Instructions for Accurate Analysis
Follow our guide to ensure accurate results. Quality matters for reliable AI predictions.
Ivy AI Skin Illness Detection Module
Software Version: v1.0
Manufacturer: Ivy AI Solutions Limited, 71–75 Shelton Street, Covent Garden, London WC2H 9JQ, UK
Image Quality Requirements
Image quality is the single most important factor in screening accuracy. Follow these requirements to ensure reliable AI results.
Why image quality matters
ScanSkinAI uses advanced computer vision (DINOv2) to analyse pigment networks, vascular patterns, surface texture, border irregularity and colour variation — features that exist at the pixel level. Dermoscopy devices like MoleScope natively capture at 1920×1920 px, but a 500×500 px / 25–75 KB thumbnail export is a ~93% reduction that destroys the diagnostic detail. The AI will still produce a result from a low-quality image, but that result is unreliable and should not inform clinical decisions.
Minimum image specifications
| Parameter | Minimum required | Recommended |
|---|---|---|
| Resolution | 1024 × 768 px | 2048 × 1536 px or higher |
| File size | 150 KB | 500 KB – 5 MB |
| File format | JPEG or PNG | JPEG quality ≥80%, or PNG |
| Colour space | sRGB | sRGB |
| Max file size | — | 20 MB |
Resolution
Minimum
1024 × 768 px
Recommended
2048 × 1536 px or higher
File size
Minimum
150 KB
Recommended
500 KB – 5 MB
File format
Minimum
JPEG or PNG
Recommended
JPEG quality ≥80%, or PNG
Colour space
Minimum
sRGB
Recommended
sRGB
Max file size
Minimum
—
Recommended
20 MB
Auto-rejection: images below 1024×768 pixels or below 100 KB will be flagged as insufficient quality. Results from such images should be considered unreliable.
Images we cannot reliably predict
Wide-angle / distant photos
Lesion must fill ≥50% of frame — entire-limb shots lose pixel-level detail.
Surgical pen markings
Ink borders are misread as pathological features. Capture before marking.
Hair-obscured lesions
Hair covering >30% of the area hides texture, borders and colour patterns.
Thumbnails / screenshots
EMR / dermoscopy thumbnail exports (e.g. 500×500 from MoleScope) destroy fine structure.
Photos of screens or prints
Adds moiré, reflections and colour shifts — AI analyses the screen, not the lesion.
Blurred or out-of-focus
Motion or focus failure removes the pixel sharpness AI requires.
Overexposed / flash-washed
Specular highlights clip pixels to white — detail is permanently lost.
Clinical (macro) photography
- Hold the camera 10–30 cm from the lesion
- Lesion should fill at least 50% of the frame
- Include 1–2 cm of surrounding normal skin
- Use diffuse, even lighting — avoid direct flash
- Tap to focus on the lesion before capturing
- Keep the camera steady — brace your hand or surface
- Capture BEFORE any pen markings or surgical prep
Dermoscopic photography
Thumbnail trap: if your dermoscopy export is under 200 KB or 500×500 px, you're exporting a thumbnail — re-export as "Original" or "Full Resolution".
- Export the full-resolution original from your device
- If file is <200 KB, it is almost certainly a thumbnail — re-export
- If image is 500×500 px, it is a thumbnail (native is ~1920×1920)
- Do NOT resize, crop, rotate or filter before upload
- Clean the dermatoscope lens before each capture
- Maintain steady, even skin contact to avoid motion artefacts
- Save JPEG at 95–100% quality or PNG — never "optimise for web"
Scalp / hair-covered areas
- Part the hair fully to expose the entire lesion
- Use dermoscopy where available for scalp lesions
- If hair cannot be cleared, note that result may be unreliable
- Ensure the lesion — not surrounding hair — is in focus
Common mistakes → do this instead
Exporting thumbnails from EMR / PACS / dermoscopy software
Use "Export Original" or "Full Resolution"
Resizing images before upload
Upload the unmodified original at native resolution
Taking a screenshot of the image viewer
Download or export the actual file directly
Photographing a screen or printout
Transfer the digital file — do not re-photograph
Applying heavy JPEG compression
Save at quality ≥80%, or use PNG
Using "optimise for web" / "compress" exports
Always use full-quality export
Marking the skin before photographing
Capture image before pen markings
Photographing from too far away
Move closer — lesion fills ≥50% of the frame
Not focusing on the lesion
Tap to focus, wait for lock, then capture
Pre-upload checklist
Important limitations
ScanSkinAI is an AI-powered screening tool designed to assist clinical assessment. It is not a diagnostic device and does not replace histopathological examination, clinical judgment or specialist dermatological evaluation. Conditions that appear visually similar may require biopsy for definitive differentiation. If image quality requirements are not met, the screening result should be considered unreliable and should not inform clinical decisions.
Precision
Medical-Grade
Instant
Step-by-Step Guides
Select a scan type below for detailed instructions
Video Tutorial: Skin Analysis Guide
Watch this comprehensive guide to perform accurate skin analysis
Capture HD Photo
Use your smartphone to take a clear, high-definition photo of the affected skin area. Good lighting is essential for accurate AI analysis.
Pro Tips:
- Use natural daylight when possible
- Avoid flash — it can create glare
- Keep camera steady to prevent blur
- Ensure the area is clean and dry
Upload or Take Photo
Upload from your gallery or capture directly using the in-app camera.
Pro Tips:
- Supported formats: JPG, PNG
- Max file size: 10MB
- Multiple angles can help
- Save originals for your records
Select Location & Language
Pinpoint the body location using the interactive 3D model and choose your preferred output language.
Pro Tips:
- Use the 3D model for exact location
- Search for quick body part selection
- 13+ languages supported
- Accurate location improves results
Submit for AI Analysis
Confirm image quality, location, and language, then submit. We detect 31 skin conditions with ongoing development.
Pro Tips:
- Review all previous steps before submitting
- Ensure image meets quality standards
- Results available within seconds
- AI confidence score included
Get Personalised Care Plan
Complete your skin type assessment and risk profile to receive customised skincare recommendations.
Pro Tips:
- Answer skin type questionnaire
- Complete risk profile assessment
- Review personalised recommendations
- Save your customised care plan
How to Read Your AI Results
Understand AI prediction, confidence, probability ranking and the result summary in plain language.
1. AI Prediction
The single most likely condition our AI identified from your image — a screening signal, not a diagnosis.
- Trained on thousands of dermatology-grade photos.
- Best treated as a "second opinion" to discuss with a clinician.
2. Confidence Score
How certain the AI is about its top prediction, shown 0–100%.
3. Top Probability
After ranking every possible match internally, the AI shows you only the single highest-probability condition with its likelihood %.
Example only. The displayed result is the strongest match the AI is willing to surface.
4. Plain-Language Summary
Translates the AI output into actionable guidance you can act on today.
- What was detected and where on the body.
- Severity level (low / medium / high / critical).
- Next steps — self-care vs. seeing a dermatologist.
- Red-flag signals when urgent care is recommended.
Severity Badges Explained
ScanSkinAI is a wellness screening tool — not a medical diagnosis. A low confidence score, missing image area, or rare condition can affect accuracy. Always consult a qualified healthcare professional for diagnosis and treatment.
Fitzpatrick Scale
Skin colour & UV response classification
Very Fair
Fair
Light
Moderate
Dark
Very Dark
31 Detectable Conditions
Comprehensive range of skin conditions with medical-grade precision
Skin Cancers
Pre-cancerous
Infectious Diseases
Inflammatory Conditions
Genetic Disorders
Benign Lesions
Image Quality is Critical
Blurry, poorly lit, or unclear images will impact accuracy.
Independently Validated
Our AI was tested by an independent dermatologist reviewing 100 real user scans
Errs on Caution
Only 2% under-triage — the AI would rather send you to a doctor than miss something
Real Photos
Tested on actual user photos taken on phones — not perfect lab images
Independent
Reviewed by a dermatologist with 25+ years experience and no ties to us
ScanSkinAI is a wellness screening tool — not a medical diagnosis. Always consult a healthcare professional for medical advice. View full evidence base →
Ready to Analyse?
Follow the guide above for accurate results