MammoScreen® BD is a software application intended for use with compatible full field digital mammography and digital breast tomosynthesis systems. MammoScreen BD evaluates the breast tissue composition to provide an ACR BI-RADS 5th Edition breast density category. The device is intended to be used in the population of asymptomatic women undergoing screening mammography who are at least 40 years old.

MammoScreen BD only produces adjunctive information to aid interpreting physicians in the assessment of breast tissue composition. It is not a diagnostic software.

Patient management decisions should not be made solely based on analysis by MammoScreen BD.

Device Description

MammoScreen BD is a software-only device (SaMD) using artificial intelligence to assist radiologists in the interpretation of mammograms. The MammoScreen BD software is to automatically process a mammogram to assess the density of the breasts.

For each examination, MammoScreen BD outputs the breast density in accordance with the American College of Radiology (ACR) Breast Imaging Reporting and Data System (BI-RADS) Atlas 5th Edition breast density categories "A" through "D".

MammoScreen BD takes as input a folder with images in DICOM formats and outputs a breast density assessment in a form of a JSON file. MammoScreen BD outputs can be integrated with compatible third-party software such as the MammoScreen Web-UI interface, PACS viewer (using DICOM Structured Report or DICOM Secondary Capture SOP Class UIDs), patient worklists, or within reporting software.

AI/ML Overview

Here is a detailed breakdown of the acceptance criteria and study information for MammoScreen BD, based on the provided document:

1. Table of Acceptance Criteria and Reported Device Performance

The primary acceptance criteria for the initial clearance of MammoScreen BD were related to the accuracy and agreement with ground truth established by radiologists for classifying breast density into four BI-RADS categories.

Acceptance Criteria (from PCCP section for future modifications)	Primary Objective Reported Device Performance (4-class task)	Primary Objective Reported Device Performance (Binary task)
Quadratic Kappa on GE mammograms superior to 0.85	Quadratic Cohen's Kappa: 89.03 (95% CI: 87.43 - 90.56)	Quadratic Cohen's Kappa: 84.50 (95% CI: 81.46, 87.36)
Linear Kappa, Accuracy, and Density Bins (A, B, C, D)	Accuracy: 84.68 (95% CI: 82.68, 86.67)	Accuracy: 92.29 (95% CI: 90.82, 93.77)

Note: The document explicitly states "Acceptance criteria of the updated device" under the PCCP for future modifications. While the document does not explicitly state the acceptance criteria for the initial clearance in a separate section, the reported performance metrics (Quadratic Cohen's Kappa and Accuracy for both 4-class and binary classification) are implicitly the metrics against which the device's performance was judged for its initial clearance, demonstrating its effectiveness based on comparison to the ground truth.

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

Sample Size for Test Set: 922 women/exams. (Total of 922 exams with 4 views each).
Data Provenance: Retrospectively collected from two US screening centers and one French screening center.
- 52.6% of cases (485 patients) originated from the USA.
- 47.4% of cases (437 patients) originated from France.
- The provenance did not intersect any clinical centers used for algorithm development, mitigating a center-induced bias.

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

Number of Experts: 5 breast radiologists.
Qualifications of Experts: The document specifies "5 breast radiologists" but does not provide details on their years of experience or specific board certifications.

4. Adjudication Method for the Test Set

Adjudication Method: "Consensus among the visual assessment of 5 breast radiologists." The exact method (e.g., majority vote, sequential review with tie-breaking) is not explicitly detailed beyond "consensus."

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

No, a multi-reader multi-case (MRMC) comparative effectiveness study evaluating human readers with AI assistance versus without AI assistance was not conducted or reported in this document. The study focuses on the standalone performance of the AI algorithm against expert consensus.

6. Standalone (Algorithm Only) Performance Study

Yes, a standalone performance study was conducted. The "Primary Objectives" and "Performance Data" sections directly evaluate "the accuracy and the reproducibility of MammoScreen BD algorithm in assessing the breast density category" in terms of agreement with the ground truth established by the consensus of 5 radiologists.
- For the 4-class task, the algorithm achieved a quadratic Cohen's kappa of 89.03 and an accuracy of 84.68%.
- For the binary classification task (dense vs. non-dense), the algorithm achieved a quadratic Cohen's kappa of 84.50 and an accuracy of 92.29%.

7. Type of Ground Truth Used

Type of Ground Truth: Expert Consensus. Specifically, "ground truth (GT) established by consensus among the visual assessment of 5 breast radiologists."

8. Sample Size for the Training Set

Sample Size for Training Set: 32,368 patients, comprising 108,775 studies.

9. How Ground Truth for the Training Set Was Established

The document states that the training data was derived from "De-identified screening mammograms... retrospectively collected from 32,368 patients in 2 different US sites."
It does not explicitly state how the ground truth for the training set was established. It only describes the density distribution (A: 12.79%, B: 34.58%, C: 42.94%, D: 9.38%) within the training data, implying these were pre-existing labels. It's common for such labels to be derived from radiologist reports or existing clinical records, but the specific method of ground truth establishment for the training set is not detailed.

Summary

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

October 2, 2024

Image /page/0/Picture/1 description: The image shows the logos of the Department of Health and Human Services and the Food and Drug Administration (FDA). The Department of Health and Human Services logo is on the left, and the FDA logo is on the right. The FDA logo is a blue square with the letters "FDA" in white. To the right of the square, the words "U.S. FOOD & DRUG ADMINISTRATION" are written in blue.

Therapixel Quentin De Snoeck RAQA Director/CISO 455 Promenade des Anglais Nice. 06200 FRANCE

Re: K241561

Trade/Device Name: MammoScreen BD Regulation Number: 21 CFR 892.2050 Regulation Name: Medical Image Management And Processing System Regulatory Class: Class II Product Code: QIH Dated: September 4, 2024 Received: September 4, 2024

Dear Quentin De Snoeck:

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 (the 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 available 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.

FDA's substantial equivalence determination also included the review and clearance of your Predetermined Change Control Plan (PCCP). Under section 515C(b)(1) of the Act, a new premarket notification is not required for a change to a device cleared under section 510(k) of the Act, if such change is consistent with an established PCCP granted pursuant to section 515C(b)(2) of the Act. Under 21 CFR 807.81(a)(3), a new

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

premarket notification is required if there is a major change or modification in the intended use of a device, or if there is a change or modification in a device that could significantly affect the safety or effectiveness of the device, e.g., a significant change or modification in design, material, chemical composition, energy source, or manufacturing process. Accordingly, if deviations from the established PCCP result in a major change or modification in the intended use of the device, or result in a change or modification in the device that could significantly affect the safety or effectiveness of the a new premarket notification would be required consistent with section 515C(b)(1) of the Act and 21 CFR 807.81(a)(3). Failure to submit such a premarket submission would constitute adulteration and misbranding under sections 501(f)(1)(B) and 502(o) of the Act, respectively.

Additional information about changes that may require a new premarket notification are provided in the FDA guidance documents entitled "Deciding When to Submit a 510(k) for a Change to an Existing Device" (https://www.fda.gov/media/99812/download) and "Deciding When to Submit a 510(k) for a Software Change to an Existing Device" (https://www.fda.gov/media/99785/download).

Your device is also subject to, among other requirements, the Quality System (QS) regulation (21 CFR Part 820), which includes, but is not limited to, 21 CFR 820.30. Design controls; 21 CFR 820.90. Nonconforming product; and 21 CFR 820.100, Corrective and preventive action. Please note that regardless of whether a change requires premarket review, the QS regulation requires device manufacturers to review and approve changes to device design and production (21 CFR 820.30 and 21 CFR 820.70) and document changes and approvals in the device master record (21 CFR 820.181).

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 Part 803) for devices or postmarketing safety reporting (21 CFR Part 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 Part 4, Subpart A) for combination products: and, if applicable, the electronic product radiation control provisions (Sections 531-542 of the Act); 21 CFR Parts 1000-1050.

All medical devices, including Class I and unclassified devices and combination product device constituent parts are required to be in compliance with the final Unique Device Identification System rule ("UDI Rue"). The UDI Rule requires, among other things, that a device bear a unique device identifier (UDI) on its label and package (21 CFR 801.20(a)) unless an exception or alternative applies (21 CFR 801.20(b)) and that the dates on the device label be formatted in accordance with 21 CFR 801.18. The UDI Rule (21 CFR 830.300(a) and 830.320(b)) also requires that certain information be submitted to the Global Unique Device Identification Database (GUDID) (21 CFR Part 830 Subpart E). For additional information on these requirements, please see the UDI System webpage at https://www.fda.gov/medical-device-advicecomprehensive-regulatory-assistance/unique-device-identification-system-udi-system.

Also, please note the regulation entitled, "Misbranding by reference to premarket notification" (21 CFR 807.97). For questions regarding the reporting of adverse events under the MDR regulation (21 CFR Part

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

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,

Yanna S. Kang -S

Yanna Kang, Ph.D. Assistant Director Mammography and Ultrasound Team DHT8C: Division of Radiological Imaging and Radiation Therapy Devices OHT8: Office of Radiological Health Office of Product Evaluation and Quality Center for Devices and Radiological Health

Enclosure

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

Indications for Use

Submission Number (if known)

K241561

Device Name

MammoScreen BD

Indications for Use (Describe)

MammoScreen BD only produces adjunctive information to aid interpreting physicians in the assessment of breast tissue composition. It is not a diagnostic software.

Patient management decisions should not be made solely based on analysis by MammoScreen BD.

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

escription Use (Part 21 CFR 801 Subpart D)

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."

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

Image /page/4/Picture/0 description: The image shows the logo for Therapixel. The logo consists of a blue pixelated square to the left of the word "THERAPIXEL" in a sans-serif font. Below the word "THERAPIXEL" is the tagline "TECHNOLOGY. FOR LIFE. FOR ALL." in a smaller sans-serif font.

510(k) Summary

K241561

This 510(k) summary of safety and effectiveness information is prepared in accordance with the requirements of 21 CFR § 807.92.

Applicant Information:

Therapixel 455 Promenade des Anglais, 06200 Nice France Phone: +33 9 72 55 20 39

Submission Correspondent:

Quentin de Snoeck RAQA Director/CISO Email: qdesnoeck@therapixel.com Phone: +33 9 72 55 20 39

Date Summary Prepared: October 2, 2024

Device Information:

Trade Name:	MammoScreen® BD
Common Name:	Breast Density Assessment Software
Device Classification Name:	Automated radiological image processing software
Regulation Number:	21 CFR §892.2050, Medical image management and processingsystem
Regulation Class:	Class II
Product Code:	QIH
Submission type	Traditional 510(k)

Predicate Device:

The predicate device is WRDensity by Whiterabbit.ai, cleared under K202013 (Product code QIH).

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

Image /page/5/Picture/0 description: The image is the logo for Therapixel. The logo consists of a blue pixelated square to the left of the word "THERAPIXEL" in a sans-serif font. Below the word "THERAPIXEL" is the tagline "TECHNOLOGY. FOR LIFE. FOR ALL." in a smaller sans-serif font.

Device Description

Note that MammoScreen BD outputs should be used as complementary information by radiologists while interpreting breast density. Patient management decisions should not be made solely based on analysis by MammoScreen BD, the medical professional interpreting the mammogram remains the sole decision-maker.

Indication for Use:

MammoScreen BD only produces adjunctive information to aid interpreting physicians in the assessment of breast tissue composition. It is not a diagnostic software.

Patient management decisions should not be made solely based on analysis by MammoScreen BD.

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

Image /page/6/Picture/0 description: The image shows the logo for Therapixel. The logo consists of a blue pixelated square to the left of the word "THERAPIXEL" in a sans-serif font. Below the word "THERAPIXEL" is the tagline "TECHNOLOGY. FOR LIFE. FOR ALL." in a smaller font. The logo is simple and modern, and the blue color suggests trust and reliability.

Predicate device comparison:

The predicate device is WRDensity by Whiterabbit.ai, cleared under K202013.

	Subject DeviceMammoScreen BD	Predicate DeviceWRDensity by Whiterabbit.ai (K202013)
ClassificationName	Automated Radiological Image ProcessingSoftware	Automated Radiological Image ProcessingSoftware
ProductCode	QIH	QIH
RegulationNumber	892.2050	892.2050
MedicalDevice Class	Class II	Class II
Indicationsfor Use	MammoScreen® BD is a software application intended for use with compatible full field digital mammography and digital breast tomosynthesis systems. MammoScreen BD evaluates the breast tissue composition to provide an ACR BI-RADS 5th Edition breast density category. The device is intended to be used in the population of asymptomatic women undergoing screening mammography who are at least 40 years old. MammoScreen BD only produces adjunctive information to aid interpreting physicians in the assessment of breast tissue composition. It is not a diagnostic software.Patient management decisions should not be made solely based on analysis by MammoScreen BD.	WRDensity is a software application intended for use with compatible full-field digital mammography and digital breast tomosynthesis systems. WRDensity provides an ACR BI-RADS 5th Edition breast density category to aid interpreting physicians in the assessment of breast tissue composition. WRDensity produces adjunctive information. It is not a diagnostic aid.
PatientPopulation	Asymptomatic women undergoing mammography	Symptomatic and asymptomatic women undergoing mammography
End Users	Interpreting physicians	Interpreting physicians
Image SourceModalities	FFDM and DBT (using 2D synthetic (2DSM))	FFDMHologic Selenia DimensionsHologic Lorad SeleniaSynthetic 2D (2DSM)Hologic C-View
Input: ImageData Format	DICOM Digital mammography images - For presentation: RCC, LCC, RMLO, LMLO	DICOM Digital mammography images - For presentation: RCC, LCC, RMLO, LMLO
	Subject DeviceMammoScreen BD	Predicate DeviceWRDensity by Whiterabbit.ai (K202013)
MeasurementScale	4-category breast density scale from 5th Ed.ACR BI-RADS Atlas 2013	4-category breast density scale from 5th Ed. ACRBI-RADS Atlas 2013
Output Data	BIRADS 5th Ed. For each patient:MammoScreen BD breast density assessment	BIRADS 5th Ed. For each patient: Whiterabbit.aiWRDensity Breast Density Level, and BreastDensity Level Probability
OutputDevice	Third-party Software dedicated tomammography	Mammography Workstation, PACS, RIS
OutputFormat	JSON format presented in a radiologist'scompatible system (third-party software)	DICOM Structured Report and SecondaryCapture Text labels presented in a radiologist'sPACS and RIS patient worklist
Deployment	Virtual Machine Software	Virtual Machine Software
AssessmentScope	Results per exam	Results per exam
AssessmentType	Image feature-based with deep learning	Image feature-based with deep learning
AnatomicalLocation	Breast	Breast
PCCP	Predetermined Change Control Plan (PCCP)including:	No PCCP
	• Support of GE mammograms (no re-trainingrequired)• Support of Siemens mammograms (re-training required)• Pre-training of backbone usingUnsupervised Machine Learning (asopposed to Supervised Machine Learning)

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

Image /page/7/Picture/0 description: The image shows the logo for Therapixel. The logo consists of a blue pixelated design on the left, followed by the word "THERAPIXEL" in a bold, sans-serif font. Below the company name is the tagline "TECHNOLOGY. FOR LIFE. FOR ALL." in a smaller font.

The indication for use of MammoScreen BD is similar to that of the predicate device. Both devices are intended for concurrent use by physicians interpreting breast images to help them with assessing the breast density. The devices are not intended as a replacement for the review of a physician or their clinical judgement, since they are not diagnostic software.

The assessment of the breast density is very similar to the predicate device The overall design of MammoScreen BD is the same as the design of the predicate device. Both medical devices use algorithms to provide an assessment of the breast density to the user in a similar manner and according to the ACR BIRADS 5th Edition. The technological characteristics are the same. The unsubstantial difference regarding the output format do not raise different questions about the safety and effectiveness of the device as compared to the predicate device.

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

Image /page/8/Picture/0 description: The image shows the logo for Therapixel. The logo consists of a blue pixelated design on the left, followed by the word "THERAPIXEL" in black, with a blue square above the "A". Below the word, there is a tagline that reads "TECHNOLOGY. FOR LIFE. FOR ALL." in a smaller font.

Training dataset

De-identified screening mammograms used for training were retrospectively collected from 32,368 patients in 2 different US sites.

A detailed description of the training data is available in the Table below:

Total number of studies	108,775
Density distribution	A: 12.79%
	B: 34.58%
	C: 42.94%
	D: 9.38%
	Unknown (excluded): 0.31%
Patient ages	First quartile (Q1): 47.0
	Mean: 56.0
	Third quartile (Q3): 64.0
Patient Race / Ethnicity	White: 49.13%
	Asian: 7.63%
	Black or african american: 0.43%
	native hawaiian or pacific islander: 0.09%
	Unknown: 42.72%
Manufacturer	Hologic: 61.63%
	GE: 38.37%

Performance Data

MammoScreen BD is a software-only device.

Safety and performance of MammoScreen BD have been evaluated and verified in accordance with software specifications and applicable performance standards through software verification and validation testing. Tests have been performed in compliance with the following recognized consensus standard:

IEC 62304:2006/A1:2016- Medical device software Software life-cycle processes ●

Primary Objectives

The purpose of the validation testing is to measure the accuracy and the reproducibility of MammoScreen BD algorithm in assessing the breast density category.

The primary objective was to evaluate the accuracy of MammoScreen BD in assessing the breast density value in terms of agreement between MammoScreen BD and the ground truth (GT) established by consensus among the visual assessment of 5 breast radiologists.

The standalone performance for density assignment will be measured with:

A confusion matrix between the density value assigned by MammoScreen BD and the ground truth.
Accuracy, defined as the number of correctly classified examinations divided by the total . number of examinations included in the dataset.

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

Image /page/9/Picture/0 description: The image shows the logo for Therapixel. The logo consists of a blue pixelated square design on the left, followed by the word "THERAPIXEL" in black, with a blue square above the letter "A". Below the word "THERAPIXEL" is the text "TECHNOLOGY. FOR LIFE. FOR ALL." in a smaller font size.

. Weighted (linear and quadratic) Cohen's kappa to give an overall estimation of the agreement.
Retrospective data of 922 women having undergone mammography at two US screening centers and one French screening center with either 2DSM (obtained from a DBT acquisition) or FFDM (total of 922 exams with 4 views each) were collected. 52.6% of the cases originate from the USA (47.4% from France). Among them, 196 patients had a 2DSM examination, 493 patients had an FFDM examination, and 233 patients had both. The provenance of the data does not intersect any of the clinical centers used for the development of the algorithm, mitigating the risk of a centerinduced bias.

The whole dataset was acquired with Hologic. For the purposes of this study only FFDM and 2DSM were considered. Details of the dataset are reported in the table below:

Sources	Origins	Patients	Studies	Images	Manufacturer (images)	Modalities (images)	Breast Densities (studies)	Study Dates	Patient Ages (y/o)	Patient Races
FR_1: 437(47.40%)	USA: 485(52.60%)	unknown: 478(51.84%)	unknown: 478(51.84%)	benign: 2,446(57.23%)	hologic: 4,274(100.00%)	FFDM: 3,053(71.43%)	C: 366(39.70%)	min: 2015-05-22max: 2023-02-15	min: 40q25: 52.0mean: 59.7q75: 67.0max: 90	unknown: 451(48.92%)
US_1: 249(27.01%)	FRA: 437(47.40%)	benign: 426(46.20%)	benign: 426(46.20%)	unknown: 1,774(41.51%)		2DSM: 1,221(28.57%)	B: 353(38.29%)			white: 273(29.61%)
US_2: 228(24.73%)		malignant: 14(1.52%)	malignant: 14(1.52%)	malignant: 39(0.91%)			D: 106(11.50%)			asian: 102(11.06%)
US_3: 8(0.87%)		undetermined: 4(0.43%)	undetermined: 4(0.43%)	undetermined: 15(0.35%)			A: 97(10.52%)			black or africanamerican: 88(9.54%)
										american indian oralaska native: 4(0.43%)
										native hawaiian orpacific islander: 4(0.43%)
922	922	922	922	4,274	4,274	4,274	922			922

Note that the same patients may have both FFDM and 2DSM examinations.

On the four-class task, MammoScreen BD achieved a quadratically-weighted Cohen's kappa of 89.03, 95% confidence interval [87.43 - 90.56]. A confusion matrix demonstrating the level of agreement between the breast density level and the radiologist consensus for each BI-RADS breast density category can be found in the right part of the figure below:

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

Image /page/10/Picture/0 description: The image shows the logo for Therapixel. The logo consists of a blue pixelated square to the left of the word "THERAPIXEL" in black font. Below the word "THERAPIXEL" is the phrase "TECHNOLOGY. FOR LIFE. FOR ALL." in a smaller, lighter font.

Image /page/10/Figure/1 description: The image contains two plots: a bar chart on the left and a confusion matrix on the right. The bar chart, titled "Global Analysis", displays the distribution of categories A, B, C, and D, with percentages and counts for each. The confusion matrix shows the agreement between predicted and consensus categories, with percentages and counts for each cell, along with Cohen's Kappa values (quadratic and linear) and their confidence intervals.

Accuracy = 84.68 with CI = [82.68, 86.67]

Accuracy = 92.29 with CI = [90.82, 93.77]

(Left) Distribution of breast density based on the consensus established among 5 radiologists. (Right) Confusion matrix comparing the performance of ManmoScreen-BD against the radiologist consensus assessment of breast density for the fourclass BI-RADS breast density task. The number of exams within each bin is shown in parentheses.

On the binary classification task, MammoScreen BD achieved a quadratically-weighted Cohen's kappa of 84.50, 95% confidence interval [81.46, 87.36]. The confusion matrix is presented in the figure below:

Image /page/10/Figure/4 description: The image contains two charts that show a binary analysis and a confusion matrix. The bar chart on the left shows that 48% (551) of the data is non-dense and 52% (604) is dense. The confusion matrix on the right shows the predicted TPX-BD versus the consensus, with 88% (485) of the non-dense data being correctly predicted as non-dense and 96% (581) of the dense data being correctly predicted as dense. The Cohen's Kappa (quadratic) and Cohen's Kappa (linear) are both 84.50 with CI = [81.46, 87.36].

(Left) Distribution of breast density ground truth labels grouping dense (BI-RADS C+D) and non-dense (BI-RADS A+B). (Right) Confusion matrix comparing the performance of MammoScreen BD against the radiologist consensus assessment of breast density for the binary breast density task. The number of exams within each bin is shown in parentheses.

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

Image /page/11/Picture/0 description: The image shows the logo for Therapixel. The logo consists of a blue pixelated square to the left of the word "THERAPIXEL" in black font. Below the word "THERAPIXEL" is the phrase "TECHNOLOGY. FOR LIFE. FOR ALL." in a smaller font.

Secondary Objectives

Same woman CC and MLO views ●

In addition to what was evaluated with the primary objectives, for a deeper understanding of MammoScreen BD, consistency was evaluated separately on CC and MLO views of the same patients (see the figure below). Consistency was assessed by evaluating the agreement, in terms of percentage of cases, between the density value for the mediolateral-oblique (MLO) and craniocaudal (CC) views of the same breast. We observe a good agreement between breast density assessment on CC and MLO views, with a minimum of 90% on breast density B and a maximum of 93% on breast density D. This indicates that MammoScreen BD behaves equally well independently on CC and MLO views.

Image /page/11/Figure/4 description: The image shows a confusion matrix and two bar charts that compare predictions on CC and MLO. The confusion matrix shows the percentage and number of predictions for each category (A, B, C, D) on MLO, given the actual category on CC. The bar charts show the distribution of predictions for each category on CC and MLO, with percentages and numbers displayed on top of each bar. For example, the bar chart on the left shows that 10% (86) of the predictions on CC are A, while the bar chart on the right shows that 10% (88) of the predictions on MLO are A.

Contingency matrix comparing the breast density assessment (four-class BI-RADS breast density) of ManmoScreen BD on CC and MLO views of the same patient. Individual assessments for CC (left) are given. The number of exams within each bin is shown in parentheses.

. Same woman 2DSM and FFDM

Reproducibility was further evaluated on 2DSM and FFDM mammograms of the same patients (see the figure below). 233 patients had both types of examinations and were used during this analysis. We observe a fair agreement between breast density assessment on 2DSM and FFDM. The highest breast density of type D (95%) and the smallest agreement is obtained on breast density of type A (79%).

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

Image /page/12/Picture/0 description: The image is a logo for Therapixel. The logo consists of a blue pixelated square to the left of the word "THERAPIXEL" in black font. Below the word "THERAPIXEL" is the phrase "TECHNOLOGY. FOR LIFE. FOR ALL." in a smaller font.

Image /page/12/Figure/1 description: The image contains a confusion matrix and two bar charts that compare predictions on FFDM and 2DSM. The confusion matrix shows the percentage and number of predictions for each class (A, B, C, D) on FFDM, with the rows representing the actual classes and the columns representing the predicted classes. The bar chart on the left shows the distribution of classes in 2DSM, with class B having the highest percentage (35%) and class A having the lowest (8%). The bar chart on the right shows the distribution of classes in FFDM, with class C having the highest percentage (39%) and class A having the lowest (7%).

Contingency matrix comparing the breast density assessment (four-class BI-RADS breast density) of MammoScreen BD on 2DSM and FFDM of the same patients. Individual distribution of assessments for 2DSM (right) are given. The number of exams within each bin is shown in parentheses.

Subgroup analysis ●

o Image modality

When considering FFDM mammogram only, MammoScreen BD achieved a quadraticallyweighted Cohen's kappa of 89.11, 95% CI [87.01 - 91.00] (see the figure below).

Image /page/12/Figure/6 description: The image contains two plots that show the results of an analysis on FFDM. The bar plot on the left shows the distribution of the data, with the highest percentage being 40% for category C and the lowest being 10% for category A. The confusion matrix on the right shows the accuracy and Cohen's Kappa values, with an accuracy of 84.85 and Cohen's Kappa (quadratic) of 89.11.

(Left) Density label distribution on FFDM mammograms only. (Right) Confusion matrix comparing the performance of MammoScreen BD against the radiologist consensus assessment of breast density on FFDM mammograms.

The same metric achieves a value of 88.89, 95% CI [85.88 - 91.50] when evaluated on 2DSM mammograms (see the figure below). Overall, performances on FFDM and 2DSM are comparable with those observed in the primary objectives.

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

Image /page/13/Picture/0 description: The image is a logo for Therapixel. The logo consists of a blue pixelated plus sign on the left, followed by the word "THERAPIXEL" in black, sans-serif font. Below the word "THERAPIXEL" is the tagline "TECHNOLOGY. FOR LIFE. FOR ALL." in a smaller, lighter font. The logo is clean and modern, and the blue and black color scheme is professional and trustworthy.

Image /page/13/Figure/1 description: The image contains two charts. The first chart is a bar chart titled "Analysis on 2DSM" showing the distribution of data across four categories labeled A, B, C, and D, with corresponding percentages and counts: A (9%, 38), B (39%, 168), C (38%, 161), and D (14%, 62). The second chart is a confusion matrix displaying the agreement between predicted and consensus categories, with Cohen's Kappa (linear) = 82.71 and CI = [78.30, 86.60], showing the percentage and count for each combination of predicted and consensus categories.

Accuracy = 84.38 with CI = [80.65, 87.65] Cohen's Kappa (quadratic) = 88.89 with Cl = [85.88, 91.50]

(Left) Density label distribution on 2DSM mammograms only. (Right) Confusion matrix comparing the performance of MammoScreen BD against the radiologist consensus assessment of breast density on 2DSM mammograms.

O Age

Performances of MammoScreen BD in terms of quadratically-weighted Cohen's kappa on different age groups were evaluated as follows:

Age < 55: 87.93 (95% CI: 85.19 - 90.26) ● 55 ≤ Age < 65: 91.22 (95% CI: 88.05 - 93.77) ● Age ≥ 65: 87.15 (95% CI: 82.70 - 90.96)
MammoScreen BD performances on subgroups based on age are comparable with performances observed in the primary objectives. A slightly lower agreement of MammoScreen BD compared to radiologists is observed on the oldest age group (age ≥ 65) with a quadratically-weighted Cohen's kappa of 87.15.

Please see the related figures below.

{14}------------------------------------------------

Image /page/14/Picture/0 description: The image is a logo for Therapixel. The logo consists of a blue pixelated square to the left of the word "THERAPIXEL" in a sans-serif font. Below the word "THERAPIXEL" is the tagline "TECHNOLOGY. FOR LIFE. FOR ALL." in a smaller font.

Accuracy = 82.63 with CI = [79.47, 85.61] Cohen's Kappa (quadratic) = 87.93 with Cl = [85.19, 90.26] Cohen's Kappa (linear) = 80.93 with Cl = [77.14, 84.48]

Accuracy = 87.99 with CI = [84.09, 91.56] Cohen's Kappa (quadratic) = 91.22 with Cl = [88.05, 93.77] Cohen's Kappa (linear) = 86.46 with Cl = [81.93, 90.27]

Image /page/14/Figure/2 description: The image contains two plots. The plot on the left is a bar chart titled "Global Analysis - age < 55", showing the distribution of data across four categories (A, B, C, and D). The percentages and counts for each category are as follows: A (9%, 53), B (32%, 182), C (42%, 239), and D (17%, 96). The plot on the right is a confusion matrix comparing "Consensus" and "Predicted TPX-BD" across the same four categories, with cells showing percentages and counts of agreement and disagreement between the two classifications.

(Left) Density label distribution on patients younger than 55 years old. (Right) Confusion matrix comparing the performance of MammoScreen BD against the radiologist consensus assessment of breast density on such patients.

Image /page/14/Figure/4 description: The image contains two plots. The plot on the left is a bar chart titled "Global Analysis - 55 <= age < 65", showing the distribution of categories A, B, C, and D, with percentages and counts of 8% (26), 37% (115), 42% (129), and 12% (38), respectively. The plot on the right is a confusion matrix labeled "Predicted TPX-BD" versus "Consensus", displaying the classification accuracy for each category. The matrix shows high accuracy along the diagonal, with values of 92% for A, 86% for B, 88% for C, and 89% for D.

B C D (Left) Density label distribution on patients between 55 and 65 years old. (Right) Confusion matrix comparing the performance

of MammoScreen BD against the radiologist consensus assessment of breast density on such patients.

{15}------------------------------------------------

Image /page/15/Picture/0 description: The image shows the logo for Therapixel. The logo consists of a blue pixelated plus sign on the left, followed by the word "THERAPIXEL" in black, with a blue pixel above the "A". Below the word "THERAPIXEL" is the tagline "TECHNOLOGY. FOR LIFE. FOR ALL." in a smaller, sans-serif font.

Image /page/15/Figure/1 description: The image shows the accuracy and Cohen's Kappa values with confidence intervals. The accuracy is 85.50 with a confidence interval of [81.04, 89.59]. Cohen's Kappa (quadratic) is 87.15 with a confidence interval of [82.70, 90.96]. Cohen's Kappa (linear) is 81.60 with a confidence interval of [75.94, 86.83].

Image /page/15/Figure/2 description: The image contains two plots. The plot on the left is a bar chart titled "Global Analysis - age >= 65", showing the distribution of categories A, B, C, and D, with percentages 13%, 51%, 31%, and 5% respectively. The plot on the right is a confusion matrix for predicted TPX-BD, showing the agreement between consensus categories A, B, C, and D and predicted categories A, B, C, and D. For example, 85% of category A are predicted as A, and 83% of category B are predicted as B.

(Left) Density label distribution on patients old. (Right) Confision matrix comparing the performance of MammoScreen BD against the radiologist consensus assessment of breast density on such patients.

Breast thickness O

Performances of MammoScreen BD in terms of quadratically-weighted Cohen's kappa on different breast thickness groups were evaluated as follows:

Thickness < 50 mm: 85.81 (95% CI: 81.99 - 89.30) 50 mm ≤ Thickness < 70 mm: 86.90 (95% CI: 84.25 - 89.34) ● . Thickness ≥ 70 mm: 89.29 (95% CI: 85.19 - 92.89)
A lower agreement on low-dense breasts (density A) on thickness less than 50 mm is observed. but few data were available (7 cases). Similarly, a lower agreement on high-dense breasts (density D) on thickness greater than 70 mm is observed. Likewise, few samples were available (4 cases). Overall, MammoScreen BD performances on subgroups based on breast thickness are comparable with performances observed in the primary objectives.

Please see the related figures below.

{16}------------------------------------------------

Image /page/16/Picture/0 description: The image is a logo for Therapixel. The logo consists of a blue pixelated cross to the left of the word "THERAPIXEL" in a sans-serif font. Below the word "THERAPIXEL" is the tagline "TECHNOLOGY. FOR LIFE. FOR ALL." in a smaller sans-serif font.

Accuracy = 83.39 with CI = [79.62, 87.15] Cohen's Kappa (quadratic) = 85.81 with Cl = [81.99, 89.30] Cohen's Kappa (linear) = 79.61 with Cl = [74.67, 84.49]

Image /page/16/Figure/2 description: The image contains two plots. The first plot is a bar chart titled "Global Analysis - thickness < 50". The x-axis represents categories A, B, C, and D, while the y-axis represents the percentage and count for each category. The percentages are 2%, 25%, 45%, and 27% for categories A, B, C, and D, respectively. The second plot is a confusion matrix showing the predicted TPX-BD versus consensus, with percentages and counts for each cell in the matrix.

(Left) Density label distribution on screen with breast thickness lower than 50mm. (Right) Confusion matrix comparing the performance of MammoScreen BD against the radiologist consensus of breast density on such screens.

Image /page/16/Figure/4 description: The image contains two plots. The plot on the left is a bar chart titled "Global Analysis - 50 <= thickness < 70". The x-axis represents categories A, B, C, and D, while the y-axis represents the percentage and count for each category. The plot on the right is a confusion matrix showing the predicted TPX-BD against the consensus, with categories A, B, C, and D on both axes, and the cells showing percentages and counts.

Accuracy = 84.80 with CI = [82.08, 87.52] Cohen's Kappa (quadratic) = 86.90 with CI = [84.25, 89.34] Cohen's Kappa (linear) = 81.09 with Cl = [77.55, 84.41]

(Left) Density label distribution on screen with breast thickness between 50mm (Right) Confusion matrix comparing the performance of MammoScreen BD against the radiologist consensus assessment of breast density on such screens.

{17}------------------------------------------------

Image /page/17/Picture/0 description: The image shows the logo for Therapixel. The logo consists of a blue pixelated square to the left of the word "THERAPIXEL" in a sans-serif font. Below the word "THERAPIXEL" is the tagline "TECHNOLOGY. FOR LIFE. FOR ALL." in a smaller font.

Accuracy = 86.26 with CI = [81.52, 91.00] Cohen's Kappa (quadratic) = 89.29 with Cl = [85.19, 92.89] Cohen's Kappa (linear) = 84.12 with Cl = [78.53, 89.55]

Image /page/17/Figure/2 description: The image contains two plots. The first plot is a bar chart titled "Global Analysis - thickness >= 70". The bar chart shows the distribution of four categories, A, B, C, and D, with percentages of 30%, 41%, 27%, and 2% respectively. The second plot is a confusion matrix showing the predicted TPX-BD values, with categories A, B, C, and D on both axes, and the matrix shows the percentages and counts of each category.

(Left) Density label distribution on screen with breast than 70mm. (Right) Confusion matrix comparing the performance of MammoScreen BD against the radiologist consensus of breast density on such screens.

Race O

Performances of MammoScreen BD in terms of quadratically-weighted Cohen's kappa on different race groups were evaluated as follows:

White: 87.88 (95% CI: 81.82 - 93.18)
Asian: 87.11 (95% CI: 80.22 - 93.04)
Black or African American: 90.82 (95% CI: 85.58 94.97) ●
. American Indian or Alaska native or native Hawaiian or Pacific islander: 53.85 (95% CI: 0.00-83.33)

A lower agreement of MammoScreen BD compared to radiologists is observed on the last group including American Indians, Alaska natives, native Hawaiians and Pacific islanders with a quadratically-weighted Cohen's kappa of 53.85. However, this result is mainly due to the limited size of this group (8 patients in total). Overall, MammoScreen BD performances on subgroups based on race are comparable with performances observed in the primary objectives and does not appear to be biased with respect to race.

Please see the related figures below.

{18}------------------------------------------------

Image /page/18/Picture/0 description: The image is a logo for Therapixel. The logo consists of a blue pixelated square design on the left, followed by the word "THERAPIXEL" in a bold, sans-serif font. Below the word "THERAPIXEL" is the tagline "TECHNOLOGY. FOR LIFE. FOR ALL." in a smaller, lighter font.

Accuracy = 83.50 with CI = [79.80, 86.70] Cohen's Kappa (quadratic) = 87.72 with Cl = [84.44, 90.37] Cohen's Kappa (linear) = 81.24 with Cl = [76.68, 85.03]

Image /page/18/Figure/2 description: The image contains two plots. The plot on the left is a bar chart titled "Global Analysis - ['white']" showing the distribution of four categories labeled A, B, C, and D. The percentages and counts for each category are: A (15%, 59), B (43%, 176), C (34%, 137), and D (8%, 34). The plot on the right is a confusion matrix labeled "Predicted TPX-BD" showing the relationship between predicted and consensus categories A, B, C, and D, with percentages and counts for each cell, along with a color bar indicating the percentage scale.

(Left) Density label distribution on White patients. (Right) Confusion matrix comparing the performance of MammoScreen BD against the radiologist consensus assessment of breast density on such screens.

Image /page/18/Figure/4 description: The image contains two charts related to a global analysis of 'asian'. The first chart is a bar graph showing the distribution of categories A, B, C, and D, with corresponding percentages and counts: A (2%, 2), B (39%, 52), C (46%, 61), and D (13%, 17). The second chart is a confusion matrix displaying the predicted TPX-BD against the consensus, with values indicating the percentage and count for each category combination, such as A-A (100%, 2), B-B (87%, 45), C-C (93%, 57), and D-D (71%, 12).

Accuracy = 87.88 with CI = [81.82, 93.18] Cohen's Kappa (quadratic) = 87.11 with CI = [80.22, 93.04] Cohen's Kappa (linear) = 83.02 with Cl = [74.75, 90.64]

(Left) Density label distribution on Asian patients. (Right) Confusion matrix comparing the performance of MammoScreen BD against the radiologist consensus assessment of breast density on such screens.

{19}------------------------------------------------

Image /page/19/Picture/0 description: The image is a logo for Therapixel. The logo features a series of blue squares on the left side, arranged in a pattern resembling a plus sign. To the right of the squares is the word "THERAPIXEL" in a bold, sans-serif font. Below the word "THERAPIXEL" is the tagline "TECHNOLOGY. FOR LIFE. FOR ALL." in a smaller, lighter font.

Accuracy = 84.21 with CI = [76.84, 91.58] Cohen's Kappa (quadratic) = 90.82 with Cl = [85.58, 94.97] Cohen's Kappa (linear) = 84.47 with Cl = [76.45, 91.51]

Image /page/19/Figure/2 description: The image contains two plots. The first plot is a bar chart titled "Global Analysis - ['black or african american']". The x-axis represents categories A, B, C, and D, with corresponding percentages and counts: A (21%, 20), B (34%, 32), C (36%, 34), and D (9%, 9). The second plot is a confusion matrix showing the relationship between "Consensus" and "Predicted TPX-BD" across categories A, B, C, and D. The matrix displays percentages and counts, with high accuracy along the diagonal, indicating good agreement between consensus and predicted values for categories A, B, C, and D.

(Left) Density label distribution on Black patients. (Right) Confusion matrix comparing the performance of MammoScreen BD against the radiologist consensus assessment of breast density on such screens.

Accuracy = 62.50 with CI = [25.00. 87.50] Cohen's Kappa (quadratic) = 53.85 with CI = [0.00, 83.33] Cohen's Kappa (linear) = 40.00 with Cl = [0.00, 78.95]

Image /page/19/Figure/5 description: The image contains two charts related to a global analysis of 'American Indian or Alaska Native' and 'Native Hawaiian or Pacific Islander' populations. The first chart is a bar graph showing the distribution of data across categories A, B, C, and D, with corresponding percentages and counts: A at 12% (1), B at 38% (3), C at 50% (4), and D at 0% (0). The second chart is a confusion matrix labeled 'Predicted TPX-BD' against 'Consensus', displaying the agreement between predicted and actual categories, with values indicating the percentage and count of each combination.

(Left) Density label distribution on American Indians, Alaska native, native Hawaiian or pacific islander patients. (Right) Confusion matrix comparing the performance of MammoScreen BD against the radiologist consensus assessment of breast density on such screens.

The performance testing results indicate that MammoScreen BD does not pose any new concerns regarding safety or effectiveness. Additionally, it was established that MammoScreen BD is noninferior to the targeted performance.

{20}------------------------------------------------

Image /page/20/Picture/0 description: The image shows the logo for Therapixel. The logo consists of a blue pixelated square design on the left, followed by the word "THERAPIXEL" in a sans-serif font. Below the word is the tagline "TECHNOLOGY. FOR LIFE. FOR ALL."

Predetermined Change Control Plan (PCCP)

MammoScreen BD is powered by machine-learning neural architectures. Therapixel will make future algorithm improvements under a PCCP. The plan describes the future modifications, assesses their impact, and a modification protocol details how data management, re-training, performance evaluation and update procedures will be handled.

below explains, for each Predetermined change, the data used The table for development/modification and testing, and the monitoring methodology.

	Modification 1:Support of GEmammograms (nore-trainingrequired)	Modification 2:Support of SiemensMammograms (re-training required)	Modification 3:Pre-training ofbackbone usingUnsupervisedMachine Learning(as opposed toSupervisedMachine Learning)
Data used fordevelopment andmodifications,representing thetarget population	No new datacollection is foreseenfor this change.	Siemens, FFDMSiemens, 2DSM	No new datacollection is foreseenfor this change.
Comprehensive testingfor planned changes	Agreement testingagainst the visualassessment of 5radiologists on GEmammograms.	Agreement testingagainst the visualassessment of 5radiologists onHologicmammograms.Agreement testingagainst the visualassessment of 5radiologists onSiemensmammograms.	Agreement testingagainst the visualassessment of 5radiologists onHologic and newmanufacturer(s)mammograms.
Acceptance criteriaof the updated device	Quadratic kappa onGE mammogramssuperior to 0.85.Linear kappa,accuracy, anddensity bins (A, B, Cand D)	Quadratic kappa onSiemens andHologicmammogramssuperior to 0.85Linear kappa,accuracy, and densitybins (A, B, C and D)	Quadratic kappa onHologicmammogramssuperior to 0.85.Linear kappa,accuracy, and densitybins (A, B, C and D)
Characterization ofthe device before andafter implementationof changes	mammograms non-inferior to thecomparator device(K241561) onHologicmammograms.	bins (A, B, C and D) onSiemens andHologicmammograms non-inferior to thecomparator device(K241561) onHologicmammograms.	on Hologicmammograms non-inferior to thecomparator device(K241561) onHologicmammograms.
	The device will be accessible to more centers, and thus to morewoman. Prevents obsolescence of MammoScreen BD.Better representation of breast tissue diversity leading to higheroverall performances and a better generalization on unseen data.
Monitoring, detection,and response todeviations in deviceperformance.	Therapixel monitors customer sites. The distribution of breast densityassessment obtained is determined on a representative screeningdistribution, which serves as a Reference Distribution. Devicemonitoring compares breast density assessment in real conditions tothe reference distribution and alerts of any deviations. Theinvestigation can result in a field-safety notice, a Medical DeviceReport.

{21}------------------------------------------------

Image /page/21/Picture/0 description: The image is a logo for Therapixel. The logo features the word "THERAPIXEL" in a sans-serif font, with a small blue square above the "A". To the left of the word is a blue pixelated square. Below the word "THERAPIXEL" is the text "TECHNOLOGY. FOR LIFE. FOR ALL."

Conclusions

Performance testing results demonstrated that the device is safe and effective.

Therapixel has applied a risk management process following FDA-recognized standards to identify, evaluate, and mitigate all known hazards related to MammoScreen BD. All identified risks are effectively mitigated, and it can be concluded that the residual risk is outweighed by the benefits. Considering all data in this submission, the data provided in this 510(k) supports the safe and effective use of MammoScreen BD for its indications for use and substantial equivalence to the 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).