AI Dental Disease Detection

Detect
problems.
Protect
smiles.

Advanced AI that analyses dental images with clinical accuracy to detect cavities, periapical lesions and more.

Real-time AI Analysis

Caries Detected

92%

🦷 Cavities High ↗

📊 Periapical Lesion Medium ↗

✨ Enamel Wear Low ↗

🛡

Clinically Validated

Before

AI Analysis

↔

How it Works

☁️

Upload Image

X-ray / Photo

🤖

AI Scans

Detects issues

🔍

Analyzes

In real-time

📋

Get Report

Instant insights

"

Early detection leads to better outcomes.

+2.5K

Happy Patients

▶

🎯

87.9%

Overall Accuracy

🦷

90.2%

Caries Accuracy

📊

81.2%

Periapical Accuracy

⚠️

2.3%

Missed Rate

i This benchmark is from FasterRCNN + Resnet50

Diagnostic Process

How DentOmni AI Works

Our system combines advanced deep learning models with state-of-the-art computer vision to analyze dental radiographs in seconds.

STEP 01

📤

Upload Radiograph

Simply upload a panoramic or periapical dental X-ray. The image is securely processed and optimized for the AI networks.

STEP 02

🤖

AI Processing

Our dual-model pipeline evaluates the image. Faster R-CNN localizes lesions with bounding boxes, while ResNet-50 validates classification probability.

STEP 03

🔬

Clinical Calibrations

Probability scores are run through thresholding logic to determine pathological presence, eliminating false positives.

STEP 04

📋

Report & Sign-off

An annotated diagnostic image is generated instantly, along with a clinician-ready PDF report available for immediate download.

Visual Processing Pipeline

Patient X-Ray

DICOM/JPEG/PNG

➔

Faster R-CNN
Lesion Localization

➔

ResNet-50
Classification Verification

➔

Annotated Image

Localized Bounding Boxes

➔

DentOmni Report

PDF Download

Test the Diagnostic Engine

Experience real-time clinical AI dental diagnostics yourself. Upload an image to analyze caries and periapical lesions.

Training Output & Metrics Library

Model Metrics & Architecture

Explore the deep learning metrics, activation maps, features, predictions, and loss charts obtained during our training runs. Toggle between models below.

Confusion Matrix

Detailed classification accuracy metrics across diagnosed classes, illustrating precise true vs. predicted counts.

Click to enlarge 🔍

Training Loss Curves

Monitors convergence of classification and bounding box regression losses across 70 training epochs.

Click to enlarge 🔍

Feature Maps

Visualizes feature maps from intermediate layers, depicting edges and local patterns detected by convolutions.

Click to enlarge 🔍

FPN Feature Pyramid

Shows multi-scale feature pyramids that capture both low-level detail (small caries) and high-level semantics.

Click to enlarge 🔍

Grad-CAM Heatmap

Displays network focus regions on radiographs, validating that the AI targets the correct pathological anatomy.

Click to enlarge 🔍

Precision-Recall Curves

Evaluates precision-recall trade-offs at varying Intersection over Union (IoU) thresholds for disease classes.

Click to enlarge 🔍

Confidence Distributions

Illustrates the separation in AI confidence scores between healthy tissues and confirmed pathological cases.

Click to enlarge 🔍

Accuracy Line Graphs

Tracks step-wise validation accuracy and mean Average Precision (mAP) gains over the training run.

Click to enlarge 🔍

K-Fold Cross Validation

Compares performance scores across K-Fold splits, proving the model's reliability across diverse data sets.

Click to enlarge 🔍

IoU Box Plots

Displays statistical distributions of overlap scores, highlighting high bounding box localization precision.

Click to enlarge 🔍

Dataset Class Balance

Details target labeling distributions and loss weight scaling applied to resolve dataset imbalance.

Click to enlarge 🔍

Test Predictions

Displays sample bounding box outputs on fresh, unseen panoramic X-rays during model validation.

Click to enlarge 🔍

Confusion Matrix

ResNet-50 classification accuracy breakdown across dental conditions, highlighting per-class precision and recall.

Click to enlarge 🔍

Dataset Distribution

Visual breakdown of sample counts across training, validation, and test splits used during ResNet-50 fine-tuning.

Click to enlarge 🔍

Dataset Class Balance

Pie chart of class proportions in the dataset, revealing imbalance characteristics addressed during ResNet-50 training.

Click to enlarge 🔍

K-Fold Accuracy

Accuracy progression across K-Fold cross-validation folds, demonstrating stable generalization of the ResNet-50 model.

Click to enlarge 🔍

Performance Bar Chart

Side-by-side metric comparison (accuracy, precision, recall, F1) per class, validating ResNet-50 diagnostic performance.

Click to enlarge 🔍

Performance Box Plot

Statistical variance and median performance across cross-validation runs, proving ResNet-50 metric consistency.

Click to enlarge 🔍

Training Performance Curves

Epoch-wise accuracy and loss convergence tracking for ResNet-50, visualizing learning rate effectiveness and overfitting prevention.

Click to enlarge 🔍

AI Dental Scan

Upload your dental X-ray and select a detection model

Select Detection Model

Faster R-CNN + ResNet-50

Two-stage detector with Feature Pyramid Network. Produces bounding boxes with precise localisation. Best for dental lesion detection.

⭐ Recommended · 87.9% Overall Accuracy

ResNet-50 Standalone

Single-stage image-level classifier. Faster inference. Classifies the whole image — no bounding box output.

⚡ Fast · Image-level Classification

🩻

Drop your X-ray here or click to browse

Supports panoramic and periapical dental X-rays (Multiple select enabled)

JPEGPNG DICOMTIFFBMP

🎯 Detection Results

Research Publications

📄 IEEE Report

Peer-reviewed research on DentOmni's AI diagnostic methodology, model architecture, and clinical validation.

🖥️

Best on Desktop

This IEEE report is formatted for full-page viewing. Please open DentOmni on a laptop or computer to read it properly.

💡 Open on Computer for Best Experience

⬇️ Download Report

Rendering Scientific Publication...

Detect
problems.
Protect
smiles.