PowerLook® Tomo Detection V2 Software is a computer-assisted detection and diagnosis (CAD) software device intended to be used concurrently by interpreting physicians while reading digital breast tomosynthesis (DBT) exams from compatible DBT systems. The system detects soft tissue densities (masses, architectural distortions and asymmetries) and calcifications in the 3D DBT slices. The detections and Certainty of Finding and Case Scores assist interpreting physicians in identifying soft tissue densities and calcifications that may be confirmed or dismissed by the interpreting physician.

Device Description

PLTD V2 detects malignant soft-tissue densities and calcifications in digital breast tomosynthesis (DBT) image. The PLTD V2 software allows a interpreting physician to quickly identify suspicious soft tissue densities and calcifications by marking the detected areas in the tomosynthesis images. When the PLTD V2 marks are displayed by a user, the marks will appear as overlays on the tomosynthesis images. The PLTD V2 marks also serve as a navigation tool for users, because each mark is linked to the tomosynthesis plane where the detection was identified. Users can navigate to the plane associated with each mark by clicking on the detection mark. Each detected region will also be assigned a "score" that corresponds to the PLTD V2 algorithm's confidence that the detected region is a cancer (Certainty of Finding Score). Certainty of Finding scores are relative scores assigned to each detected region and a Case Score is assigned to each case regardless of the number of detected regions. Certainty of Finding and Case Scores are computed by the PLTD V2 algorithm and represent the algorithm's confidence that a specific finding or case is malignant. The scores are represented on a 0% to 100% scale. Higher scores represent a higher algorithm confidence that a finding or case is malignant. Lower scores represent a lower algorithm confidence that a finding or case is malignant.

AI/ML Overview

Here's a breakdown of the acceptance criteria and the study proving the device meets them, based on the provided text.

1. Acceptance Criteria and Reported Device Performance

The device is a Computer-Assisted Detection and Diagnosis (CAD) software for digital breast tomosynthesis (DBT) exams. The acceptance criteria are largely demonstrated through the multi-reader multi-case (MRMC) pivotal reader study and standalone performance evaluations.

Table of Acceptance Criteria and Reported Device Performance:

Criteria Category	Metric	Acceptance Criteria (Implied / Stated)	Reported Device Performance (with CAD vs. without CAD)
Pivotal Reader Study (Human-in-the-Loop)
Radiologist Performance	Case-level Area Under the Receiver Operating Characteristic (ROC) Curve (AUC)	Non-inferiority to radiologist performance without CAD. Implicit superiority is also a desirable outcome.	AUC with CAD: 0.852 AUC without CAD: 0.795 Average difference: 0.057 (95% CI: 0.028, 0.087); p < 0.01 (non-inferiority and difference)
Radiologist Reading Time	Reading Time	Superiority (shorter time) compared to reading without CAD.	Reading time improved 52.7% with CAD (95% CI: 41.8%, 61.5%; p < 0.01).
Sensitivity (Case-Level)	Case-level Sensitivity	Non-inferiority (margin delta = 0.05). Implicit superiority is also a desirable outcome.	Average case-level sensitivity with CAD: 0.850 Average case-level sensitivity without CAD: 0.770 Average increase: 0.080 (95% CI: 0.026, 0.134); p < 0.01 (non-inferiority and difference)
Sensitivity (Lesion-Level)	Lesion-level Sensitivity	Non-inferiority (margin delta = 0.05). Implicit superiority is also a desirable outcome.	Average per-lesion sensitivity with CAD: 0.853 Average per-lesion sensitivity without CAD: 0.769 Average increase: 0.084 (95% CI: 0.029, 0.139); p < 0.01 (non-inferiority and difference)
Specificity (Case-Level)	Case-level Specificity	Non-inferiority (margin delta = 0.05).	Specificity with CAD: 0.696 Specificity without CAD: 0.627 Average increase: 0.069 (95% CI: 0.030, 0.108); p < 0.01 (non-inferiority)
Recall Rate (Case-Level)	Recall Rate in Non-Cancer Cases	Non-inferiority (margin delta = 0.05). Lower recall rates are better.	Average recall rate with CAD: 0.309 Average recall rate without CAD: 0.380 Average reduction: 0.072 (95% CI: 0.031, 0.112); p < 0.01 (non-inferiority)
Standalone Performance (Algorithm Only)
Hologic DBT (Screening)	Case-Level Sensitivity, Lesion-Level Sensitivity, FP Rate in Non-Cancer Cases, Specificity	Met design specifications (details in User Manual, not specified here).	Met design specifications (specific values not provided in document, but deemed acceptable).
GE DBT (Screening)	Case-Level Sensitivity, Lesion-Level Sensitivity, FP Rate in Non-Cancer Cases, Specificity	Met design specifications (details in User Manual, not specified here).	Met design specifications (specific values not provided in document, but deemed acceptable).
GE vs. Hologic DBT	Sensitivity, FPPI, AUC (non-inferiority)	Non-inferiority between GE DBT and Hologic DBT performance within specified margins.	Differences in Sensitivity, FPPI, and AUC were within the margin of non-inferiority (specific values not provided, but deemed acceptable).

2. Sample Sizes and Data Provenance

Pivotal Reader Study (Test Set):
- Sample Size: 260 Hologic digital breast tomosynthesis (DBT) cases. This included 65 cancer cases with 66 malignant lesions.
- Data Provenance: Retrospective. The country of origin is not explicitly stated, but given the FDA submission, it's likely primarily from the US or regions with similar regulatory standards.
Standalone Performance Studies (Test Set):
- Hologic DBT: 655 Hologic DBT cases, including 235 cancer cases with 242 malignant lesions.
- GE DBT: 610 GE DBT cases, including 204 cancer cases with 221 malignant lesions.
- Data Provenance: Retrospective. No country of origin specified.
Training Set Sample Size:
- The document does not explicitly state the sample size for the training set. It mentions the algorithm uses "deep learning technology," implying a substantial training set would have been used.

3. Number of Experts and Qualifications for Test Set Ground Truth

Pivotal Reader Study & Standalone Studies: The document does not explicitly state the number of experts used to establish the ground truth for the test sets. However, for the pivotal reader study, it states: "The purpose of the pivotal study was to compare clinical performance of radiologists using CAD... to that of radiologists using DBT without CAD." This implies the radiologists in the study were the interpreting physicians who would typically establish ground truth in a clinical setting based on their diagnostic judgment and follow-up (biopsy, etc.). It mentions that the ground truth for cancer cases required "malignant lesion localization."
Qualifications of Experts: The readers in the pivotal study were "24 tomosynthesis radiologist readers." While specific experience levels (e.g., "10 years of experience") are not provided, "radiologist" implies medical doctors specialized in radiology, typically with several years of post-medical school training and board certification.

4. Adjudication Method for Test Set

The document does not explicitly describe an adjudication method (e.g., 2+1, 3+1 consensus) for establishing the ground truth for the test sets. It implies that the ground truth for malignancy was established via biopsy or other confirmed pathology (referred to as "malignant lesions" and "cancer cases"). For the reader study, the "ground truth" for evaluating reader performance was the confirmed cancer status of the cases and lesions.

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

Yes, an MRMC comparative effectiveness study was done. This was the "Pivotal Reader Study."
Effect Size of Human Reader Improvement with AI vs. Without AI Assistance:
- AUC: Radiologists' per-subject average AUC improved by 0.057 when using CAD (0.852 with CAD vs. 0.795 without CAD).
- Reading Time: Reading time improved 52.7% with CAD.
- Case-Level Sensitivity: Average sensitivity increased by 0.080 (0.850 with CAD vs. 0.770 without CAD).
- Lesion-Level Sensitivity: Average sensitivity increased by 0.084 (0.853 with CAD vs. 0.769 without CAD).
- Case-Level Specificity: Average specificity increased by 0.069 (0.696 with CAD vs. 0.627 without CAD).
- Recall Rate in Non-Cancer Cases: Average recall rate was reduced by 0.072 (0.309 with CAD vs. 0.380 without CAD).

6. Standalone Performance (Algorithm Only without Human-in-the-Loop Performance)

Yes, standalone studies were done.
- A standalone study was conducted for Hologic DBT cases (N=655).
- A standalone study was conducted for GE DBT cases (N=610).
- A comparison of performance between GE DBT and Hologic DBT was also performed to show non-inferiority.

7. Type of Ground Truth Used

The ground truth used for the cancer cases (in both standalone and reader studies) appears to be pathology-confirmed malignancy (referred to as "cancer cases" and "malignant lesions"). The document states that "malignant lesion localization was required for a reader to correctly detect cancer in a case," implying confirmed cancerous findings.

8. Sample Size for the Training Set

The document does not specify the sample size for the training set. It mentions the use of "deep learning technology," which typically requires large datasets for training.

9. How Ground Truth for the Training Set Was Established

The document does not explicitly state how the ground truth for the training set was established. Given it's a deep learning-based CAD system for cancer detection, it's highly probable that ground truth for the training data would also have been established through pathology-confirmed diagnoses (biopsy results) and/or by expert radiologists' review and consensus with prolonged follow-up to confirm benign status.

Summary

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

Image /page/0/Picture/0 description: The image shows the logo of the U.S. Food and Drug Administration (FDA). The logo consists of two parts: the Department of Health and Human Services logo on the left and the FDA logo on the right. The FDA logo is a blue square with the letters "FDA" in white, followed by the words "U.S. Food & Drug Administration" in blue.

December 6, 2018

iCAD Inc. John Delucia VP, Regulatory Affairs, Clinical Affairs and Quality Assurance 98 Spit Brook Rd. Suite 100 NASHUA, NH 03062

Re: K182373

Trade/Device Name: PowerLook® Tomo Detection V2 Software Regulation Number: 21 CFR 892.2090 Regulation Name: Radiological Computer Assisted Detection and Diagnosis Software Regulatory Class: Class II Product Code: ODO Dated: November 6, 2018 Received: November 7, 2018

Dear Mr. Delucia:

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.

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

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 devices or postmarketing safety reporting (21 CFR 4, Subpart B) for combination products (see https://www.fda.gov/CombinationProducts/GuidanceRegulatoryInformation/ucm597488.htm); good manufacturing practice requirements as set forth in the quality systems (QS) 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 http://www.fda.gov/MedicalDevices/Safety/ReportaProblem/default.htm.

For comprehensive regulatory information about mediation-emitting products, including information about labeling regulations, please see Device Advice (https://www.fda.gov/MedicalDevices/DeviceRegulationandGuidance/) and CDRH Learn (http://www.fda.gov/Training/CDRHLearn). 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 (http://www.fda.gov/DICE) for more information or contact DICE by email (DICE@fda.hhs.gov) or phone (1-800-638-2041 or 301-796-7100).

Sincerelv.

Hse 2. Nils

for Robert Ochs, Ph.D. Director Division of Radiological Health Office of In Vitro Diagnostics and Radiological Health Center for Devices and Radiological Health

Enclosure

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

Indications for Use

510(k) Number (if known) K182373

Device Name PowerLook® Tomo Detection V2 Software

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}------------------------------------------------

SECTION 1: 510(K) Summary

Date Prepared: December 5, 2018

Submitter:

iCAD, Inc. 98 Spit Brook Road Suite 100 Nashua, NH 03062

Contact Person:

John DeLucia VP, Regulatory Affairs, Clinical Affairs and Quality Assurance Email: jdelucia@icadmed.com Phone: (603) 309-1945 Fax: (603) 880-3043

Submission Date:

August 30, 2018

Trade Name:

PowerLook® Tomo Detection V2 software

Common Name:

Medical Imaging Software

Classification Name/Regulation

Radiological Computer Assisted Detection and Diagnosis Software 21 CFR 892.2090

Product Code

QDQ

Device Classification

Class II

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

Legally Marketed Devices to Which Substantial Equivalence is Claimed

PowerLook Tomo Detection (PLTD) V2 Software is substantially equivalent to the following legally marketed predicate device.

Device Name	Manufacturer	FDA DENReference #	Decision Date
OsteoDetect	Imagen	DEN180005	May 24, 2018

Device Description

Key Outputs of PowerLook® Tomo Detection V2 Software

a. Lesion Detection

PowerLook® Tomo Detection V2 software detects soft tissue densities (masses, architectural distortions and asymmetries) and calcifications in the 3D digital breast tomosynthesis images. The PowerLook Tomo Detection V2 algorithm uses deep learning technology to process feature computations and uses pattern recognition to identify suspicious breast lesions appearing as soft tissue densities or clusters of calcifications. Each detected region in the tomosynthesis data is identified or represented by marking the contour of the lesion in the tomosynthesis slice where it was detected.

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

b. Certainty of Finding and Case Scores

Certainty of Finding scores are relative scores assigned to each detected region and a Case Score is assigned to each case regardless of the number of detected regions. Certainty of Finding and Case Scores are computed by the PowerLook Tomo Detection V2 algorithm and represent the algorithm's confidence that a specific finding or case is malignant. The scores are represented on a 0% to 100% scale. Higher scores represent a higher algorithm confidence that a finding or case is malignant. Lower scores represent a lower algorithm confidence that a finding or case is malignant. The scores are based on a population with 50% prevalence of cancer and should be interpreted as the probability of the finding or case correctly being identified as malignant in a population of 50% cancers and 50% non-cancers. The scores serve as a guide to interpreting physicians to aid in determining if a suspicious finding or case needs further work-up. These scores are not intended to be the clinically used "probability of malignancy". Certainty of Finding and Case Scores are not calibrated to the prevalence in the intended use population or to the prevalence in the pivotal reader study outlined in the Assessment of Clinical Performance Data section, and consequently, the Certainty of Finding and Case Scores are in general higher than the actual probability of malignancy in an intended use population with less than 50% prevalence. These scores represent a relative level of concern or level of suspicion because they do not represent an absolute clinical probability of malignancy.

Supported Digital Breast Tomosynthesis Systems

The following Digital Breast Tomosynthesis systems have been tested and are compatible with PowerLook Tomo Detection V2 software:

. Hologic Selenia Dimensions
. GE Pristina

The PowerLook Tomo Detection V2 software is designed as a stand-alone executable operating within the larger software framework provided by PowerLook AMP1.

Intended Use / "Indications for Use"

PowerLook® Tomo Detection V2 software is a computer-assisted detection and diagnosis (CAD) software device intended to be used concurrently by interpreting physicians while reading digital breast tomosynthesis

1 iCAD PowerLook AMP is a Class I medical device exempt per 21 CFR § 892.2010 (Medical image storage device) and 21 CFR § 892.2020 (Medical image communications device). The PowerLook AMP is a device that provides electronic storage functions for medical images and also provides electronic transfer of medical image data between medical devices

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

(DBT) exams from compatible DBT systems. The system detects soft tissue densities (masses, architectural distortions and asymmetries) and calcifications in the 3D DBT slices. The detections and Certainty of Finding and Case Scores assist interpreting physicians in identifying soft tissue densities and calcifications that may be confirmed or dismissed by the interpreting physician.

Comparison with Predicate Device

The Summary of Substantial Equivalence Table below details the similarities and differences between the subject device. PowerLook Tomo Detection V2. and the predicate device, OsteoDetect. classified under De Novo request DEN180005.

PowerLook Tomo Detection V2 and the predicate device have the same indication for use. Both devices are intended to aid in the detection, localization, and characterization of disease specific findings on acquired medical images, per 21 CFR 892.2090.

From a technology standpoint, PowerLook Tomo Detection V2 is the same as the predicate per 21 CFR 892.2090; both devices are radiological computer assisted detection and diagnostic software, intended to aid in the detection, and characterization of disease specific findings on acquired medical images. The outputs of both devices serve as a concurrent read. The output is used to inform the clinical user (who themselves make the primary diagnostic and patient management decisions) and will not replace the clinical expertise and judgment of the clinical user. Where any differences may occur between the subject device and the predicate, special controls established for Radiological Computer Assisted Detection and Diagnosis Software are in place to further mitigate any risks in these differences.

Summary of Indications for Use

The subject device and the predicate device are both classification name and number: Radiological Computer Assisted Detection and Diagnosis Software 21 CFR 892.2090. Both devices are intended to aid in the detection, and characterization of disease specific findings on acquired medical images.

The subject device and the predicate may differ in their detection of disease specific findings however, special controls are in place to mitigate any risk for this difference.

Summary of Technological Characteristic

The technological characteristics of PowerLook Tomo Detection V2 are the same as the predicate OsteoDetect classified under De Novo request DEN180005. Per 21 CFR 892.2090, both devices are radiological computer assisted detection and diagnostic software intended to aid in the detection, localization, and characterization of disease specific findings on acquired medical

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

images. The outputs of both devices serve as a concurrent read. The output is used to inform the clinical user (who themselves make the primary diagnostic and patient management decisions) and will not replace the clinical expertise and judgment of the clinical user.

General Safety and Effectiveness Concerns

The device labeling contains instructions for use and any necessary cautions and warnings to provide for safe and effective use of this device. Risk management is ensured via a risk analysis which is used to identify and mitigate potential hazards. Any potential hazards are controlled via software development, verification and validation testing. In addition, general controls of the FD&C Act, and special controls established for Radiological Computer Assisted Detection and Diagnosis Software are in place to further mitigate any safety and or effectiveness risks.

Assessment of Non-Clinical Performance Data

PowerLook Tomo Detection V2 has been verified and validated according to iCAD's design control processes. All supporting documentation has been included in this 510(k) Premarket Notification. Verification activity included unit, integration, and system level testing. Validation testing included standalone testing and performing a pivotal reader study to compare the clinical performance of radiologists using CAD detections and Certainty of Finding and Case Scores from the PLTD V2 CAD system with DBT images to that of radiologists using DBT without CAD

. Standalone Performance on a Screening Population Dataset

Hologic DBT

A standalone study, which evaluated the performance of PLTD without a radiologist, was conducted with a sample of 655 Hologic DBT cases, including 235 cancer cases with 242 malignant lesions. These 655 cases were used in the standalone study with PLTD V2. A stratified bootstrap procedure was used to estimate over a screening patient population. The bootstrap procedure limits the number of cases in a particular category when computing performance measures.

The purpose of the standalone study was to assess the standalone performance of PLTD V2 on a screening population.

Results from the standalone study showed that Case-Level Sensitivity, Lesion-Level Sensitivity, FP Rate in Non-Cancer Cases, and Specificity met design specifications. The detailed results are in the User Manual.

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

GE DBT

A standalone study, which evaluated the performance of PLTD V2, was conducted with a sample of 610 GE DBT cases, including 204 cancer cases with 221 malignant lesions. These 610 cases were used in the standalone study PLTD V2. A stratified bootstrap procedure was used to estimate performance over a screening patient population. The bootstrap procedure limits the number of cases in a particular category when computing performance measures.

The purpose of the standalone study was to assess the standalone performance of PLTD V2 on a screening population.

Comparison of Performance Between GE DBT and Hologic DBT (Control Group)

A comparison was made between the standalone performance of PowerLook Tomo Detection 2.0 on Hologic DBT images and PowerLook Tomo Detection 2.0 on GE DBT images. For this comparison, the performance on Hologic is considered the control group and performance on GE is the test group. The test is to determine if the difference between the control group and the test group is within the margin of non-inferiority for Sensitivity and AUC, and FPPI.

Standalone testing was performed for the control group and the test group individually. Key performance measures were then compared by subtracting the test group performance from the control group performance. This was done in a way to produce not just an estimate of the mean difference but also a distribution of the expected differences. In order to show statistical significance, the two-sided 95% confidence interval boundaries must be within the margin of non-inferiority.

Three measures were used to compare the performance of PowerLook Tomo Detection 2.0 with GE DBT images to PowerLook Tomo Detection 2.0 with Hologic images. Each of the three measures produced differences that were within the margin of non-inferiority. Therefore, in the areas of Sensitivity, FPPI, and AUC, PowerLook Tomo Detection 2.0 with the GE DBT system is not inferior to PowerLook Tomo Detection 2.0 with the Hologic DBT system.

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

Assessment of Clinical Performance Data

. Pivotal Reader Study

A pivotal reader study, which was a retrospective, fully-crossed, multi-reader, multi-case (MRMC) study of iCAD's PowerLook® Tomo Detection (PLTD) V2 computer-assisted detection (CAD) system, was conducted with 24 tomosynthesis radiologist readers and an enriched sample of 260 Hologic digital breast tomosynthesis (DBT) cases, including 65 cancer cases with 66 malignant lesions. The purpose of the pivotal study was to compare clinical performance of radiologists using CAD detections and Certainty of Finding and Case Scores from the PLTD V2 CAD system with DBT images to that of radiologists using DBT without CAD. The results of the pivotal study will support a regulatory submission for the PLTD V2 CAD system.

The objectives of this pivotal reader study were the following:

A. Co-primary objectives. The co-primary objectives were to determine:

1. Whether radiologist performance when using CAD with DBT images is non-inferior to radiologist performance when using DBT images without CAD and
1. Whether radiologist reading time when using CAD with DBT images is superior to (shorter than) radiologist reading time when using DBT images without CAD.

Radiologist performance was assessed by measuring case-level area under the receiver operating characteristic (ROC) curve (AUC) for the detection of malignant lesions, where malignant lesion localization was required for a reader to correctly detect cancer in a case.

B. Secondary objectives. The secondary objectives of the pivotal reader study included the following for radiologists when using CAD with DBT compared to using DBT without CAD:

1. Superiority of case-level AUC
1. Non-inferiority (with non-inferiority margin delta = 0.05) of sensitivity at the case level
1. Superiority of sensitivity at the case level
1. Non-inferiority (with non-inferiority margin delta = 0.05) of sensitivity at the lesion level
1. Superiority of sensitivity at the lesion level
1. Non-inferiority (with non-inferiority margin delta = 0.05) of specificity (case-level)
1. Non-inferiority (with non-inferiority margin delta = 0.05) of recall rate in non-cancers (case-level)

{10}------------------------------------------------

The pivotal study results showed that both co-primary endpoints were met. Specifically, the pivotal study showed that:

1. Radiologist performance using CAD with DBT was non-inferior to, and statistically significantly superior to, radiologist performance using DBT without CAD. Radiologists had superior per-subject average area under the receiver operating characteristic (ROC) curve (AUC) with CAD, 0.852, versus without CAD, 0.795. The average difference in AUC was 0.057 (95% CI: 0.028, 0.087; non-inferiority p < 0.01 for non-inferiority margin delta = 0.05, and p < 0.01 for test of difference).
1. Radiologist reading time when using CAD with DBT is superior to (shorter than) radiologist reading time when using DBT without CAD. Reading time improved 52.7% with CAD (95% CI: 41.8%, 61.5%; p < 0.01). *

Interpreting physicians reading times may vary based on the specific functionality of the viewing application used for interpretation.

All pre-specified secondary endpoints also were met. In addition to superiority of case-level AUC, the pivotal study showed that:

Radiologists had superior sensitivity at the case level with CAD. Average sensitivity . increased by 0.080 (95% CI: 0.026, 0.134; non-inferiority p < 0.01 for non-inferiority margin delta = 0.05, and p < 0.01 for test of difference). Average case-level sensitivity was 0.770 without CAD and 0.850 with CAD.
At the lesion level, radiologists also had superior sensitivity with CAD. Average per-. lesion sensitivity across readers increased by 0.084 (95% CI: 0.029, 0.139; noninferiority p < 0.01 for non-inferiority margin delta = 0.05, and p < 0.01 for test of difference), from 0.769 without CAD to 0.853 with CAD.
Radiologists had non-inferior specificity with CAD. Specificity was 0.627 without CAD and 0.696 with CAD, for an average increase of 0.069 (95% CI: 0.030, 0.108: non-inferiority p < 0.01 for non-inferiority margin delta = 0.05).
Finally, radiologists had non-inferior recall rate in non-cancer cases with CAD. In non-● cancer cases, lower recall rates are better than higher recall rates. Average recall rate in non-cancer cases was 0.380 without CAD and 0.309 with CAD, for an average reduction of 0.072 (95% CI: 0.031, 0.112; non-inferiority p < 0.01 for non-inferiority margin delta = 0.05).

In this study the following were observed:

. Average sensitivity increased by 0.120 (SE=0.040) in the subgroup of 15 cancer cases with only calcifications.
. Average sensitivity increased by 0.068 (SE=0.031) in the subgroup of 50 cancer cases with at least one soft tissue density or mixed lesion.

{11}------------------------------------------------

Average specificity decreased by 0.027 (SE=0.038) in the subgroup of 24 benign and . recalled (non-cancer) cases with only calcifications.
Average specificity increased by 0.079 (SE=0.028) in the subgroup of 62 benign and ● recalled (non-cancer) cases with at least one soft tissue density or mixed lesion.
Average specificity increased by 0.084 (SE=0.021) in the subgroup of 109 non-cancer . cases with no lesions.

Conclusion:

Based upon the information presented in this submission, it is concluded that PowerLook® Tomo Detection V2 is substantially equivalent to the named predicate device and that the nonclinical and clinical tests that demonstrate that the device is as safe, as effective, and performs as well as or better than the legally marketed device

{12}------------------------------------------------

Summary of Substantial Equivalence:

Features andCharacteristics	Subject Device	Predicate Device	Discussion of Differences andComments
	iCAD Inc.PowerLook® Tomo Detection V2Software	Imagen Inc.OsteoDetectDEN180005
RegulationNumber/Name	21 CFR 892.2090 / Radiological Computer AssistedDetection and Diagnosis Software.	Same	NA
RegulationDescription	A radiological computer assisted detection and diagnosticsoftware is an image processing device intended to aid inthe detection, localization, and characterization of fracture,lesions, or other disease specific findings on acquiredmedical images (e.g. radiography, MR, CT). The devicedetects, identifies and characterizes findings based onfeatures or information extracted from images, andprovides information about the presence, location, andcharacteristics of the findings to the user. The analysis isintended to inform the primary diagnostic and patientmanagement decisions that are made by the clinical user.The device is not intended as a replacement for a completeclinician's review or their clinical judgment that takes intoaccount other relevant information from the image orpatient history.	Same	NA
Intended Use	PowerLook® Tomo Detection V2 software is a computer-assisted detection and diagnosis (CAD) software deviceintended to be used concurrently by interpreting physicianswhile reading digital breast tomosynthesis (DBT) examsfrom compatible DBT systems. The system detects softtissue densities (masses, architectural distortions andasymmetries) and calcifications in the 3D DBT slices. Thedetections and Certainty of Finding and Case Scores assistinterpreting physicians in identifying soft tissue densitiesand calcifications that may be confirmed or dismissed bythe interpreting physician.	OsteoDetect analyzes wrist radiographsusing machine learning techniques to identifyand highlight distal radius fractures duringthe review of posterior-anterior (PA) andlateral (LAT) radiographs of adult wrists.	Both devices have the sameintended use per 21 CFR892.2090.
Type of CADSoftware	Radiological computer assisted detection anddiagnostic software	Same	NA
Mode of Action	Image processing device intended to aid in thedetection, localization, and characterization of softtissue densities (masses, architectural distortionsand asymmetries) and calcifications in the 3DDBT slices.	Image processing device intended to aid inidentifying and highlighting distal radiusfractures during the review of posterior-anterior (PA) and lateral (LAT) radiographsof adult wrists.	Both devices are intended to aid in thedetection, localization, and characterizationof disease specific findings on acquiredmedical images per 21 CFR 892.2090.PLTD detects a different disease specificfinding. However, special controls are inplace to mitigate any risk for this difference.
Clinical Output	To inform the primary diagnostic and patientmanagement decisions that are made by theclinical user.	Same	The outputs of both devices serve as aconcurrent read. The output is used toinform the clinical user (who themselvesmake the primary diagnostic and patientmanagement decisions) and will not replacethe clinical expertise and judgment of theclinical user.
PatientPopulation	Symptomatic and asymptomatic womenundergoing mammography.	Adult men and women undergoingradiographs of adult wrists.	PLTD is intended for a different patientpopulation however special controls arein place to mitigate any risk for thisdifference.
End Users	Interpreting Physicians Radiologists	Clinicians	Same
Image SourceModalities	Digital breast tomosynthesis slices	2D X-ray images	The Image Source Modalities are different.However, special controls are in place tomitigate any risk for this difference.
Output Device	Softcopy Workstation	Softcopy Workstation, PACS, RIS	NA
Deployment	Stand-alone computer	Same	NA
Software Levelof Concern	Moderate	Moderate	NA

{13}------------------------------------------------

Regulation Number and Section

§ 892.2090 Radiological computer-assisted detection and diagnosis software.

(a)
Identification. A radiological computer-assisted detection and diagnostic software is an image processing device intended to aid in the detection, localization, and characterization of fracture, lesions, or other disease-specific findings on acquired medical images (e.g., radiography, magnetic resonance, computed tomography). The device detects, identifies, and characterizes findings based on features or information extracted from images, and provides information about the presence, location, and characteristics of the findings to the user. The analysis is intended to inform the primary diagnostic and patient management decisions that are made by the clinical user. The device is not intended as a replacement for a complete clinician's review or their clinical judgment that takes into account other relevant information from the image or patient history.(b)
Classification. Class II (special controls). The special controls for this device are:(1) Design verification and validation must include:
(i) A detailed description of the image analysis algorithm, including a description of the algorithm inputs and outputs, each major component or block, how the algorithm and output affects or relates to clinical practice or patient care, and any algorithm limitations.
(ii) A detailed description of pre-specified performance testing protocols and dataset(s) used to assess whether the device will provide improved assisted-read detection and diagnostic performance as intended in the indicated user population(s), and to characterize the standalone device performance for labeling. Performance testing includes standalone test(s), side-by-side comparison(s), and/or a reader study, as applicable.
(iii) Results from standalone performance testing used to characterize the independent performance of the device separate from aided user performance. The performance assessment must be based on appropriate diagnostic accuracy measures (
e.g., receiver operator characteristic plot, sensitivity, specificity, positive and negative predictive values, and diagnostic likelihood ratio). Devices with localization output must include localization accuracy testing as a component of standalone testing. The test dataset must be representative of the typical patient population with enrichment made only to ensure that the test dataset contains a sufficient number of cases from important cohorts (e.g., subsets defined by clinically relevant confounders, effect modifiers, concomitant disease, and subsets defined by image acquisition characteristics) such that the performance estimates and confidence intervals of the device for these individual subsets can be characterized for the intended use population and imaging equipment.(iv) Results from performance testing that demonstrate that the device provides improved assisted-read detection and/or diagnostic performance as intended in the indicated user population(s) when used in accordance with the instructions for use. The reader population must be comprised of the intended user population in terms of clinical training, certification, and years of experience. The performance assessment must be based on appropriate diagnostic accuracy measures (
e.g., receiver operator characteristic plot, sensitivity, specificity, positive and negative predictive values, and diagnostic likelihood ratio). Test datasets must meet the requirements described in paragraph (b)(1)(iii) of this section.(v) Appropriate software documentation, including device hazard analysis, software requirements specification document, software design specification document, traceability analysis, system level test protocol, pass/fail criteria, testing results, and cybersecurity measures.
(2) Labeling must include the following:
(i) A detailed description of the patient population for which the device is indicated for use.
(ii) A detailed description of the device instructions for use, including the intended reading protocol and how the user should interpret the device output.
(iii) A detailed description of the intended user, and any user training materials or programs that address appropriate reading protocols for the device, to ensure that the end user is fully aware of how to interpret and apply the device output.
(iv) A detailed description of the device inputs and outputs.
(v) A detailed description of compatible imaging hardware and imaging protocols.
(vi) Warnings, precautions, and limitations must include situations in which the device may fail or may not operate at its expected performance level (
e.g., poor image quality or for certain subpopulations), as applicable.(vii) A detailed summary of the performance testing, including test methods, dataset characteristics, results, and a summary of sub-analyses on case distributions stratified by relevant confounders, such as anatomical characteristics, patient demographics and medical history, user experience, and imaging equipment.