Search Results

DS Core Detect is a concurrent read, computer assisted detection software intended to aid the dentist in the incidental detection of permanent teeth that may be associated with periapical radiolucencies in CBCT images. The device is not intended as a replacement for a complete dentist's review or a replacement of their clinical judgment. The device is to be used by licensed dentists. The device is designed for patients who already have CBCT data obtained for other reasons. DS Core Detect does not generate X-ray images and does not provide an indication for taking such images.

The device is intended for adults aged 22 years and older with permanent teeth.

The DS Core Detect device is a cloud-based SaMD that is a Computer Aided Detection "CADe" Machine Learning "ML" enabled device for the aid in incidental detection of permanent teeth that may have periapical radiolucency "PARL" on a CBCT image. The device is designed for patients who already have CBCT data obtained from a CBCT imaging system for other reasons. DS Core Detect does not generate X-ray images and does not provide an indication for taking such images.

The device itself is accessed by the dental practitioner through their dental image viewer. From within the dental viewer the user can view the CBCT with results from the subject device. The device will provide a list of teeth that may potentially have PARL and a corresponding PARL "map" to visualize where the PARL is detected. The user has completely autonomy to confirm or reject a tooth that is flagged by the device as potentially PARL positive.

The intended users of the device are trained and licensed dentists. The intended patient population for the device is adult patients with permanent teeth at least age 22.

Here's a breakdown of the acceptance criteria and study details for the DS Core Detect device, based on the provided FDA 510(k) clearance letter:

Acceptance Criteria and Device Performance for DS Core Detect

The DS Core Detect device has two operating points: High sensitivity and Standard sensitivity. The acceptance criteria are based on the device's standalone performance in terms of sensitivity and specificity, and augmented by a Multi-Reader Multi-Case (MRMC) study to evaluate the impact on human reader performance.

1. Table of Acceptance Criteria and Reported Device Performance

The provided document does not explicitly state "acceptance criteria" in a pass/fail format with exact numerical thresholds. However, the performance metrics reported from the standalone study and the MRMC study implicitly define the expected performance for clearance. The key performance indicators are Sensitivity and Specificity at different operating points for the standalone performance, and the improvement in AUC, Sensitivity, and Specificity for human readers when aided by the device.

Table 1: DS Core Detect Performance Metrics

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

• Sample Size (Test Set): The standalone study used 306 CBCTs. The MRMC study does not explicitly state the number of cases but refers to "Tooth Level AUC results," implying that the unit of analysis is individual teeth within the CBCTs.
• Data Provenance: The data consisted of patients from various clinical sites across the U.S. It is retrospective, as the device is designed for patients who already have CBCT data obtained for other reasons. The dataset included adult patients aged 22 and older with permanent teeth.

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

The document states that the standalone study assessed performance compared to a "consensus ground truth." However, it does not specify the number of experts used to establish this ground truth, nor their specific qualifications (e.g., "radiologist with 10 years of experience").

4. Adjudication Method for the Test Set

The document mentions "consensus ground truth" for the standalone study. This implies an adjudication process where multiple experts collectively determine the ground truth. However, the specific method (e.g., 2+1, 3+1, majority vote, etc.) is not described.

5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study was done, If so, what was the effect size of how much human readers improve with AI vs without AI assistance

• Yes, an MRMC comparative effectiveness study was done. The document explicitly states: "The MRMC study evaluated reader performance with and without the assistance of the device, in a fully crossed study design."
• Effect Size of Improvement:
  • AUC Improvement: Human readers improved by 0.224 (from 0.415 unaided to 0.639 aided), with a 95% CI of [0.124, 0.324].
  • Sensitivity Improvement: Human readers improved by 0.227 (from 0.421 unaided to 0.649 aided), with a 95% CI of [0.124, 0.330].
  • Specificity Change: Human readers experienced a slight decrease in specificity of -0.016 (from 0.962 unaided to 0.946 aided), with a 95% CI of [-0.034, 0.002].

6. If a standalone (i.e. algorithm only without human-in-the-loop performance) was done

• Yes, a standalone study was done. The document states: "The standalone study assessed the performance of the device compared to a consensus ground truth." The performance metrics (Sensitivity and Specificity) for the "High sensitivity" and "Standard sensitivity" operating points are reported from this standalone study.

7. The type of ground truth used

The ground truth used for both the standalone study and likely as the reference for the MRMC study was expert consensus. It is specifically referred to as "consensus ground truth."

8. The sample size for the training set

The document does not specify the sample size for the training set. It only details the test set used for performance evaluation.

9. How the ground truth for the training set was established

The document does not specify how the ground truth for the training set was established.

Ask a specific question about this device