Complete reference for all Tricholens tools, methods, and scoring systems
Tricholens is a screening tool, not a diagnostic service. Results are estimates intended to help you track trends over time and identify areas worth discussing with a dermatologist. They should not be used as a substitute for professional medical evaluation.
Tricholens provides three standalone scalp tools plus an automated detection pipeline:
The automated detection pipeline (sections 2–7) uses computer vision to count, classify, and analyse hairs from a single trichoscopy image.
Users upload trichoscopy or close-up scalp images. If the physical dimensions of the image (in millimeters) are known, they can be provided for more accurate density calculations.
Default assumption: When the user does not provide physical dimensions, Tricholens assumes a field of view of 2.5 mm × 2.5 mm for a square image. For non-square images, the dimensions are scaled proportionally based on the aspect ratio.
Tricholens uses a computer vision object detection model trained on trichoscopy images to identify individual hair follicles. The model returns bounding boxes and segmentation polygons for each detected hair.
Each detected hair's cross-sectional width is computed directly from its segmentation polygon. The algorithm measures the polygon's narrow-axis span (perpendicular to the hair's length direction) and converts that pixel distance to millimeters using the calibrated field of view.
Hairs are classified by width in micrometers:
| Type | Width | Meaning |
|---|---|---|
| Terminal | ≥ 60 µm | Thick, healthy, pigmented scalp hair |
| Intermediate | 30–60 µm | Transitioning hair, a marker of miniaturization |
| Vellus | < 30 µm | Fine, thin hair associated with hair loss |
Hair thickness varies naturally between ethnic groups. Tricholens adjusts classification thresholds to reflect published ethnic baselines.
| Tier | Asian (80-120 µm) | Caucasian (50-100 µm) | African (40-90 µm) |
|---|---|---|---|
| Healthy Terminal | ≥ 80 µm | ≥ 65 µm | ≥ 55 µm |
| Normal | ≥ 60 µm | ≥ 50 µm | ≥ 40 µm |
| Thinning | ≥ 30 µm | ≥ 30 µm | ≥ 30 µm |
| Miniaturized | < 30 µm | < 30 µm | < 30 µm |
Every measurement is checked against physical bounds that reflect biological limits:
| Bound | Asian | Caucasian | African |
|---|---|---|---|
| Upper (max plausible) | 101 µm | 101 µm | 101 µm |
| Lower (visibility floor) | 10 µm (universal) | ||
In magnification mode, the tool calculates a suggested correction using thin-hair-first scaling (anchoring to the thinnest measurement to keep it above 10 µm). In manual FOV mode, the tool warns the user to re-check their field-of-view dimensions but cannot suggest a specific correction.
The T:V ratiois calculated as Terminal Hair Count ÷ Vellus Hair Count. A healthy scalp typically has a T:V ratio above 4.0.
When fewer than 3 terminal hairs are detected, Tricholens uses two alternative metrics:
HMI = ((terminal × 1.0 + intermediate × 0.5) ÷ total) × 100.A composite score out of 100 combining:
Score ranges: 70+ normal, 40–69 mild concern, below 40 moderate concern. Clamped between 10 and 100.
Density = Total Hairs Detected ÷ Image Area (cm²)Normal density ranges from approximately 120 to 200 hairs/cm² depending on the scalp zone, ethnicity, and age.
After detection, the image and all computed metrics are sent to a large language model which generates a screening summary. The AI acts as a screening assistant, references specific metrics, notes when fallback metrics are used, and always recommends consulting a dermatologist.
The Hair Thickness tool allows users to manually measure individual hair widths on trichoscopy or close-up scalp images with full control over calibration and measurement placement.
Two independent calibration modes:
pixelsPerMm = (magnification / 50) × 512. Auto-calibration can suggest a corrected value.pixelsPerMm = imageWidth / fovWidthMm. Auto-calibration warns but cannot suggest corrections.EXIF metadata is extracted on upload (digital zoom, focal length, device info) to help verify calibration.
Click two points across a hair shaft. Each measurement gets labeled start/end anchors (e.g., “1S”, “1E”). Start with the thinnest visible hair first for best auto-calibration results.
Multiple measurements along the same hair can be grouped into segments, rendered as filled polygons on the canvas.
Each measurement is classified in real-time using ethnic-adjusted thresholds. Users can:
A resizable chat widget docked below the canvas provides:
The Spot Analysis tool detects bright, reflective, or pigmented regions on the scalp — such as sebum buildup, oily patches, or flaking — using client-side image processing.
The tool processes the uploaded image entirely in the browser (no server round-trip):
0.299R + 0.587G + 0.114B| Parameter | Baseline Value | Description |
|---|---|---|
| Brightness threshold | 180 | Luminance value (0–255) above which a pixel is flagged. Lower = more sensitive. |
| Minimum spot size | 8 px | Connected blobs smaller than this are discarded as noise. |
These defaults work well for most trichoscopy images. Users can adjust both parameters via the settings panel, with a one-click reset to baseline available when values have been changed.
The spot coverage percentage is rated on a three-tier scale:
| Coverage | Rating |
|---|---|
| ≤ 33% | Good |
| 34–60% | Moderate |
| > 60% | Significant |
Detected spots are rendered as a pink/magenta overlay on a dimmed version of the original image. Users can toggle between overlay and original views. The image supports full zoom (scroll or buttons, 0.5x–20x) and pan (click and drag) for detailed inspection of individual spots.
All three tools (Hair Thickness, Spot Analysis, Self Assessment) include a crop tool that lets users draw a rectangle on the image to isolate a region of interest before analysis. This is useful for focusing on a specific area of the scalp or removing unwanted background. The cropped image replaces the original and can be downloaded.
The Self Assessment tool provides a structured way for users to log observed scalp conditions, rate their severity, and track changes over time.
Users tag conditions they observe in their scalp image from the following categories:
| Condition | Description |
|---|---|
| Oily Scalp | Excess sebum around follicles, shiny or greasy appearance |
| Dry Scalp | Flaky, tight skin with visible dryness and possible irritation |
| Dandruff | White or yellowish flakes attached to the scalp or hair base |
| Sensitive / Redness | Visible redness, irritation, or broken capillaries |
| Seborrheic Dermatitis | Oily, scaly patches with yellowish crusting around follicles |
| Inflammation | Red, swollen areas around follicles, possible pustules |
| Hair Loss Pattern | Visible thinning, widened parting, or exposed scalp skin |
Each tagged condition is rated on a 5-point severity scale:
| Level | Label | Penalty Weight |
|---|---|---|
| 1 | Minimal | 0 (no impact on health score) |
| 2 | Mild | 1 |
| 3 | Moderate | 2 |
| 4 | Significant | 3 |
| 5 | Severe | 4 |
The health score is a 0–100 rating computed from all tagged conditions and their severity levels. Level 1 (Minimal) carries zero penalty — tagging a condition as minimal means it's not really a concern. Penalties only accumulate from Level 2 onward.
penalty = Σ(severityLevel - 1) for each tagged condition
maxPenalty = 7 categories × 4 max penalty = 28
healthScore = 100 - (penalty / maxPenalty) × 100Score interpretation:
| Score | Label | Example |
|---|---|---|
| 80–100 | Healthy | All conditions minimal, or no conditions tagged |
| 60–79 | Mild Concerns | A few conditions at mild severity |
| 40–59 | Needs Attention | Multiple moderate conditions |
| 20–39 | Significant Concerns | Several significant conditions |
| 0–19 | Consult a Specialist | Widespread severe conditions |
Signed-in users can save assessments to their account (Firebase). Each saved report stores all tagged conditions, severity levels, the computed health score, and any notes. Past assessments are accessible from the History view, allowing users to track their scalp health over time. Reports can be individually deleted.
The uploaded scalp image is displayed alongside the assessment panel with full zoom (scroll or buttons, 0.5x–20x) and pan (click and drag) support for detailed visual comparison while tagging conditions.
For best results: Use the same device and magnification for each session, capture images in consistent lighting, and provide the physical dimensions (mm) of your field of view if known.