Search Results

DentalMonitoring is a medical device software using image processing algorithms to analyze pictures of the oral cavity (hereinafter Scans). Scans are taken using the DM App, a smartphone, and the manufacturer's proprietary hardware products. Scans are taken by the patient, a non-healthcare professional, or a healthcare professional. The Scan is taken in healthcare facilities, such as a dental practice, or in a non-healthcare environment, such as the patient's own home. For some clinical parameters. DentalMonitoring requires a 3D Model. The product is designed to assist healthcare professionals in remotely monitoring orthodontic treatments and treatment progress. The results of DentalMonitoring are intended to be used as an aid in diagnosis and monitoring, not on a stand-alone basis for clinical decision-making. DentalMonitoring is indicated for use for patients over the age of 6 and reports results solely on permanent teeth.

DentalMonitoring can monitor the following clinical parameters:

• . oral hygiene: dental plaque / food residue;
• 트 soft tissue statement: gingival recession, black triangle;
• . dental statement: closure of extraction space, tooth wear;
• . alignment: closure of all anterior spaces; and
• . dental occlusion:
  • o in 2D Monitoring: midline deviation, overbite/open bite, overjet
• o in 3D Monitoring: canine class, midline deviation, overbite/open bite, overjet
  Additionally, the following clinical parameters specific to orthodontic treatment types or phases can be monitored using DentalMonitoring:
• for aligner treatments: tracking (seat/unseat), attachment loss, button loss; .
• . for braces: bracket debonding, tie loss, self-ligating clips, passive archwire and auxiliaries; and
• for thermoformed retainers: tracking (seat/unseat). I
  Based on an initial 3D Model provided by a healthcare professional. DentalMonitoring can also provide healthcare professionals with 3D Models representative of the patient's dentition and treatment progress. This device is a prescription device and is not intended for over-the-counter use

DentalMonitoring is a standalone software with embedded artificial intelligence comprising machine learning and locked neural network algorithms. The product is composed of these parts: the Dashboard, the DM App along with a DM Cheek Retractor and DM ScanBox, and the Data Analysis Platform.

The four steps are detailed below:

• 1. Patient profile set-up: The healthcare professional sets up the patient profile through a web- based interface, Dashboard, accessible at www.dental-monitoring.com/doctor. During the patient profile set-up, the healthcare professional is prompted to set up a Protocol in order to select the clinical parameters they wish to monitor in accordance with the patient's treatment. During the set-up, the healthcare professional also determines the appropriate frequency at which the patient should use the DM App to take pictures of their intraoral cavity.
• 2. Intraoral pictures acquisition: A set of pictures of the patient's intraoral cavity, hereinafter referred to as a Scan, is taken using the DM App along with a DM Cheek Retractor and DM ScanBox. The DM Cheek Retractor is an intraoral retraction device for the cheeks and lips to allow for sufficient space for image capture. The DM ScanBox is an extraoral holding apparatus for the smartphone that attaches to the DM Cheek Retractor. The DM App guides the user through the appropriate steps in order to obtain a complete Scan. This DM App is a mobile application installed on a smartphone; thus, the pictures are captured through the smartphone's main camera allowing the procedure to be totally non-invasive to the patient.
• 3. Analysis of acquired intraoral pictures: The Scan is processed through the Data Analysis Platform. The Data Analysis Platform includes a technical processing phase and clinical processing phase, the latter being a clinical analysis to determine if any event has occurred within the clinical parameters the healthcare professional has set up to be monitored. The Data Analysis Platform uses AI comprising image processing algorithms and neural networks.
• 4. Communication of results of the analysis performed through the Data Analysis Platform are communicated to the healthcare professional through the Dashboard on one hand, and to the patient through the DM App on the other hand. Results are shared with the healthcare professional in an exhaustive fashion, providing them with detailed information. As for the patient, the results are communicated in accordance with the rules defined by the healthcare professional in the Protocol applied to the patient.

DentalMonitoring enables HCPs to adapt the use of the product according to their needs depending on each patient's orthodontic treatment. DentalMonitoring comprises two types of monitoring: 2D Monitoring and 3D Monitoring. Some clinical parameters are specific to either 2D Monitoring, or 3D Monitoring.

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

1. Table of Acceptance Criteria and Reported Device Performance

The acceptance criteria are generally demonstrated through the clinical performance results (Sensitivity, Specificity, Slope, Intercept, and Bias) of each monitored parameter. The document doesn't explicitly state quantitative acceptance thresholds for each metric before the study, but rather presents the study results as adequate to demonstrate performance. The "Final Study Results" for each indication serve as the reported device performance.

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

The studies are categorized as Retrospective and Prospective. The data provenance consistently states United States.

• Retrospective Study (21-001):

  • Data Provenance: Retrospectively collected Scans from the US.
  • Sample Size (Test Set):
    • Dental plaque / food residue: 5,064 evaluations (derived from the 2x2 table)
    • Gingival recession: 870 evaluations
    • Black triangle: 1,137 evaluations
    • Closure of extraction space: 478 evaluations
    • Tooth wear: 1,066 evaluations
    • Closure of all anterior spaces: 713 evaluations
    • Tracking (seat/unseat): 3,323 evaluations
    • Attachment loss: 765 evaluations
    • Button loss: 659 evaluations
    • Bracket debonding: 659 evaluations
    • Tie loss: 653 evaluations
    • Self-ligating clips: 647 evaluations

• Prospective Studies:

  • Occlusion Study (21-002):
    • Data Provenance: Prospectively collected from four sites located in the United States.
    • Sample Size (Test Set):
      • Canine class (3D): 215 patients, 297 results.
      • Midline deviation (2D): 277 patients, 277 results.
      • Midline deviation (3D): 291 patients, 294 results.
      • Overbite / open bite (2D): 285 patients, 301 results.
      • Overbite / open bite (3D): 287 patients, 298 results.
      • Overjet (2D): 208 patients, 245 results.
      • Overjet (3D): 263 patients, 292 results.
  • Archwire & Auxiliaries Study (21-003):
    • Data Provenance: Prospectively collected from six sites located in the United States.
    • Sample Size (Test Set):
      • Passive archwire and auxiliaries (2D & 3D): 730 evaluations.
      • Patient counts: 273 (2D) and 269 (3D).
  • Updated 3D Model Study (21-004):
    • Data Provenance: Prospectively collected from seven sites located in the United States.
    • Sample Size (Test Set): 250 patients, 536 results.

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

• Retrospective Study (21-001):

  • Number of Experts: A panel of three independent orthodontists.
  • Qualifications: "orthodontists" (no specific years of experience or board certification mentioned, but implied clinical expertise in orthodontics).
  • Adjudication Method: "In case of non-dominant results, a consensus was established by the same group of three independent orthodontists."

• Prospective Studies (21-002, 21-003, 21-004):

  • For occlusion (21-002), archwire & auxiliaries (21-003), and Updated 3D Model (21-004), the "Reference Method results were generated by measuring the occlusion parameters undergoing evaluation on 3D Models... Measurements were done manually and performed in three replicates using a CAD/CAM software." This implies that the ground truth was established through precise measurements using specialized software rather than direct human expert assessment on the dental images themselves for these specific parameters. The text doesn't specify if these measurements were performed by experts (e.g., orthodontists) or trained technicians, but the context of "Reference Method" implies a high standard of accuracy.

4. Adjudication Method for the Test Set

• Retrospective Study (21-001): The initial assessment by three independent orthodontists was followed by consensus review for "non-dominant results" by the same group of three orthodontists.
• Prospective Studies (21-002, 21-003, 21-004): The ground truth was established through manual measurements with CAD/CAM software, performed in three replicates. This is a technical measurement-based adjudication rather than human consensus on observations.

5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study was Done

No, a Multi-Reader Multi-Case (MRMC) comparative effectiveness study was not reported in this document. The studies primarily focused on the standalone performance of the AI algorithm against a "Reference Method" (which was expert consensus or precise measurements). There is no mention of human readers with AI assistance being compared to human readers without AI assistance.

6. If a Standalone (i.e., algorithm only without human-in-the-loop performance) was Done

Yes, extensive standalone performance testing of the algorithm only (without human-in-the-loop) was done. This is explicitly stated in the "AI/ML standalone testing" section and demonstrated by the clinical performance studies that compare the DentalMonitoring results directly to the "Reference Method" (ground truth). The device's indication for use also clarifies that it is "intended to be used as an aid in diagnosis and monitoring, not on a stand-alone basis for clinical decision-making," but the performance reported is of the algorithm itself.

7. The Type of Ground Truth Used

• Retrospective Study (21-001): Expert Consensus by a panel of three independent orthodontists.
• Prospective Studies (21-002, 21-003, 21-004): Precise Measurements using CAD/CAM Software on 3D models derived from intraoral scanners, performed in three replicates. This acts as a highly accurate, objective ground truth for quantitative parameters.

8. The Sample Size for the Training Set

The document does not explicitly state the sample size used for the training set. It only describes the "Clinical Investigation Program" which provides data to support performance, implying this is the test data. The "AI/ML standalone testing" section mentions that "Neural networks and algorithms were classified into three families... A unique protocol was established per family in order to perform the evaluation," but it doesn't detail the training phase or its dataset size.

9. How the Ground Truth for the Training Set Was Established

The document does not explicitly describe how the ground truth for the training set was established. The detailed descriptions of ground truth establishment are specifically for the test sets used in the retrospective and prospective clinical studies. It's generally assumed that similar rigorous methods are used for training data, but it's not detailed here.

Ask a specific question about this device