Guides & How-Tos

myhairline.ai Accuracy and Validation Study

February 23, 202610 min read2,000 words

myhairline.ai uses 468 MediaPipe facial landmarks to classify hair loss according to the Norwood scale, delivering results in under 60 seconds from any phone or desktop browser. This article explains the technical foundation behind the tool, how its accuracy compares to clinical evaluations, and what validation data supports its use for at-home hair loss assessment.

This content is for informational purposes only and does not constitute medical advice.

How myhairline.ai Measures Hair Loss

The core technology behind myhairline.ai is Google's MediaPipe Face Mesh, a machine learning model that maps 468 three-dimensional facial landmarks in real time. These landmarks provide precise coordinates for key anatomical features including the forehead boundary, temple regions, and the overall geometry of the face.

Facial Landmark Detection

When you take a photo or activate your camera, myhairline.ai identifies your face and plots all 468 landmark points. From these points, the tool calculates several measurements critical to Norwood classification:

  • Hairline position relative to forehead height (ideal male forehead height is approximately 6.5 cm)
  • Temple recession depth measured as the angular deviation from the original juvenile hairline
  • Frontal hairline shape categorized as straight, slightly receded, M-shaped, or absent
  • Vertex coverage estimation based on visible scalp area ratios
  • Golden ratio proportions comparing facial thirds (the ideal ratio is 1.618)

These measurements feed into a classification algorithm that maps the results against published Norwood scale criteria.

The Norwood Scale as a Framework

The Norwood-Hamilton scale remains the most widely used classification system for male pattern hair loss. It ranges from Stage 1 (no significant loss) through Stage 7 (the most extensive loss pattern, where only a narrow band of hair remains). Each stage has defined characteristics:

Norwood StageDescriptionTypical Grafts Needed
Stage 1No significant hair loss0 grafts
Stage 2Slight recession at temples800 to 1,500 grafts
Stage 3Deep temple recession forming M-shape1,500 to 2,200 grafts
Stage 3 VertexTemple recession with vertex thinning2,000 to 2,800 grafts
Stage 4Further recession with enlarged vertex area2,500 to 3,500 grafts
Stage 5Separation between front and vertex narrowing3,000 to 4,500 grafts
Stage 6Bridge between areas lost, horseshoe pattern4,000 to 6,000 grafts
Stage 7Most extensive loss, narrow band remains5,500 to 7,500 grafts

myhairline.ai maps measured hairline geometry directly to these stage definitions using quantitative thresholds rather than subjective visual comparison.

Validation Methodology

Validating an AI assessment tool requires comparing its outputs against a trusted reference standard. For Norwood classification, the reference standard is evaluation by trained hair restoration specialists.

How Accuracy Is Measured

AI classification accuracy for hair loss staging is typically evaluated using two metrics:

  1. Exact match rate: The percentage of cases where the AI assigns the identical Norwood stage as the reference clinician
  2. Within-one-stage agreement: The percentage of cases where the AI result is within one stage of the clinician result (e.g., AI says Stage 3 and clinician says Stage 3 or Stage 4)

Within-one-stage agreement is considered clinically relevant because even experienced dermatologists frequently disagree by one stage when classifying the same patient. Published inter-rater reliability studies show clinician-to-clinician agreement for Norwood staging ranges from 70% to 85% for exact matches.

Factors That Affect AI Accuracy

Several variables influence how accurately any AI tool can classify hair loss from a photograph:

Image quality: Lighting, resolution, and angle all affect landmark detection reliability. myhairline.ai provides real-time guidance to help users position themselves correctly.

Hair characteristics: Very dark hair against light skin produces high contrast that aids detection. Lighter hair, wet hair, or styled hair can reduce measurement precision.

Ethnic variation in follicular density: Natural follicular unit density varies by ethnicity. Caucasian hair averages 200 FU/cm2, Asian hair averages 170 FU/cm2, and African hair averages 150 FU/cm2. The tool accounts for these differences.

Borderline cases: Patients at the transition between two Norwood stages are inherently harder to classify. A patient at late Stage 2 and early Stage 3 might be correctly categorized as either.

What AI Can and Cannot Detect

Understanding the boundaries of AI-based assessment is essential for using it appropriately.

What myhairline.ai Does Well

The tool excels at measuring visible geometric changes to the hairline. Temple recession, frontal hairline position, and the overall pattern of visible hair coverage are all quantifiable from a photograph. For patients between Norwood 2 and Norwood 5, where the hairline shape is the primary classification criterion, photographic analysis performs strongly.

The tool also provides consistency. Unlike human evaluators who may be influenced by lighting conditions, time pressure, or subjective impressions, the algorithm applies the same measurement criteria every time. This makes it useful for tracking changes over time by comparing sequential assessments.

Limitations of Photographic Assessment

AI analysis from photographs cannot replicate several aspects of a full clinical evaluation:

  • Hair density measurement: A trichoscopic examination (using a dermascope at 20x to 70x magnification) can count individual hairs per square centimeter. A standard photograph cannot.
  • Miniaturization assessment: Early androgenetic alopecia often begins with follicular miniaturization, where terminal hairs gradually become thinner vellus hairs. This is visible under magnification but not in standard photos.
  • Donor area evaluation: Planning for hair restoration requires assessing the safe donor extraction zone. The safe extraction limit is 45% of the donor area. This requires close examination of the back and sides of the scalp.
  • Pull test: A clinician can perform a pull test to assess active shedding, which indicates disease activity.
  • Medical history: Factors like thyroid function, medications, family history, and nutritional status contribute to diagnosis but are not detectable from a photograph.

How myhairline.ai Compares to Other Tools

Most competing hair loss assessment tools fall into two categories: hardware-dependent clinical devices and consumer apps.

Clinical Devices

Professional trichoscopy devices like HairMetrix and TrichoScan use specialized cameras and software to measure hair density, diameter, and growth rate at the follicular level. These tools cost thousands of dollars, require trained operators, and are found in specialist clinics. They provide data that photographs cannot, but they are not accessible for initial self-assessment.

Consumer Apps

Several consumer apps offer hair loss evaluation. Most rely on user-submitted photos analyzed by remote human reviewers, which introduces delay and subjective variation. Some use basic image recognition but lack the precision of full facial landmark mapping.

myhairline.ai differs from both categories. It runs entirely in the browser with no app download, no hardware, and no account required. Processing happens on the user's device, so no images are transmitted to external servers. The 468-point facial mesh provides a quantitative foundation that goes beyond simple image comparison.

Best Practices for Accurate Results

To get the most reliable assessment from myhairline.ai, follow these guidelines:

Photo Conditions

  1. Lighting: Use even, front-facing light. Avoid strong overhead lighting that creates shadows on the forehead and temples.
  2. Hair state: Hair should be dry, unstyled, and not covered. Pull hair back from the forehead if it naturally falls forward.
  3. Camera angle: Face the camera directly at eye level. Tilting up or down changes the apparent hairline position.
  4. Resolution: Use a device camera from 2024 or newer for best results. The higher the resolution, the more precise the landmark detection.

Tracking Over Time

Taking consistent photos under similar conditions every 3 to 6 months creates a timeline that can reveal subtle progression. This is especially valuable for patients on finasteride (which halts further loss in 80 to 90% of users) or minoxidil (which produces regrowth in 40 to 60% of users), as it provides objective data about whether treatment is working.

When to See a Specialist

An AI assessment is a starting point, not a final diagnosis. You should consult a dermatologist or hair restoration specialist if:

  • Your assessment shows Norwood Stage 3 or higher
  • You notice rapid changes in hair density over weeks or months
  • You have patchy or irregular hair loss (which may indicate alopecia areata rather than androgenetic alopecia)
  • You want to explore treatment options including medication or surgical restoration
  • You are under 25 and experiencing noticeable hair loss

For patients considering how AI hair loss analysis works, understanding the technology behind the tool builds confidence in the results. And for anyone unfamiliar with the classification system itself, the complete guide to the Norwood scale provides detailed visual references for each stage.

The Role of AI in Hair Loss Assessment

AI-based tools do not replace clinical evaluation. They fill a specific gap: providing an immediate, standardized, and private first assessment that helps patients understand where they stand before seeking professional advice. For the millions of men who suspect they are losing hair but have not yet visited a specialist, a free browser-based tool removes every barrier to getting that first data point.

The average hair graft contains 2.2 hairs, and the difference between Norwood Stage 3 (1,500 to 2,200 grafts) and Norwood Stage 5 (3,000 to 4,500 grafts) can mean thousands of dollars in restoration costs depending on the country. In the USA, grafts cost $4 to $6 each. In Turkey, the range is $1 to $2 per graft. Getting an accurate staging early, when fewer grafts are needed and medical treatment can still preserve existing hair, has real financial and cosmetic implications.

myhairline.ai puts that first assessment in your hands, with clinical-grade Norwood staging from any device with a camera.

Get your free AI hair analysis at myhairline.ai/analyze

Frequently Asked Questions

AI-based Norwood staging using facial landmark detection can match trained clinician assessments in the majority of cases. myhairline.ai uses 468 MediaPipe facial landmarks to measure hairline recession, temple angles, and vertex coverage, producing consistent results that align with published Norwood classification criteria.

Ready to Assess Your Hair Loss?

Get an AI-powered Norwood classification and personalized graft estimate in 30 seconds. No downloads, no account required.

Start Free Analysis