performance metric (sensitivity, false positives per scan) for the algorithm in isolation.

7. Type of Ground Truth Used

For the standalone performance study, the document mentions "detection accuracy" based on scans from lung cancer screening, but doesn't explicitly state whether the ground truth was expert consensus, pathology, or outcomes data. However, for nodule detection, expert consensus on the presence and location of nodules from expert radiologists is a common ground truth, often verified or refined.
For the MRMC study, the ground truth for evaluating individual reader performance (from which the AUC is derived) would typically be an established reference standard (often expert consensus, sometimes supplemented by follow-up or pathology if available for some cases) created prior to the readers' evaluations.

8. Sample Size for the Training Set

The document does not provide the sample size of the training set used for developing the InferRead Lung CT.AI algorithm.

9. How Ground Truth for Training Set Was Established

The document does not provide information on how the ground truth for the training set was established.

Summary

{0}------------------------------------------------

Image /page/0/Picture/0 description: The image contains the logos of the U.S. Department of Health & Human Services and the U.S. Food & Drug Administration (FDA). The Department of Health & Human Services logo is on the left and features a stylized caduceus. The FDA logo is on the right and consists of the FDA acronym in a blue square, followed by the words "U.S. FOOD & DRUG" in large, bold, blue letters, with the word "ADMINISTRATION" underneath in smaller letters.

Beijing Infervision Technology Co., Ltd. % Matt Deng 1900 Market St., 8th Floor PHILADELPHIA, PA 19103

July 2, 2020

Re: K192880

Trade/Device Name: InferRead Lung CT.AI Regulation Number: 21 CFR 892.2050 Regulation Name: Picture archiving and communications system Regulatory Class: Class II Product Code: OEB, LLZ Dated: June 2, 2020 Received: June 3, 2020

Dear Matt Deng:

We have reviewed your Section 510(k) premarket notification of intent to market the device referenced above and have determined the device is substantially equivalent (for the indications for use stated in the enclosure) to legally marketed predicate devices marketed in interstate commerce prior to May 28, 1976, the enactment date of the Medical Device Amendments, or to devices that have been reclassified in accordance with the provisions of the Federal Food, Drug, and Cosmetic Act (Act) that do not require approval of a premarket approval application (PMA). You may, therefore, market the device, subject to the general controls provisions of the Act. Although this letter refers to your product as a device, please be aware that some cleared products may instead be combination products. The 510(k) Premarket Notification Database located at https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/pmn.cfm identifies combination product submissions. The general controls provisions of the Act include requirements for annual registration, listing of devices, good manufacturing practice, labeling, and prohibitions against misbranding and adulteration. Please note: CDRH does not evaluate information related to contract liability warranties. We remind you, however, that device labeling must be truthful and not misleading.

If your device is classified (see above) into either class II (Special Controls) or class III (PMA), it may be subject to additional controls. Existing major regulations affecting your device can be found in the Code of Federal Regulations, Title 21, Parts 800 to 898. In addition, FDA may publish further announcements concerning your device in the Federal Register.

Please be advised that FDA's issuance of a substantial equivalence determination does not mean that FDA has made a determination that your device complies with other requirements of the Act or any Federal statutes and regulations administered by other Federal agencies. You must comply with all the Act's requirements, including, but not limited to: registration and listing (21 CFR Part 807); labeling (21 CFR Part 801); medical device reporting of medical device-related adverse events) (21 CFR 803) for

{1}------------------------------------------------

devices or postmarketing safety reporting (21 CFR 4, Subpart B) for combination products (see https://www.fda.gov/combination-products/guidance-regulatory-information/postmarketing-safety-reportingcombination-products); good manufacturing practice requirements as set forth in the quality systems (OS) regulation (21 CFR Part 820) for devices or current good manufacturing practices (21 CFR 4, Subpart A) for combination products; and, if applicable, the electronic product radiation control provisions (Sections 531-542 of the Act); 21 CFR 1000-1050.

Also, please note the regulation entitled, "Misbranding by reference to premarket notification" (21 CFR Part 807.97). For questions regarding the reporting of adverse events under the MDR regulation (21 CFR Part 803), please go to https://www.fda.gov/medical-device-safety/medical-device-reportingmdr-how-report-medical-device-problems.

For comprehensive regulatory information about medical devices and radiation-emitting products, including information about labeling regulations, please see Device Advice (https://www.fda.gov/medicaldevices/device-advice-comprehensive-regulatory-assistance) and CDRH Learn (https://www.fda.gov/training-and-continuing-education/cdrh-learn). Additionally, you may contact the Division of Industry and Consumer Education (DICE) to ask a question about a specific regulatory topic. See the DICE website (https://www.fda.gov/medical-device-advice-comprehensive-regulatoryassistance/contact-us-division-industry-and-consumer-education-dice) for more information or contact DICE by email (DICE@fda.hhs.gov) or phone (1-800-638-2041 or 301-796-7100).

Sincerely,

For

Thalia T. Mills, Ph.D. Director Division of Radiological Health OHT7: Office of In Vitro Diagnostics and Radiological Health Office of Product Evaluation and Quality Center for Devices and Radiological Health

Enclosure

{2}------------------------------------------------

Indications for Use

510(k) Number (if known)

K192880

Device Name InferRead Lung CT.AI

Indications for Use (Describe)

Type of Use (Select one or both, as applicable)

☒ Prescription Use (Part 21 CFR 801 Subpart D)
-------------------------------------------------------------------------------------------------------------------------

☐ Over-The-Counter Use (21 CFR 801 Subpart C)
------------------------------------------------------------------------------------------------------------------------

CONTINUE ON A SEPARATE PAGE IF NEEDED.

This section applies only to requirements of the Paperwork Reduction Act of 1995.

DO NOT SEND YOUR COMPLETED FORM TO THE PRA STAFF EMAIL ADDRESS BELOW.

The burden time for this collection of information is estimated to average 79 hours per response, including the time to review instructions, search existing data sources, gather and maintain the data needed and complete and review the collection of information. Send comments regarding this burden estimate or any other aspect of this information collection, including suggestions for reducing this burden, to:

Department of Health and Human Services Food and Drug Administration Office of Chief Information Officer Paperwork Reduction Act (PRA) Staff PRAStaff(@fda.hhs.gov

"An agency may not conduct or sponsor, and a person is not required to respond to, a collection of information unless it displays a currently valid OMB number."

{3}------------------------------------------------

Summary of 510(k)

Beijing Infervision Technology Co., Ltd. K192880

This 510(k) Summary is in conformance with 21CFR 807.92

Submitter:	Beijing Infervision Technology Co., Ltd.Room B401, 4th Floor, Building 1,No.12 Shangdi Information Road.Haidian District, Beijing, 100085Phone: +86 10-86462323
Primary Contact:	Mr. Matt DengEmail: dyufeng@infervision.comPhone: 919-491-5457
Company Contact:	Xiaoyan FanProject Leader
Date Prepared:	July 1, 2020

Device Name and Classification

Trade Name:	InferRead Lung CT.AI
Classification:	Class II
Regulation Number:	21 CFR 892.2050, Picture archiving and communications system
Classification Panel:	Radiology
Product Code:	OEB, LLZ

Predicate Device:

Trade Name	ClearRead CT
Classification	Class II
510(k) Number	K161201
Regulation Number	21 CFR 892.2050, Picture archiving and communications system
Classification Panel	Radiology Panel
Product Code	OEB, LLZ

{4}------------------------------------------------

Device Description

Indications for Use

InferRead Lung CT.AI is comprised of computer assisted to aid the radiologist in the detection of pulmonary nodules during the review of CT examinations of the chest on an asymptomatic population. InferRead Lung CT.AI requires that both lungs be in the field of view. InferRead Lung CT.Al provides adjunctive information and is not intended to be used without the original CT series.

Risk Analysis Method

The InferRead Lung CT.AI was assessed to determine risks to health associated with the use of the device. Risks related to safety and usability were considered. A risk analysis was conducted in accordance with ISO 14971:2007, Medical devices – Application of risk management to medical devices. Several risks were assessed, including, but not limited to device malfunction and improper use.

Substantial Equivalence

InferRead Lung CT.AI is substantially equivalent to the ClearRead CT (K161201) currently on the market.

The table below provides a detailed comparison of InferRead Lung CT.AI to the predicate device.

Item	InferRead Lung CT.AI(Subject Device)	ClearRead CT (K161201)(Predicate Device)	Comparison
Indications for Use	InferRead Lung CT.AI iscomprised of computerassisted reading toolsdesigned to aid theradiologist in the detectionof pulmonary nodules	ClearRead CT™ iscomprised of computerassisted reading toolsdesigned to aid theradiologist in the detection ofpulmonary nodules during	The indications for use of InferRead LungCT.AI are identical to the indications foruse of the previously cleared ClearReadCT.

	during the review of CT examinations of the chest on an asymptomatic population. 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.	review of CT examinations of the chest on an asymptomatic population. The ClearRead CT requires both lungs be in the field of view. ClearRead CT provides adjunctive information and is not intended to be used without the original CT series.
Intended Use	Computer assisted reading tools designed to aid the radiologist in the detection of pulmonary nodules during review of CT examinations of the chest.	Computer assisted reading tools designed to aid the radiologist in the detection of pulmonary nodules during review of CT examinations of the chest.	The intended use of InferRead Lung CT.AI is identical to the intended use of the previously cleared ClearRead CT.
Accessories/ToolsRequired by theUser (Platform)	Must be used in conjunction with a PACS system or an Image Viewer that reads DICOM images.	Must be used in conjunction with a PACS system or an Image Viewer that reads DICOM images.	The accessories required by the user for InferRead Lung CT.AI are identical to the accessories required by the user for the previously cleared ClearRead CT.
User AccessPoint	Post Processing Application	Post Processing Application	The user access point of InferRead Lung CT.AI is identical to the user access point of the previously cleared ClearRead CT.
Image Input	DICOM	DICOM	The image input of InferRead Lung CT.AI is identical to the image input of the previously cleared ClearRead CT.
Type of Scans	CT	CT	The type of scans for InferRead Lung CT.AI are identical to the type of scans for the previously cleared ClearRead CT.
AutomaticallyLocate andIdentify Lung	Yes	Yes	The function of automatically locating and identifying lung nodules for InferRead Lung CT.AI is identical to the function of

			lung nodules for the previously clearedClearRead CT.
Modifies theOriginal CTScan	No	Yes	According to the device description of thepredicate device found in its 510Ksummary, "ClearRead CT is a dedicatedpost-processing application that generatesa secondary vessel suppressed Lung CTseries with CADe marks and associatedregion descriptors intended to aid theradiologist in the detection of pulmonarynodules." ClearRead CT modifies theoriginal scan by performing vesselsuppression. On the contrary, InferReadLung CT.AI does not modify the originalscan, only shows the locations ofpulmonary nodules.This difference does not affect theintended use or safety and effectiveness ofthe device.
Requires aDisjointComparison withthe Original CTScan	No	Yes	Since ClearRead CT creates a secondaryvessel suppressed Lung CT series, itrequires the user to have an original CTseries on a separate window. There are 2series open at the same time. We refer tothis setup as "disjoint comparison". On thecontrary, InferRead Lung CT.AI does notrequire 2 series open, as its CADe marksoverlay with the original CT scan and canbe toggled on and off. Therefore,InferRead Lung CT.AI does not require"disjoint comparison"This difference does not affect theintended use or safety and effectiveness ofthe device.
			InferRead Lung CT.AI display the originalCT scans.
Nodule Marking	A bounding box is providedaround nodules	A bounding box is providedaround nodules	The function of providing a bounding boxaround nodules for InferRead Lung CT.AIis identical to the indications for use of thepreviously cleared ClearRead CT.
Provides NoduleCharacteristics	Yes, the maximum axialplane longest diameter,mean diameter and volumeinformation are provided.	Yes, the volume, maximumaxial plane diameter,minimum axial planediameter, and averagedensity in Hounsfield unitsare provided. (from website)	InferRead Lung CT.AI has mean diametermeasurement function that is not providedby predicate device. Mean diameter is theaverage of maximum axial plane diameterand minimum axial plane diameter. Andthis difference does not affect the safetyand effectiveness of the device.
DetectionTarget(s)	Solid, Sub Solid (part solidand ground glass) nodules	Solid, Sub Solid (part solidand ground glass) nodules	The detection targets of InferRead LungCT.AI are identical to the detection targetsof the previously cleared ClearRead CT.They have the same definition principlesfor actionable nodules classification.
Size of DetectionTargets	4mm and above, supportsvisualization of nodulessmaller than 4mm	5mm and above, supportsvisualization of nodulessmaller than 5mm	The InferRead Lung CT.AI detects smallernodules. This difference does not affectthe intended use or safety andeffectiveness of the device. This differencehas been addressed with the completion ofstand-alone performance characteristicstesting.

Detailed Comparison of the Subject and Predicate Devices

Infervision InferRead Lung CT.AI Traditional 510(k)

{5}------------------------------------------------

{6}------------------------------------------------

{7}------------------------------------------------

{8}------------------------------------------------

Testing Summary

Non-clinical performance evaluation

Software testing was performed in accordance with General Principles of Software Validation; Final Guidance for Industry and FDA Staff (January 11, 2002). Software testing which included unit testing, software integration testing and software system testing was performed on InferRead Lung CT.AI. It was demonstrated that InferRead Lung CT.AI, when used according to operating instructions, met all requirement specifications. All system functionalities were tested and passed. Measurement performance was validated on phantom and clinical data to assess reproducibility and accuracy. Standalone performance testing which included chest CT scans from patients who underwent lung cancer screening was performed to validate detection accuracy of InferRead Lung CT.AI. Results showed that InferRead Lung CT.AI had similar nodule detection sensitivity and FP/scan compared to those of the predicate device. Based on the results of verification and validation tests it is concluded that InferRead Lung CT.AI is effective and safe in the detection of nodules.

Clinical performance evaluation

A pivotal reader study which was a retrospective, fully crossed, multi-reader multi-case (MRMC) study was conducted to validate that the device conformed to the defined user needs and intended uses. A total of 10 board-certified radiologists and a collection of 249 scans were involved in the reader study. The purpose of the reader study was to validate that with the aided of InferRead Lung CT.AI radiologists' nodule detection performance could significantly improve without significantly increasing reading time at a significance level alpha of 0.05 (two-sided). The reader study measured the area under the curve (AUC) of the localization receiver operating characteristic (LROC) response when using InferRead Lung CT.AI relative to the unaided read. The study also measured the radiologists' interpretation time when using InferRead Lung CT.AI relative to unaided interpretations. Results showed that InferRead Lung CT.AI was found to significantly increase the AUC (Aided - Unaided: 0.073, 95%CI: 0.020, 0.125), indicating the detection performance through using the device is superior to the unaided read for detecting nodules. Moreover, InferRead Lung CT.AI was also found to decrease reading times (Aided - Unaided: -23s, 95%Cl: -42, -3). In conclusion, the pivotal study showed that with the aided of InferRead Lung CT.AI radiologists' nodule detection performance could significantly improve without significantly increasing reading time.

Substantial Equivalence Conclusions

In conclusion, the intended use for InferRead Lung CT.AI is the same as that of the previously cleared ClearRead CT (K161201). The technological characteristics demonstrate that the InferRead Lung CT.AI is substantially equivalent to the previously cleared ClearRead CT (K161201), and the testing shows that the InferRead Lung CT.AI is substantially equivalent to the previously cleared ClearRead CT (K161201) and assures that the InferRead Lung CT.AI is as safe and effective as the previously cleared ClearRead CT (K161201).

Conclusion

{9}------------------------------------------------

The 510(k) Pre-market Notification for InferRead Lung CT.AI contains adequate information and data to determine that InferRead Lung CT.AI is as safe and effective as the legally marketed predicate device.

Regulation Number and Section

§ 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).