Denti.AI Detect

Denti.AI Technology, Inc. • Radiology • 10/06/2023

510(k)IEC 62304 Class B

Device Purpose

Denti.AI Detect is an AI-powered software designed to enhance diagnostic capabilities for dental professionals by serving as a second reader for 2D dental radiographs to identify caries, periapical radiolucencies, and measure bone levels. Its methodology utilizes computer-assisted detection algorithms, with studies demonstrating its ability to improve reader performance in identifying dental pathologies.

Summary

Denti.AI Detect is a cloud-based, Computer-Assisted Detection (CADe) software intended for use by dental professionals, including dentists and dental specialists, as a second reader for 2D extraoral (panoramic) and intraoral (bitewing, periapical) dental radiographs. It is designed to assist in analyzing images from patients aged 22 years and older with permanent dentition by detecting and highlighting uncategorized regions of interest (ROIs) within the teeth area, specifically caries and periapical radiolucency. The software also aids in the measurements of mesial and distal bone levels associated with each tooth. Denti.AI Detect supports various image file formats such as JPEG, TIFF, PNG, BMP, and DICOM, and is not intended to replace a clinician's complete review or clinical judgment. The device's performance was evaluated through stand-alone bench testing and a clinical Multi-Reader Multi-Case (MRMC) study. For caries and periapical radiolucency detection, a study with 709 images yielded an across-category weighted Alternative Free-Response Receiver Operating Characteristic (wAFROC) Area Under the Curve (AUC) of 0.737 (95% CI: 0.713, 0.761). The bone measurement study, using 193 images, reported high sensitivity for Cementoenamel Junction (CEJ)-Bone measurements (e.g., 98.1% for bitewing) with Mean Absolute Errors (MAE) around 0.5-0.7 mm depending on image type and specific measurement. The MRMC study, involving 24 dental professionals and 154 images, demonstrated a statistically significant improvement in their diagnostic performance for detecting periapical radiolucencies and caries when assisted by Denti.AI Detect (aided wAFROC 0.809 vs. unaided wAFROC 0.784, p=0.029).

AI Regulatory Concerns

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

Keywords

CADeDental RadiographyCaries DetectionPeriapical RadiolucencyBone Level MeasurementArtificial IntelligenceSecond ReaderwAFROC AUCMRMC StudyDental Diagnostics

Device Metadata

Submission Number

K230144

Company

Denti.AI Technology, Inc.

Panel

Radiology

Product Code

N/A

Data Type

Images

Clinical Function

Assessment

AI Function

Analysis

AI Subclass

detection

Relevant Clinical Questions

Q1. How does the performance of Denti.AI Detect vary across different types and qualities of dental radiographs (e.g., underexposed, overexposed, varying sensor brands)?; What is the false positive rate for caries and periapical radiolucency detection, and how does it impact typical dental workflows?; Beyond the reported wAFROC AUC of 0.737 for combined caries and periapical radiolucency, what were the specific performance metrics (e.g., sensitivity, specificity, AUC) for each finding type individually?; How does the device's bone level measurement accuracy (e.g., MAE of 0.513 mm for bitewing CEJ-Bone) compare to manual measurements by experienced periodontists, and what is the clinical significance of this difference?; Considering the MRMC study showed a statistically significant improvement (difference 0.025 in wAFROC), what is the practical clinical impact of this improvement on patient outcomes or diagnostic efficiency?

Risk Classification

ESL Analysis

IEC 62304 ClassB

FDA Device ClassII

FDA Class II via 510(k) pathway; IEC 62304 Class B (software safety classification); Authorized before §524B enforcement

Learn how Parasoft automates IEC 62304 compliance testing →

Cybersecurity Evidence

Based on AI-generated device summaries, not the full FDA submission. Items not detected here may still be present in the complete submission package.

Cyber Score15/100

SBOM Evidence⚠ Not mentioned in summary

Cyber Risk Assessment⚠ Not mentioned in summary

Postmarket Plan⚠ Not mentioned in summary

§524B ApplicableNo

Supply Chain BOM Evidence

Based on the CycloneDX / OWASP ECMA-424 xBOM ecosystem. Detection is from the AI-generated summary.

SBOMSoftware Bill of Materials

- Not detected

ML-BOMMachine Learning Bill of Materials(algorithm)

✓ Detected

SaaSBOMSoftware-as-a-Service Bill of Materials(cloud)

✓ Detected

HBOMHardware Bill of Materials(sensor)

✓ Detected

CBOMCryptography Bill of Materials

- Not detected

VDR/VEXVulnerability Disclosure Report / Vulnerability Exploitability eXchange

- Not detected

OBOMOperations Bill of Materials

- Not detected

Analysis:

• Analysis is based on the AI-generated device summary, not the full FDA submission package. Click "Analyze Full FDA PDF" below to scan the actual submission document.

• Evidence of ML-BOM, SaaSBOM, HBOM detected in AI-generated device summary.

• This device was authorized before §524B took effect. Cybersecurity evidence requirements were less prescriptive at the time.

• Postmarket cybersecurity monitoring plan not clearly documented in AI-generated device summary.

• Not detected in AI-generated device summary: SBOM, CBOM, VDR/VEX, OBOM. These may be present in the full FDA submission.

Preparing a similar submission?

ESL takes complete ownership of your software evidence work - SBOM & CVE remediation with SBOMator™, static analysis & unit testing with Parasoft, traceability, and full IEC 62304 lifecycle support.

Assess Your Readiness →Or contact ESL directly