Inside Point-n-Capture: How AI Detects Wound Borders and Calculates Area
product
A technical look at how Point-n-Capture segments wound borders, derives scale from a smartphone, and turns a photo into structured length, width, and area.
Point-n-Capture is often described in three words — point, capture, measure — but the value sits in what happens between the shutter and the structured note. This deep dive walks through the mechanics.
Border segmentation
When the image is captured, a segmentation model identifies the wound edge against the surrounding skin. The output is a pixel-level mask, not a bounding box, so irregular shapes — Charcot ulcers, dehisced surgical sites, pressure injuries with satellite areas — are traced along their true perimeter rather than approximated as a rectangle.
Scale without a fiducial
Rather than requiring a sticker or ruler in the frame, Point-n-Capture uses depth data from the smartphone camera to establish real-world scale. This is what lets clinicians skip the fiducial step and still get length, width, and area in centimeters rather than pixels.
From mask to measurement
Once the mask and scale are in place, the system derives:
- Length and width along the wound's principal axes.
- Area as the true masked region, not L × W.
- A draft wound description covering location, size, and visible tissue types for the clinician to review.
Where the measurement goes
A measurement disconnected from the chart is a photo in a folder. Point-n-Capture writes into the same record that the AI scribe for wound care is building, so imaging, narrative, and structured fields land together. Over multiple visits, the captures feed the healing analytics dashboard so trajectory — not just today's number — drives decisions.
Review and override
The clinician sees the detected border overlaid on the image and can adjust it before confirming. Ambiguous edges should be reviewed rather than accepted. The full workflow is at WoundScribe AI.