(258 days)
Yes
The document explicitly states that the device uses an "Artificial intelligence algorithm trained on a database..."
No.
The device is a computer-assisted detection and diagnosis software that aids radiologists in interpreting medical images, not directly treating a medical condition.
Yes
The "Intended Use / Indications for Use" section states that the device is "intended to be used... to assist radiologists in the interpretation of a prostate MRI examination" and provides "automatic detection and segmentation of regions suspicious for cancer," and a "PI-RADS interpretation suggestion." These functions are directly related to diagnosing disease.
Yes
The device is described as a "plug-in Radiological Computer Assisted Detection and Diagnosis Software device" and a "Computer Assisted Detection and Diagnosis algorithm designed to plug into a hosting workflow". It processes existing image data and provides outputs for interpretation by a radiologist within a separate hosting application. There is no mention of accompanying hardware or hardware components included with the device itself.
Based on the provided information, this device is not an IVD (In Vitro Diagnostic).
Here's why:
- IVD Definition: In Vitro Diagnostics are devices intended for use in the collection, preparation, and examination of specimens taken from the human body (such as blood, urine, or tissue) to provide information for the diagnosis, treatment, or prevention of disease.
- Device Function: The Prostate MR AI device analyzes medical images (MRI scans), not biological specimens. It processes existing image data to assist radiologists in interpreting those images.
- Intended Use: The intended use is to be a "concurrent reading aid to assist radiologists in the interpretation of a prostate MRI examination." This is a function related to image analysis and interpretation, not the analysis of biological samples.
Therefore, while it is a medical device used for diagnostic purposes, its method of operation and the type of input it uses (medical images) classify it as a Radiological Computer Assisted Detection and Diagnosis Software rather than an In Vitro Diagnostic.
No
The clearance letter does not explicitly state that the FDA has reviewed and approved or cleared a Predetermined Change Control Plan (PCCP) for this specific device. The section "Control Plan Authorized (PCCP) and relevant text" is listed as "Not Found."
Intended Use / Indications for Use
Prostate MR AI is a plug-in Radiological Computer Assisted Detection and Diagnosis Software device intended to be used
• with a separate hosting application
• as a concurrent reading aid to assist radiologists in the interpretation of a prostate MRI examination acquired according to the PI-RADS standard
• in adult men (40 years and older) with suspected cancer in treatment naïve prostate glands
The plug-in software analyzes non-contrast T2 weighted (T2W) and diffusion weighted image (DWI) series to segment the prostate gland and to provide an automatic detection and segmentation of regions suspicious for cancer. For each suspicious region detected, the algorithm moreover provides a lesion Score, by way of PI-RADS interpretation suggestion.
Outputs of the device should be interpreted consistently with ACR recommendations using all available MR data (e.g., dynamic contrast enhanced images [if available]).
Patient management decisions should not be made solely based on analysis by the Prostate MR AI algorithm.
Product codes (comma separated list FDA assigned to the subject device)
ODO
Device Description
This premarket notification addresses the Siemens Healthineers Prostate MR AI (VA10A) Radiological Computer Assisted Detection and Diagnosis Software (CADe/CADx).
Prostate MR AI is a Computer Assisted Detection and Diagnosis algorithm designed to plug into a hosting workflow that assists radiologists in the detection of suspicious lesions and their classification. It is used as a concurrent reading aid to assist radiologists in the interpretation of a prostate MRI examination acquired according to the PI-RADS standard.
The automatic lesion detection requires transversal T2W and DWI series as inputs. The device automatically exports a list of detected prostate regions that are suspicious for cancer (each list entry consists of contours and a classification by Score and Level of Suspicion (LoS)), a computed suspicion map, and a per-case LoS. The results of the Prostate MR AI plug-in (with the case-level LoS, lesion center points, lesion diameters, lesion ADC median, lesion 10th percentile, suspicion map, and non-PZ segmentation considered optional) are to be shown in a hosting application that allows the radiologist to view the original case, as well as confirm, reject, or edit lesion candidates with their contours and Scores as generated by the Prostate MR AI plug-in. Moreover, the radiologist can add lesions with contours and PI-RADS scores and finalize the case. In addition, the outputs include an automatically computed prostate segmentation, as well as sub-segmentations of the peripheral zone and the rest of the prostate (non-PZ).
The algorithm will augment the prostate workflow of currently cleared syngo.MR General Engine if activated via a separate license on the General Engine.
Mentions image processing
Yes
Mentions AI, DNN, or ML
Yes
Input Imaging Modality
non-contrast T2 weighted (T2W) and diffusion weighted image (DWI) series
Anatomical Site
Prostate gland
Indicated Patient Age Range
Adult men (40 years and older)
Intended User / Care Setting
Radiologists qualified to read prostate MRI
Description of the training set, sample size, data source, and annotation protocol
The cases for the reader study were kept completely separate from those used for the training of the Prostate MR AI algorithm.
Description of the test set, sample size, data source, and annotation protocol
Automatic prostate segmentation
222 transversal T2 series from more than 10 clinical sites.
The reference standard was established through pixel-wise consensus, which was built based on annotation results of three expert radiologists.
Prostate lesion detection and classification
- 105 cases from 6 sites against a ground truth generated by radiologists.
- 115 cases from 6 sites against a ground truth generated by prostate biopsy.
- 340 cases from the multi-reader multi-case study.
The reference standard for the radiology ground truth was established through consensus reading of three expert radiologists in prostate MRI reading. The reference standard for the pathology ground truth was established through biopsy results for the same patient.
Clinical Test (Reader Study)
340 cases selected retrospectively from two US sites. Cases were consecutive and not enriched for positive cases. Additional consecutive patient cases specifically from men of African descent were included to ensure that at least 13% of the study cases are of Black or African American ethnicity.
The primary endpoint reference standard: Biopsy results (with the criterion Gleason Grade Group GGG greater or equal to 1, i.e., any cancer on biopsy), and in case they were not available, PSA density and follow-up.
The secondary endpoint reference standard: consensus lesions with a consensus PI-RADS of at least 3 from majority voting among 3 experienced radiologists acting as Truthers.
Summary of Performance Studies (study type, sample size, AUC, MRMC, standalone performance, key results)
Automatic prostate segmentation
Nonclinical Test
Sample Size: 222 transversal T2 series
Key results:
- The median of the Dice score between the AI algorithm results and the corresponding ground truth masks exceeds the threshold of 0.9.
- The median of the normalized volume difference between the algorithm results and the corresponding ground truth masks is within a ±5% range.
- The AI algorithm results as compared to any individual reader are statistically non-inferior based on variabilities that existed among the individual readers within the 5% margin of error and 5% significance level.
Prostate lesion detection and classification
Nonclinical Test
Sample Size: 105 cases for radiology ground truth, 115 cases for pathology ground truth, 340 for MRMC study.
Key results:
- The case level sensitivity of the lesion detection is equal or greater than 0.80 for both radiology and pathology ground truth.
- The false positive rate per case of the lesion detection is smaller than one false positive per case for radiology ground truth.
- The accuracy of the PI-RADS classification of radiology ground truth lesions detected by the algorithm is equal or greater than 0.8.
- The non-inferior performance of the subject device in GE vs Siemens and African American vs non-African American cases, and in cases with peripheral zone vs non- peripheral lesions.
Clinical Study
Reader study with 4080 reads
Study Type: Multi-Reader Multi-Case (MRMC) study in a paired split-plot design, combining two fully-crossed MRMC (sub-)studies. 12 Radiologists.
Sample Size: 340 cases
AUC:
Fully inclusive analysis (primary endpoint):
- radiologists' AUC (unaided) = 0.676
- radiologists' AUC (aided) = 0.701
- mean difference: +0.025, 95% C.I. [0.001, 0.049]
Maximally restrictive analysis (primary endpoint): - radiologists' AUC (unaided) = 0.658
- radiologists' AUC (aided) = 0.695
- mean difference: +0.037, 95% C.I. [0.011, 0.063]
AUwAFROC (secondary endpoint):
Fully inclusive analysis: - improved in aided reading by 0.0350 (95% C.I.:[0.0020, 0.0681], P=0.037)
Maximally restrictive analysis: - improved in aided vs. unaided reading by 0.302 (95% C.I.: [0.0080,0.0520], P=0.008)
Standalone Performance (AI standalone performance curve shown in ROC plots with 95% confidence interval envelope).
Key results: - In either analysis (fully inclusive or maximally restrictive), the improvement in AUROC for the primary endpoint was statistically significant and the primary endpoint thus met.
- In either analysis (fully inclusive or maximally restrictive), the improvement in AUwAFROC for the secondary endpoint was statistically significant and the secondary endpoint thus met.
- Supplemental analysis showed improvement in Fleiss' Kappa for interreader agreement in per-case PI-RADS scores for aided vs unaided reads (0.087, 95% C.I. [0.051, 0.125], P
§ 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.
0
Image /page/0/Picture/0 description: The image shows the logo of the U.S. Food and Drug Administration (FDA). On the left is the Department of Health and Human Services logo. To the right of that is the FDA logo, which is a blue square with the letters "FDA" in white. To the right of the blue square is the text "U.S. FOOD & DRUG ADMINISTRATION" in blue.
March 5, 2025
Siemens Healthcare GmbH Abhineet Johri Regulatory Affairs Manager Henkestr. 127 Erlangen, 91052 Germany
Re: K241770
Trade/Device Name: Prostate MR AI (VA10A) Regulation Number: 21 CFR 892.2090 Regulation Name: Radiological Computer Assisted Detection And Diagnosis Software Regulatory Class: Class II Product Code: ODO Dated: February 6, 2025 Received: February 6, 2025
Dear Abhineet Johri:
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.
1
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 (QS) 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 Rule"). 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 803), please go to https://www.fda.gov/medical-device-safety/medical-device-reportingmdr-how-report-medical-device-problems.
2
For comprehensive regulatory information about mediation-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,
D.R.K.
Daniel M. Krainak, Ph.D Assistant Director Magnetic Resonance and Nuclear Medicine 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
510(k) Number (if known) K241770
Device Name Prostate MR AI (VA10A)
Indications for Use (Describe)
Prostate MR AI is a plug-in Radiological Computer Assisted Detection and Diagnosis Software device intended to be used · with a separate hosting application
· as a concurrent reading aid to assist radiologists in the interpretation of a prostate MRI examination acquired according to the PI-RADS standard
· in adult men (40 years and older) with suspected cancer in treatment naïve prostate glands
The plug-in software analyzes non-contrast T2 weighted (T2W) and diffusion weighted image (DWI) series to segment the prostate gland and to provide an automatic detection and segmentation of regions suspicious for cancer. For each suspicious region detected, the algorithm moreover provides a lesion Score, by way of PI-RADS interpretation suggestion.
Outputs of the device should be interpreted consistently with ACR recommendations using all available MR data (e.g., dynamic contrast enhanced images [if available]).
Patient management decisions should not be made solely based on analysis by the Prostate MR AI algorithm.
| 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."
4
Image /page/4/Picture/1 description: The image contains the logo for Siemens Healthineers. The word "SIEMENS" is written in teal, and the word "Healthineers" is written in orange below it. To the right of the words is a graphic of orange dots arranged in a circular pattern.
510(k) Summary K241770 Prostate MR AI (VA10A)
In accordance with 21 CFR §807.92, the following summary of safety and effectiveness is provided.
SUBMITTER I.
21CFR § 807.92(a)(1)
21CFR § 807.92(a)(3)
Siemens Healthcare GmbH Henkestr. 127 91052 Erlangen Germany
Contact: Mr. Abhineet Johri Phone: +1 (484) 680-8723 Email: abhineet.johri@siemens-healthineers.com
Date Prepared: May 17, 2024
DEVICE II.
21CFR § 807.92(a)(2) Device Trade Name Prostate MR AI (VA10A) Classification Name Radiological Computer Assisted Detection/Diagnosis Software For Lesions Suspicious For Cancer Device Classification Panel Radiology 892.2090 Regulation Number Product Code QDQ
III. LEGALLY MARKETED PREDICATE DEVICES
| Predicate Device | |
|---|---|
| Device Trade Name | Transpara™ |
| 510(k) Number | K181704 |
| Regulation Number | 892.2090 |
| Product Code | QDQ |
| This predicate has not been subject to a design-related recall |
Reference Device
ProstatID™M Device Trade Name
5
Image /page/5/Picture/0 description: The image shows the Siemens Healthineers logo. The word "SIEMENS" is written in teal, and the word "Healthineers" is written in orange below it. To the right of the word "Healthineers" is a cluster of orange dots.
| 510(k) Number | K212783 |
|---|---|
| Regulation Number | 892.2090 |
| Product Code | QDQ |
| MINIMAL A SECTION A CONSULTION CONSULTION CONSULTION CONSULTION CONSULTION CONSULTION CONSULTION CONSULTION CONSULTION CONSULTION CONSULTION CONSULTION CONSULTION CONSULTION |
This predicate has not been subject to a design-related recall.
IV. DEVICE DESCRIPTION SUMMARY
21CFR § 807.92(a)(4)
This premarket notification addresses the Siemens Healthineers Prostate MR AI (VA10A) Radiological Computer Assisted Detection and Diagnosis Software (CADe/CADx).
Prostate MR AI is a Computer Assisted Detection and Diagnosis algorithm designed to plug into a hosting workflow that assists radiologists in the detection of suspicious lesions and their classification. It is used as a concurrent reading aid to assist radiologists in the interpretation of a prostate MRI examination acquired according to the PI-RADS standard.
The automatic lesion detection requires transversal T2W and DWI series as inputs. The device automatically exports a list of detected prostate regions that are suspicious for cancer (each list entry consists of contours and a classification by Score and Level of Suspicion (LoS)), a computed suspicion map, and a per-case LoS. The results of the Prostate MR AI plug-in (with the case-level LoS, lesion center points, lesion diameters, lesion ADC median, lesion 10th percentile, suspicion map, and non-PZ segmentation considered optional) are to be shown in a hosting application that allows the radiologist to view the original case, as well as confirm, reject, or edit lesion candidates with their contours and Scores as generated by the Prostate MR AI plug-in. Moreover, the radiologist can add lesions with contours and PI-RADS scores and finalize the case. In addition, the outputs include an automatically computed prostate segmentation, as well as sub-segmentations of the peripheral zone and the rest of the prostate (non-PZ).
The algorithm will augment the prostate workflow of currently cleared syngo.MR General Engine if activated via a separate license on the General Engine.
INTENDED USE/INDICATIONS FOR USE V.
21CFR § 807.92(a)(5)
| Predicate Device
Transpara™
K181704 | Reference Device
ProstatID™
K212783 | Subject Device
Prostate MR AI (VA10A) |
|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| The ScreenPoint Transpara™
system is intended for use as a
concurrent reading aid for
physicians interpreting screening
mammograms, to identify
regions suspicious for breast
cancer and assess their
likelihood of malignancy.
Output of the device includes
marks placed on suspicious soft
tissue lesions and suspicious
calcifications; region-based
scores, displayed upon the
physician's query, indicating the | ProstatID™ is a radiological
computer assisted detection
(CADe) and diagnostic (CADx)
software device for use in a
healthcare facility or hospital to
assist trained radiologists in the
detection, assessment and
characterization of prostate
abnormalities, including cancer
lesions using MR image data | Prostate MR AI is a plug-in
Radiological Computer Assisted
Detection and Diagnosis
Software device intended to be
used
• with a separate hosting
application
• as a concurrent reading aid
to assist radiologists in the
interpretation of a prostate
MRI examination acquired |
| likelihood that cancer is present | with the following indications
for use. | according to the PI-RADS
standard |
| in specific regions; and an | ProstatID analyzes T2W, DWI
and ADC MRI data. ProstatID
does not include DCE images in
its analysis. | in adult men (40 years and
older) with suspected cancer
in treatment naïve prostate
glands |
| overall score indicating the | ProstatID software is intended
for use as a concurrent reading
aid for physicians interpreting
prostate MRI exams of patients
presented for high-risk screening
or diagnostic imaging, from
compatible MRI systems, to
identify regions suspicious for
prostate cancer and assess their
likelihood of malignancy. | The plug-in software analyzes
non-contrast T2 weighted (T2W)
and diffusion weighted image
(DWI) series to segment the
prostate gland and to provide an
automatic detection and
segmentation of regions
suspicious for cancer. For each
suspicious region detected, the
algorithm moreover provides a
lesion Score, by way of PI-
RADS interpretation suggestion. |
| likelihood that cancer is present
on the mammogram. Patient
management decisions should
not be made solely on the basis
of analysis by Transpara™™. | Outputs of the device include the
volume of the prostate and
locations, as well as the extent of
suspect lesions, with index
scores indicating the likelihood
that cancer is present, as well as
an exam score by way of PI-
RADS interpretation suggestion.
"Extent of suspect lesions" refers
to both the assessment of the
boundary of a particular
abnormality, as well as
identification of multiple
abnormalities. In cases where
multiple abnormalities are
present, ProstatID can be used to
assess each abnormality
independently. | Outputs of the device should be
interpreted consistently with
ACR recommendations using all
available MR data (e.g.,
dynamic contrast-enhanced
images [if available]). |
| | Outputs of this device should be
interpreted with all available MR
data consistent with ACR
clinical recommendations (e.g.,
dynamic contract enhanced
images if available) in context of
PI-RADs v2, and in conjunction
with bi-parametric MRI acquired
with either surface or endorectal
MRI accessory coils from
compatible MRI systems. | Patient management decisions
should not be made solely based
on analysis by the Prostate MR
AI algorithm. |
Indications for Use Comparison
6
Image /page/6/Picture/0 description: The image contains the logo for Siemens Healthineers. The word "SIEMENS" is written in teal, and the word "Healthineers" is written in orange below it. To the right of the words is a graphic of orange dots.
7
| intended as a replacement for |
|---|
| interpreting prostate |
| abnormalities using MR image |
| data consistent with clinical |
| recommendations (including |
| DCE); nor should patient |
| management decisions be made |
| solely on the basis of ProstatID. |
The indication for Use of Prostate MR AI is similar to that of the predicate device. Both devices are designed for use by medical professionals who analyze radiological images, assisting them in pinpointing and characterizing abnormalities. The devices are both intended to be used concurrently with image interpretation but are not meant to replace a clinician's evaluation or clinical judgement. Thus, the subject and predicate devices are both intended to perform the same type of function and serve the same fundamental role in medical practice.
There are distinctions in the disease-specific abnormalities these devices can identify, the types of medical images they can process, and the specific patient populations they are intended for. The core functionalities of lesion identification and interpretation for medical images remain consistent across the differences. The new concerns regarding the safety and effectiveness of the device raised by these distinctions are assessed and resolved in the device designs to ensure the substantial equivalence.
Indications for Use/Intended Use Comparison Summary and Conclusion
The Indications for Use were assessed in accordance with the following FDA Guidance Documents:
- The 510(k) Program: Evaluating Substantial Equivalence in Premarket Notifications [510(k)] .
The results of this evaluation determined that the Indications for Use for the subject device and the predicate device are fundamentally equivalent, and only include differences in modality type, body region, and vendors. As such, Siemens Healthineers is of the opinion that the Intended Use and Indications for Use are similar to the predicate device.
THE PREDICATE DEVICES
21CFR § 807.92(a)(6)
| Attribute | Predicate Device
Transpara™
K181704 | Reference Device
ProstatID™
K212783 | Subject Device
Prostate MR AI (VA10A) | Equivalency
Analysis |
|-------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| General Information | | | | |
| Regulation
number | § 892.2090 Radiological
Computer Assisted Detection
and Diagnosis Software | § 892.2090 Radiological
Computer Assisted Detection and
Diagnosis Software | § 892.2090 Radiological
Computer Assisted Detection
and Diagnosis Software | Identical |
| Classification | Class II | Class II | Class II | Identical |
| Product Code | QDQ | QDQ | QDQ | Identical |
| Clinical Characteristics | | | | |
| Attribute | Predicate Device
Transpara™
K181704 | Reference Device
ProstatIDTM
K212783 | Subject Device
Prostate MR AI (VA10A) | Equivalency
Analysis |
| Intended Use
(short) | A concurrent reading aid for
physicians interpreting
screening FFDM acquired with
compatible mammography
systems, to identify findings and
assess their level of suspicion. | A concurrent reading aid for
physicians interpreting prostate
MRI exams of patients presented
for high-risk screening or
diagnostic imaging, to identify
regions suspicious for prostate
cancer and assess their likelihood
of malignancy. | A concurrent reading aid for
physicians interpreting prostate
MRI examinations acquired
according to the PI-RADS
standard, to identify findings and
assess their level of suspicion. | Equivalent -
Justified in
Section V. |
| Intended
patient
population | Women undergoing screening
mammography | Population of biological males
with a prostate gland undergoing
screening or clinical MRI exams.
This includes biological males of
all ages with clinical indicators
suggestive of possible prostate
cancer or with family history of
prostate cancer. | Adult men (40 years and older)
with suspected prostate cancer
undergoing prostate MRI
without prior treatment of the
prostate gland (treatment-naïve). | Equivalent -
Images
captured from
different
patient
populations
are
standardized
to be
processible by
CAD devices. |
| Anatomical
region of
interest | Breast | Prostate gland | Prostate gland | Equivalent -
Images
captured from
different
anatomical
regions are
standardized
to be
processible by
CAD devices. |
| Intended
Users | physicians qualified to read
screening mammograms | Physicians qualified to read and
interpret prostate MRI exams
consistent with ACR
recommendations in the context
of PI-RADS v2 | Radiologists qualified to read
prostate MRI | Equivalent -
- Users are
qualified to
read radiology
images using
CAD devices |
| Mode of
action | Software that applies algorithms
for recognition of suspicious
calcifications and soft tissue
lesions to detect and
characterize findings in
radiological breast images and
provide information about the
presence, location, and
characteristics of the findings to
the user. | Software that applies algorithms
for recognition of suspicious
tissue regions in Prostate MR
images to provide information
about the presence, location, and
level of suspicion of the findings. | Software that applies algorithms
for recognition of suspicious
tissue regions in Prostate MR
images to provide information
about the presence, location, and
level of suspicion of the
findings. | Equivalent -
Both predicate
and subject
devices use
algorithms to
detect findings
and provide
diagnosis. |
| Method of
Use | Concurrent | Concurrent | Concurrent | Identical |
| Attribute | Predicate Device
TransparaTM
K181704 | Reference Device
ProstatIDTM
K212783 | Subject Device
Prostate MR AI (VA10A) | Equivalency
Analysis |
| Visualization
Features | Computer aided detection
(CAD) marks to highlight
locations where the device
detected suspicious
calcifications or soft tissue
lesions. Decision support is
provided by region scores on a
scale ranging from 0-100, with
higher scores indicating a higher
level of suspicion. | ProstatID does not include a
standalone graphical user
interface. Rather, ProstatID
outputs are in DICOM format and
may be viewed on DICOM-
compliant image viewers. | Prostate MR AI does not include
a standalone graphical user
interface. Rather, it is a plug-in
device that is intended to be used
with a separate hosting
application that allows the user
to view the original case, as well
as confirm, reject, edit, or add
lesions, their contours and
Scores. | Equivalent -
The difference
between
predicate and
subject
devices is
justified by
using the
reference
device which
also does not
have UI while
it retains the
same level of
safety and
effectiveness. |
| Technical Characteristics | | | | |
| Design | Software only device | Software only device | Software only device | Identical |
| Automatic
Segmentation | Yes | Yes | Yes | Identical |
| Algorithm | Artificial intelligence algorithm
trained with large datasets of
biopsy proven examples of
breast cancer, benign lesions
and normal tissue. | Neural network trained on a
database of reference normal
tissues and abnormalities with
known ground truth. | Artificial intelligence algorithm
trained on a database of prostate
MR image series acquired
according to the PI-RADS
standard (non-contrast T2W and
DWI image series), and
corresponding radiological
and/or biopsy findings. | Identical - all
trained AI
algorithms |
| Alteration
original
image | No | No | No | Identical |
| Data
acquisition
protocol | Screening mammograms | Prostate MRI image series | Prostate MRI image series
acquired according to the PI-
RADS standard | Equivalent -
Acquired
images are
qualified for
CAD device
processing |
| Input | Medical images provided in a
DICOM format | Medical images provided in a
DICOM format | Medical images provided in a
DICOM format | Equivalent -
Devices all
use
standardized |
| Attribute | Predicate Device
Transpara™
K181704 | Reference Device
ProstatID™
K212783 | Subject Device
Prostate MR AI (VA10A) | Equivalency
Analysis |
| Output | • Marks placed on suspicious
soft tissue lesions and
suspicious calcifications
• Region-based scores
indicating the likelihood that
cancer is present
• Overall score indicating the
likelihoood that cancer is
present on the mammogram | • Marks locations suspicious
of lesions
• Provides region scores with
higher scores indicating a
higher level of suspicion
• Provides single exam score
that synthesizes features | • Automatically segments the
contours of the prostate gland
• Automatically segments the
parts of the prostate that belong
to the periheral zone (PZ) and
that do not belong to the
peripheral zone (non-PZ),
respectively
• Calculation of a "Suspicion
Map" that indicates lesions
suspicious for cancer
• For each detected lesion:
o Lesion contours
o Rating of severity ("Score")
on a scale from 3 to 5 (in
steps of 1). Score is
generated by an algorithm
trained on the correlation of
prostate MRI with PI-RADS
scores provided by
radiologists and results of
lesion-targeted biopsy.
• A "Level of Suspicion"
(LoS) on a scale from 60 to
100 (in steps of 1) as a fine
granular measure of the
algorithm's suspicion for the
presence of a significant
lesion, based on training on
PI-RADS scores provided by
radiologists and results of
lesion-targeted biopsy | Equivalent -
as far as the
different
body regions
allow. PZ vs.
non-PZ is
prostate
specific and
is relevant for
PI-RADS
evaluation.
Instead of
marks for
suspicious
locations in
the image, the
subject
device
provides
lesion
contours and
a "suspicion
map". |
| Attribute | Predicate Device
Transpara™
K181704 | Reference Device
ProstatID™
K212783 | Subject Device
Prostate MR AI (VA10A) | Equivalency
Analysis |
| Score | Finding level:
Continuous score 1-100
indicating the level of suspicion
of malignancy (from low
suspicion to high suspicion).
Breast level:
None
Exam level:
10-point scale score indicative
of higher frequency of cancer
positive | Finding level:
Scores on a continuous scale
from 0 to 1 that accompany the
overlay markings of suspicious
locations
Case level:
Suggested level of suspicion
(LoS) or overall PI-RADS exam
score | Finding level:
Rating of severity ("Score") on a
scale from 3 to 5 (in steps of 1).
The Score is generated by an
algorithm trained on the
correlation of prostate MRI with
PI-RADS scores provided by
radiologists and results of lesion-
targeted biopsy
A "Level of Suspicion" (LoS) on
a scale from 60 to 100 (in steps
of 1) as a more granular measure
of the algorithm's suspicion for
the presence of a significant
lesion, based on training on PI-
RADS scores provided by
radiologists and results of lesion-
targeted biopsy
Prostate level (= Exam Level):
LoS derived from Finding Level
results as maximum LoS over all
findings (on a scale of 60-100 in
steps of 1), or (in the absence of
findings) as maximum of
internal Suspicion Map over
prostate segmentation (on a scale
of 1-59 in steps of 1) | Equivalent -
The
differences
between
predicate and
subject
devices can be
justified by
using the
reference
device which
apply
continuous
finding scale
and share the
same PI-
RADS LoS
exam score. |
| Finding
discovery | Findings are by-default
displayed when score is equal or
higher than 5.
Upon user request for findings
of score equal or less than 4. | Findings are added to post-
processed T2W image in DICOM
format as a colorized translucent
overlay to highlight locations, as
overlay scores on a continuous
scale from 0 to 1, and as a
suggested LoS or overall PI-
RADS exam score. | Upon activation of the plug-in,
findings are displayed in the
hosting workflow as
segmentations on the T2W
image series with corresponding
ratings by Scores (value range:
3-5), and as an overall exam
Score. | Equivalent -
Finding
results of both
predicate and
subject
devices can be
displayed. |
| Attribute | Predicate Device
Transpara™™
K181704 | Reference Device
ProstatID™
K212783 | Subject Device
Prostate MR AI (VA10A) | Equivalency
Analysis |
| Performance | Reader study with 6720 reads:
• fully crossed multi-reader,
multi-case design
• 240 cases
• 14 radiologists | Reader study with 2700 reads:
• 150 cases were read by 9
trained physicians in two
separate reads - first without
ProstatID, and second with
ProstatID. | Reader study with 4080 reads:
• split-plot multi-reader, multi-
case design
(comprising 2 fully crossed
multi-reader, multi-case splits)
• 2× 170 = 340 cases from 2
consecutively acquired cohorts
• 2× 6 = 12 radiologists
• Two analysis scenarios:
a) all cases, all readers
b) only cases with biopsy, or a
negative MRI and ≥12
months negative follow-up
by PSA or MRI (exclusion
of 34 cases), and only case-
reader pairs in which
readers were not involved
in prostate MRI reading at
the institution providing the
case during the time of data
collection (exclusion of
459 case-reader pairs) | Equivalent -
The difference
between the
predicate and
subject
devices can be
justified by
comparing
with the
performance
of the
reference
device. Reader
studies of both
reference and
subject
devices used
similar
statistically
meaningful
case numbers,
qualified
readers,
acceptance
criteria and
yielded
comparable
study results. |
| | Case-level ROC:
• radiologists' AUC unaided =
0.866
• radiologists' AUC aided =
0.886
• mean difference: +0.020,
95% C.I. [0.010, 0.030] | Case-level ROC for discriminating
Gleason Score ≥ 7 in those 130 out
of the 150 cases that had biopsy
results:
• radiologists' AUC unaided =
0.629
• radiologists' AUC aided =
0.671
• mean difference: +0.042,
95% C.I. [0.005, 0.080] | Case-level ROC for
discriminating Gleason Score ≥ 6:
• radiologists' AUC (unaided) =
a) 0.676; b) 0.658
• radiologists' AUC (aided) =
a) 0.701; b) 0.695
• mean difference:
a) +0.025,
95% C.I. [0.001, 0.049]
b) +0.037,
95% C.I. [0.011, 0.063] | |
| | | Lesion-level wAFROC:
• radiologists' AUC unaided =
0.387
• radiologists' AUC aided =
0.430
• mean difference: +0.043,
95% C.I. [0.003, 0.083] | Lesion-level wAFROC:
• radiologists' AUC unaided =
a) 0.734; b) 0.772
• radiologists' AUC aided =
a) 0.769; b) 0.834
• mean difference:
a) +0.035,
95% C.I. [0.002, 0.068]
b) +0.030,
95% C.I. [0.008, 0.052] | |
VI. COMPARISON OF FEATURES AND SPECIFICATIONS WITH
8
9
10
Healt
Prostate MR AI (VA10A) Traditional 510(k) Submission
11
Image /page/11/Picture/0 description: The image contains the logo for Siemens Healthineers. The word "SIEMENS" is written in teal, and the word "Healthineers" is written in orange below it. To the right of the words is a graphic of orange dots arranged in a circular pattern.
12
Prostate MR AI (VA10A) Traditional 510(k) Submission
The predicate and subject devices are both Computer Assisted Detection and Diagnosis Software devices to assist radiologists to detect and diagnose diseases based on the radiological images. They share
13
Image /page/13/Picture/1 description: The image shows the Siemens Healthineers logo. The word "SIEMENS" is written in teal, and the word "Healthineers" is written in orange. To the right of the word "Healthineers" are several orange dots.
significant similarities in the functionalities of detection, assessment and characterization of human tissue abnormalities on radiological images using AI/ML augmented technologies.
While technological characteristics of the predicate device Transpara™, e.g., network architecture and training process, remain unknown to the public, the verification and validation testing of the subject device Prostate MR AI demonstrate that the device can perform prostate gland segmentation, lesion detection and classification as intended, meeting all the design inputs. The risks associated with the algorithm development are mitigated as far as possible. The detection and diagnosis accuracy of the subject device was assessed to validate the appropriateness and implementation of the intended use. In addition, a Multi-Reader/Multi-Case Study was performed to demonstrate that radiological reading aided by Prostate MR AI yields better diagnostic performance than unaided reading. Evidence provided within this submission demonstrates conformance with special controls for software as medical devices. The differences in technological characteristics between the subject device Prostate MR AI and the predicate device do not constitute any new intended use and do not raise new questions of safety and effectiveness.
The other differences between the subject device and the predicate device, notably the body regions targeted, are justified based on the reference device ProstatIDTM.
In summary, Siemens is of the opinion that Prostate MR AI (VA10A) does not raise new or different questions of safety or effectiveness and is substantially equivalent to the currently marketed predicate device Transpara™ (K181704).
VII. PERFORMANCE DATA
The following performance data were provided in support to demonstrate similarities to the predicate / previously cleared device.
Summary of Software Verification and Validation
No performance standards for CADe/CADx have been issued under the authority of Section 514. Nonclinical testing was conducted for the device Prostate MR AI (VA10A) during product development. The features described in this Premarket Notification were supported with verification and validation testing.
Siemens Healthineers claims conformance to the following recognized consensus standards:
- ISO 14971 Third Edition 2019-12
- IEC 62304 Edition 1.1 2015-06 CONSOLIDATED VERSION
- IEC 82304-1 Edition 1.0 2016-10 .
Software documentation for Basic Documentation Level per FDA's Guidance Document "Content of Premarket Submissions for Device Software Functions" issued on June 14, 2023 is also included as part of this submission. The performance data demonstrates continued conformance with special controls for medical devices containing software. Non-clinical tests were conducted on the device Prostate MR AI during product development.
The Risk Analysis was completed, and risk control implemented to mitigate identified hazards. The testing results support that all the software specifications have met the acceptance criteria. Testing for verification and validation for the device was found acceptable to support the claims of substantial equivalence.
14
Nonclinical Test Summary
21CFR § 807.92(b)(2)
Automatic prostate segmentation
To monitor the performance of the automatic prostate segmentation, an automated test routine was established that compared the segmentation result of 222 transversal T2 series from more than 10 clinical sites against ground truth generated by radiologists. The image data base of this formal test includes ~31% cases acquired with a 1.5T system and ~69% cases acquired with a 3T system. Of the test data 41%, 27%, and 32% were scanned on Siemens, Philips, and GE MR systems, respectively.
The reference standard was established through pixel-wise consensus, which was built based on annotation results of three expert radiologists. Two metrics were used for the evaluation. One was based on the Dice score of pairs of segmentation masks, and the other was based on the normalized volume difference based on the computed volumes of masks.
The overall results demonstrate the following:
- . The median of the Dice score between the AI algorithm results and the corresponding ground truth masks exceeds the threshold of 0.9.
- . The median of the normalized volume difference between the algorithm results and the corresponding ground truth masks is within a ±5% range.
- . The AI algorithm results as compared to any individual reader are statistically non-inferior based on variabilities that existed among the individual readers within the 5% margin of error and 5% significance level.
Prostate lesion detection and classification
To monitor the performance of the automatic prostate lesion and classification, an automated test routine was established that compared the result of the automatic lesion detection and classification result for
- . 105 cases from 6 sites against a ground truth generated by radiologists,
- 115 cases from 6 sites against a ground truth generated by prostate biopsy, as well as ●
- . 340 cases from the multi-reader multi-case study.
The reference standard for the radiology ground truth was established through consensus reading of three expert radiologists in prostate MRI reading. The reference standard for the pathology ground truth was established through biopsy results for the same patient. Therefore, in the pathology ground truth, only case level annotation was available, while in the radiology ground truth, lesion level annotation was also included. Sensitivity and false positive rate per case was used to evaluate the performance of prostate lesion detection and classification, and accuracy was used to determine the performance of the PI-RADS classification.
The overall results demonstrated the following:
- . The case level sensitivity of the lesion detection is equal or greater than 0.80 for both radiology and pathology ground truth.
- . The false positive rate per case of the lesion detection is smaller than one false positive per case for radiology ground truth.
- . The accuracy of the PI-RADS classification of radiology ground truth lesions detected by the algorithm is equal or greater than 0.8.
15
- The non-inferior performance of the subject device in GE vs Siemens and African American vs ● non-African American cases, and in cases with peripheral zone vs non- peripheral lesions
All pre-specified criteria for non-clinical testing were therefore met.
Clinical Test Summary
21CFR § 807.92(b)(2)
For assessment of the clinical performance of the device, a reader study was conducted with the objective to determine whether individual radiologists perform better with than without Prostate MR AI in the task of identifying cases of treatment-naïve men that are suspicious of prostate cancer based on a prostate MRI examination acquired according to the PI-RADS standard. In order to compare the reading performance of radiologists with and without the aid of Prostate MR AI, a study design with independent arms for aided and unaided reading is appropriate to test both reading conditions. The study was set up as a multi-reader multi-case (MRMC) study in a paired split-plot design, which essentially combined two fully-crossed MRMC (sub-)studies conducted in parallel, each using half of the overall readers and half of the overall cases.
Within each of the two MRMC sub-studies, multiple radiologists performed two reads of multiple prostate MRIs, one without and one with the support of Prostate MR AI. Between the sessions there was a wash-out time interval of at least 28 days. All of the 12 Readers were American Board of Radiology certified and selected to reflect a spectrum of experience and practice type background.
Study cases were selected retrospectively to be representative of the population of U.S. men undergoing prostate MRI without prior treatment of the prostate gland.
For inclusion in the reader study, 340 cases were selected. The cases from two US sites were consecutive and not enriched for positive cases. To adequately represent African American patients, who are more likely to be diagnosed with prostate cancer, present at an earlier age, and are more likely to have advanced disease at diagnosis', additional consecutive patient cases specifically from men of African descent were included to ensure that at least 13% of the study cases are of Black or African American ethnicity2.
The cases for the reader study were kept completely separate from those used for the training of the Prostate MR AI algorithm.
The primary endpoint of the study was based on the comparison of case-level diagnostic performance of aided and unaided reads using the reader-provided case-level LoS (RLoS), that is, a PI-RADS scale with additional intermediate steps of 0.5. The resulting finer granularity, which is not used in clinical practice, was introduced for improved accuracy of the comparative assessment. Biopsy results (with the criterion Gleason Grade Group GGG greater or equal to 1. i.e. any cancer on biopsy), and in case they were not available, PSA density and follow-up were used to determine the reference standard.
However, as consecutive cases were used in order to avoid selection bias, not all cases had unquestionable ground truth in terms of biopsy information or a negative MRI with negative follow-up of at least 12 months by PSA density or MRI. Therefore, two alternative evaluations were performed in
1 Smith ZL, Eggener SE, Murphy AB: African-American Prostate Cancer Disparities. Curr Urol Rep 18, 81 (2017)
2 https://www.census.gov/quickfacts/fact/table/US/IPE120219 (accessed on Feb 6th, 2022) gives an estimate of 13.4% for the U.S. population of Black or African American ethnicity for July 15, 2021
16
parallel: a fully inclusive and a maximally restrictive analysis.
In the fully inclusive scenario, all cases and readers were considered, using a probabilistic treatment if definite reference standard labels were not available.
In the maximally restrictive scenario, only cases with biopsy information, or a negative MRI and 12month negative follow-up by PSA density or MRI, were used. Moreover, reader-case pairs were excluded if readers were involved in clinical prostate MRI evaluation at the institution that provided the cases during the period of data collection.
In the fully inclusive analysis, the average area under the ROC (Receiver Operating Characteristic) curve (AUROC) improved from 0.6758 in unaided reading to 0.7010 in aided reading, with a difference of 0.0252 (95% C.I. [0.0011, 0.0493]; P=0.040).
In the maximally restrictive analysis, AUROC improved from 0.6579 in unaided reading to 0.6948 in aided reading, with a difference of 0.0368 (95% C.I. [0.0108, 0.0628]; P=0.006).
In either analysis, the improvement was statistically significant and the primary endpoint thus met.
The following figures show the pooled ROC curves and RLoS ≥ 3 operating points for the discrimination of any cancer on biopsy (Gleason Grade Group ≥ 1) in unaided (orange) and aided reading (blue). The grey curve denotes AI standalone performance (with 95% confidence interval envelope); left: fully inclusive analysis, right: maximally restrictive analysis.
Image /page/16/Figure/9 description: The image contains two identical ROC curves, which plot sensitivity versus 1-specificity. Both curves show a blue line, an orange line, and a gray dotted line. The blue line is generally above the orange line, indicating better performance. There are also gray dots on the plot.
In the fully inclusive analysis, the average sensitivity/specificity of the Readers at a case-level RLOS threshold of ≥ 3 was 0.57 (95% C.I.: [0.49, 0.64]) / 0.72 (95% C.I. [0.64, 0.79]) in unaided and 0.60 (95% C.I .: [0.53, 0.68]) / 0.73 (95% C.I .: [0.65, 0.80]) in aided reading.
For the secondary endpoint (an analysis of the lesion-level reading performance in unaided and aided reading), AUwAFROC (area under the average weighted alternative free receiver operating characteristic) figures of merit were determined, using as reference standard consensus lesions with a consensus PI-RADS of at least 3 from majority voting among 3 experienced radiologists acting as Truthers. Against this, the Readers' corresponding true positives (correct lesion localizations) and false positive lesion detections were held, with their respective ratings.
Although in this analysis the reference standard is not in question for any case, for full equivalence to the primary analysis the evaluation was performed both for the fully inclusive and the maximally restrictive analysis scenario.
In the fully inclusive analysis, AUwAFROC improved in aided reading by 0.0350 (95% C.I.:[0.0020, 0.0681], P=0.037).
17
Image /page/17/Picture/1 description: The image contains the logo for Siemens Healthineers. The word "SIEMENS" is written in teal, and the word "Healthineers" is written in orange below it. To the right of the words is a graphic of orange dots.
In the maximally restrictive analysis, AUwAFROC improved in aided vs. unaided reading by 0.302 (95% C.I.: [0.0080,0.0520], P=0.008).
In either analysis, the improvement was statistically significant and the secondary endpoint thus met.
In a supplemental analysis for the fully inclusive analysis scenario, Fleiss' Kappa for interreader agreement in per-case PI-RADS scores was 0.283 (95% C.I.: [0.242, 0.322]) for unaided reads, and 0.371 (95% C.I.: [0.326, 0.411]) for aided reads, with a difference of 0.087 (95% C.I. [0.051, 0.125]). The improvement in Fleiss' Kappa between unaided and aided reads was statistically significant (P