AI-Powered Dermatology

Take Control of Your Skin Health

Instant AI screening for 8 skin lesion types using deep learning. Upload a dermatoscopic image and receive a detailed risk assessment in seconds.

Start Screening Learn More

88.2% Accuracy

0.9757 AUC Score

25K+ Images Trained

8-Class Classification

Differentiates between Melanoma, BCC, SCC, Actinic Keratosis, Benign Keratosis, Dermatofibroma, Vascular Lesion, and Melanocytic Nevus.

Clinical Metadata Fusion

Goes beyond image analysis by incorporating patient age, sex, and anatomical site — the same factors dermatologists consider in clinical practice.

Confidence-Calibrated

Provides per-class probability distributions and malignancy risk scores. Uncertain cases are flagged for professional review rather than giving false confidence.

AI Screening

Upload Your Skin Lesion Image

Our EfficientNet-B4 model analyzes dermatoscopic images at 456×456 resolution and provides a differential diagnosis across 8 categories.

Enter your details

Age and sex help the model make a more accurate assessment.

Upload a clear image

Use a well-lit, close-up photo of the affected skin area.

Get instant results

Receive classification, confidence scores, and recommended next steps.

This is a screening tool only. It does not replace professional medical advice. Always consult a dermatologist for proper diagnosis.

Skin Checkup

Fill in your details and upload a photo

Age

Gender

Upload Image of Affected Area

Drag & drop or click to upload

Supported: JPG, PNG, WEBP

Understanding the Risk

What You Should Know About Skin Cancer

Melanoma

The deadliest form of skin cancer. Develops in melanocytes and can spread rapidly to other organs. Survival rate drops from 99% (early stage) to 30% (late stage).

High Risk

Basal Cell Carcinoma

The most common skin cancer. Typically grows slowly and rarely metastasizes, but can cause significant local tissue damage if untreated.

Moderate Risk

Squamous Cell Carcinoma

Second most common skin cancer. Can metastasize to lymph nodes and distant organs. Often found on sun-exposed areas like head, neck, and hands.

High Risk

Actinic Keratosis

Pre-cancerous rough patches caused by UV damage. Approximately 10% may progress to squamous cell carcinoma if left untreated.

Pre-cancerous

Analysis Results

Architecture

EfficientNet-B4 ImageNet Pretrained 456×456 Input Multi-Modal Fusion Focal Loss (γ=2.0) TTA (8 transforms)

Dual-input architecture fusing EfficientNet-B4 image features (1792-dim) with clinical metadata (9-dim: age, sex, anatomical site) via a Dense fusion head. Trained with 2-phase transfer learning.

Overall Test Set Metrics

88.2% Test Accuracy

0.9757 AUC (macro)

0.8365 F1 (macro)

0.8806 F1 (weighted)

                            91.9%
                            Binary Accuracy
                        

                            0.9709
                            Binary AUC
                        

Per-Class Performance

Class	Precision	Recall	F1-Score	AUC	Support
MEL Melanoma	0.844	0.770	0.805	0.960	679
NV Melanocytic Nevus	0.915	0.946	0.930	0.977	1,931
BCC Basal Cell Carcinoma	0.910	0.916	0.913	0.993	499
AK Actinic Keratosis	0.754	0.685	0.718	0.954	130
BKL Benign Keratosis	0.784	0.847	0.814	0.978	393
DF Dermatofibroma	0.936	0.806	0.866	0.988	36
VASC Vascular Lesion	0.944	0.895	0.919	0.987	38
SCC Squamous Cell Carcinoma	0.845	0.638	0.727	0.941	94
Macro Avg	0.866	0.813	0.837	0.976	3,800

Analyzing Image

Running inference through EfficientNet-B4...