(256 days)
DeepLook PRECISE is a software device that is integrated into medical Image Viewers and PACS workstations to assist trained professionals in measuring dimensions of objects within a region of interest (ROI) that is identified by the user in DICOM images. The generated information consists of an estimated greatest long-axis and greatest short-axis dimensions, area, volume, and margin of the objects. For illustration purposes, DeepLook PRECISE can optionally provide a colorization of the interior area defined by a margin. DeepLook PRECISE does not make clinical decisions nor is a decision-support tool. The information provided by the software and must not be used in isolation when making patient management decisions.
Lossy compressed mammographic images and digitized film screen images must not be reviewed for primary image interpretation. Mammographic images may only be interpreted using and FDA cleared monitor that offers at least 5 Megapixel resolutions and meets other technical specifications reviewed and accepted by FDA. Typical users of this system are trained physicians and radiologists.
DeepLook PRECISE, is not intended for use on mobile devices.
DeepLook PRECISE, is a software application that works embedded in PACS/EIS or OEM viewers. It provides automated measurement, replacing manual digital calipers currently used to measure objects in digital medical imaging.
It is a Windows OS service that uses XML messaging to receive Commands from the viewer application: it processes results independently and returns requested data to the viewer application via the same XML messaging. The software is not compiled within the viewer's application. As a result, integration is simplified and limited to establishing XML protocols for reciprocal Commands.
DeepLook PRECISE can operate on an individual workstation, local servers or in Cloudbased applications.
The software uses patented shape-recognition processes to analyze Regions of Interest (ROIs) identified by a trained medical user (i.e., imaging technologist, radiologist, phvsician or researcher). When requesting a measurement, the user triggers a set of XML Commands via the viewer interface. The primary Command is a request for a measurement of an object located within an ROI designated by the user using a mouse to click on the location of a suspected object.
DeepLook receives the Command, processes the area within the ROI and assembles the candidate shapes (and all relevant metrics) and returns a full set of displays (bundled and prioritized) that depict the possible boundaries of the object. To facilitate initial viewing, DeepLook PRECISE designates a default shape; this shape is recommended as the best depiction of the targeted margin of the object. The selection of the default shape from the display stack is determined by a set of deterministic algorithms that sort for the best candidate shape based on shape recognition ratios developed by DeepLook. Based on training and professional skill, a user can use simple keystroke Commands or a track ball or mouse wheel to move through the entire stack of alternative margins (shapes) and select the one that they conclude best represents the targeted object.
Once a candidate shape has been chosen, the user has the option to extend or contract any specific section of the margin in order to include or exclude a feature they deem relevant. Any alteration of the contours of the displayed margin will recalculate the overall measurement metrics (i.e., calculate two new axis measurements, and the area and estimated volume). The results will instantly appear with the modified margin.
The Commands to accept the default shape-display or select alternative shape-displays and any modification of selected shape-displays are all executed using keystroke or track ball functions that are initiated by the user through the viewer interface. Each resulting shape includes all relevant measurement metrics when displayed, allowing for quick comparison and selection.
For illustrative purposes, the software offers a colorization display of the internal shapes within a candidate margin: each of the shapes within the outer margin of a targeted object or anatomical structure is assigned a calibrated color. This is offered solely to assist the user in distinguishing the shape components when making a final selection. The colorization is not a decision-support or diagnostic tool.
The measurements and graphical display can be saved. Depending on the viewer configuration, this data can be saved 1) to the hard drive of the workstation; 2) to a server on the premises or in the Cloud; 3) in a structured report; or 4) configured to comply with DICOM data fields and saved to the PACS. The location of the saved display will be determined by the viewer manufacturer and/or the user.
The displays of DeepLook PRECISE described above offer the user consistent measurement of each shape. The modification functions provide the user with maximum flexibility to adjust the default shape or any other shapes in the display stack. The user can also decline all suggested shapes generated by DeepLook PRECISE and use standard mouse-operated digital calipers to measure the target object.
The final configuration of the user interface (i.e., keystrokes, track ball and hot-key Commands) will be determined by each vendor that integrates DeepLook PRECISE. To assist the vendor during integration, each XML Command and display option is illustrated separately in this manual. This provides the integrator with the option of selecting some or all the functions and to determine the best Commands to incorporate, consistent with its own user interface.
DeepLook PRECISE is stand-alone Windows OS service designed to permit maintenance and performance improvements without the need to modify established XML protocols used to send and receive Commands. This eliminates the need for additional vendor integrations when patches and/or updates are required to DeepLook PRECISE's Windows OS service.
The provided document, a 510(k) Premarket Notification for the DeepLook PRECISE device, primarily focuses on demonstrating substantial equivalence to a predicate device rather than providing a detailed study report of the device's performance against specific acceptance criteria. While it mentions nonclinical testing and meeting "predetermined acceptance criteria," it does not explicitly list the acceptance criteria or present the quantitative results of the study in a way that directly maps to such criteria.
Therefore, I cannot fully complete the requested table of acceptance criteria and reported device performance or provide detailed answers to many of the questions regarding the specifics of the performance study based solely on the provided text.
However, I can extract and infer some information and highlight what is missing.
Here's an attempt to answer your questions based on the provided text:
Acceptance Criteria and Device Performance Study Information for DeepLook PRECISE
Based on the provided K202084 510(k) Premarket Notification document, the following observations can be made regarding acceptance criteria and the performance study:
1. Table of Acceptance Criteria and Reported Device Performance:
The document states: "Validation testing indicated that as required by the risk analysis, designated individuals performed all verification and validation activities and that the results demonstrated that the predetermined acceptance criteria were met."
However, the specific "predetermined acceptance criteria" for performance (e.g., accuracy of measurements, precision, sensitivity, specificity, or inter-rater reliability metrics) and the quantitative results are NOT provided or detailed in the document. The document focuses on demonstrating substantial equivalence to a predicate device, and the nonclinical testing section broadly states that "all functions and has passed all predetermined testing criteria."
Therefore, a table of specific acceptance criteria and detailed reported device performance cannot be accurately constructed from the provided text.
2. Sample Size Used for the Test Set and Data Provenance:
The document does not explicitly state the sample size used for the test set or provide details on the data provenance (e.g., country of origin, retrospective or prospective nature of the data). It broadly refers to "nonclinical testing results" provided in the 510(k) submission, but these are not included in the provided snippets.
3. Number of Experts Used to Establish Ground Truth for the Test Set and Their Qualifications:
The document does not specify the number of experts used or their qualifications for establishing ground truth for the test set. It mentions "trained professionals in measuring dimensions of objects" and "trained physicians and radiologists" as typical users, but not those involved in ground truth establishment for the testing itself.
4. Adjudication Method for the Test Set:
The document does not describe any adjudication method for the test set.
5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study Was Done, and Effect Size:
The document does not mention a multi-reader multi-case (MRMC) comparative effectiveness study or any effect size related to human readers improving with AI vs. without AI assistance. The device is described as a "software device that is integrated into medical Image Viewers and PACS workstations to assist trained professionals," implying a human-in-the-loop use case, but the testing details for this specific interaction are not provided.
6. If a Standalone (Algorithm Only Without Human-in-the-Loop Performance) Study Was Done:
The description of the device's operation ("DeepLook receives the Command, processes the area within the ROI and assembles the candidate shapes... user can use simple keystroke Commands or a track ball or mouse wheel to move through the entire stack of alternative margins... Once a candidate shape has been chosen, the user has the option to extend or contract any specific section of the margin...") strongly suggests that the device is intended for human-in-the-loop use and not as a standalone diagnostic algorithm. The provided "Performance Data" section vaguely refers to "nonclinical testing," but does not distinguish between standalone and human-in-the-loop performance studies or provide data for either.
7. The Type of Ground Truth Used:
The document does not explicitly state the type of ground truth used (e.g., expert consensus, pathology, outcomes data). Given the device measures dimensions, common ground truth methods could include manual expert measurements with high precision, or potentially pathology if correlated with surgical removal. However, this is not detailed.
8. The Sample Size for the Training Set:
The document does not provide information regarding the sample size used for the training set.
9. How the Ground Truth for the Training Set Was Established:
The document does not provide information regarding how the ground truth for the training set was established.
Summary of what is present and what is missing:
- Present:
- Confirmation that nonclinical testing was performed and that predetermined acceptance criteria were met.
- References to relevant standards (ISO 14971, NEMA PS 3.1-3.20 (DICOM), IEC 62304, FDA Guidance on Cyber Security, FDA Guidance for Content of Premarket Submissions for Software Contained in Medical Devices).
- A high-level description of the device's function: assisting trained professionals in measuring dimensions (long-axis, short-axis, area, volume, margin) of user-identified ROIs in DICOM images.
- Acknowledgement that the device is not a decision-support tool and the information provided "must not be used in isolation when making patient management decisions."
- Missing (Critical for understanding the performance study details):
- Quantitative acceptance criteria for performance metrics (e.g., accuracy, precision).
- Specific quantitative performance results against these criteria.
- Details on the test dataset (sample size, data characteristics, provenance).
- Details on ground truth establishment (number/qualifications of experts, adjudication methods).
- Information on training data.
- Specifics of any MRMC or standalone performance studies, including effect sizes.
This document serves as an FDA 510(k) clearance letter and summary, which typically focuses on demonstrating substantial equivalence rather than providing a comprehensive scientific publication of the device's performance study results. Detailed study reports would be contained within the full 510(k) submission, which is not publicly released in its entirety.
§ 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).