POSTOPERATIVE WOUND. Choose "History of trauma: surgery" and enter the date of surgery, and the app stops guessing what the finding is and describes how it is healing - against the published Southampton wound scoring system (0 normal healing to V dehiscence) and the CDC/NHSN definition of surgical-site infection. From the photograph it measures the erythema of the surroundings, the crust along the wound line, a yellowish coating and sheen; discharge, wound edges and local warmth, swelling and tenderness come from three new examination questions. From colour alone the grade never exceeds II: pus and dehiscence are never claimed from a photograph. The timeline by day after surgery says what is expected (sutures day 10-14, crust within 3 weeks, infection window 30 days). This is an orientational grading for documentation, not a diagnosis of infection - the decision to open, swab and treat is clinical. Verified on a real postoperative wound on day 12; 283 regression tests.

BODY REGION IN THE CLINICAL CONTEXT. A photograph of one patch of skin does not say where on the body it lies, and location excludes whole groups of findings more reliably than colour: a ligature or restraint furrow away from the neck and limbs is impossible, abrasions are typical over joints, a Mongolian spot is lumbosacral. Enter the region (head, neck, trunk, arm, hand, leg, knee/elbow, foot) and the candidate list is reordered - with the reason shown whenever something is excluded. The app does not infer the region from the image; it asks. NO FALSE CERTAINTY. When even the best option stays below 50 %, the headline now reads "Uncertain finding - options only" and lists the two best, instead of announcing a "recognised pattern". A reddened elbow (erythema) had come out as a "band-like mark" at 30 % under a headline that sounded certain. Exclusions with their reasons are now always shown, not only when nothing else remains. SCALE FROM LIDAR (iPhone Pro). The Camera tab now reads the distance to the measured point from the LiDAR scanner and, with the lens calibration data, turns it into a scale: millimetres per pixel with no ruler and no tapping. "Take photo (LiDAR)" captures a real photograph, saves it to Photos with depth data and opens it in the Photograph tab with the scale set, so the lesion-diameter estimate is available at once; the source of the scale is written into the record. Not provided when zoomed in - a silently wrong scale is worse than none. This is time-of-flight measurement and a lens table, not image recognition. LESION DIMENSIONS AND UNITS. Length and width can be entered manually (ruler or forensic scale at the finding) - the only reliable size for an elongated wound or band. Choose millimetres, centimetres or inches in Info; the record and the CSV export carry the values. LARGER VIEW. The loupe can be doubled with one tap to inspect detail (e.g. pinpoint haemorrhages), and on tall iPhones the photograph gets more of the screen. Still no neural network, no trained model, no CoreML and no Vision - verified by 260+ regression tests.

CALIBRATION TARGET WITHOUT TAPPING. The T2/T3 calibration target from the aPrehliadky project is now found in the photograph automatically. With a ColorChecker Classic Mini in the T3 cut-out the app computes the full colour-correction matrix from its 24 patches; without it, or with the older T2 target, it balances white from the grey row. The fiducials and patches it read are drawn over the photograph, so you can see exactly what was measured. SCALE FROM THE TARGET. The fiducials have a known spacing (T3: 247 mm), so the millimetre scale is set at the same time - no need to tap two points on a ruler. PRINT THE TARGET FROM THE APP. Info > Calibration offers both targets as PDF (T3: A4 landscape with a cut-out for the ColorChecker Mini; T2: A4). The layout is the same definition the detector uses, so what you print is what the app finds. MORE HONEST PERCENTAGES. A colour deviation that only just crosses the "less yellow than skin" threshold, with no dark component, no longer yields a confident contusion - typically a narrow wound whose measured area is mostly surrounding skin. The assessment says so and points to the examination questions. The target is located by plain geometry - thresholds, connected components and a homography. QR codes are not decoded and no image-recognition framework is used; a regression test guards that the app's sources import neither Vision nor CoreML nor CoreImage. Verified by 255 regression tests.

EXPERT MODE — the app can now learn from your own verified cases. Switch it on in Info and any measurement can be stored as a reference fingerprint: twelve measured numbers and your note, never a photograph and nothing about the person examined. The fingerprint joins recognition immediately, and when several independent examples of the same finding agree, the conclusion is reported as better supported. Fingerprints can be exported and imported (.ch3lab) — a backup you can carry to another device. Recognition thresholds are now verified on real clinical photographs — not estimated. The verification series exposed seven faults you would otherwise have met in the field: a fresh livid bruise dropped out of recognition, a resorbing haematoma carried no evidence of blood, and solar lentigo was rejected outright — although melanin adds yellow while blood takes it away. Recognition now covers most of the atlas: petechiae, cyanosis, Mongolian spot, second-degree burn blisters, haemangioma, post-inflammatory pigmentation, pustules, scaly dermatoses, exanthems, putrefactive change, electrical marks and ligature furrows. Atlas expanded to 67 findings — a new "Benign skin growths" category: warts, molluscum contagiosum and skin tags, each with its forensic pitfalls. Forensic mode stays focused on injuries: a non-traumatic finding (mole, keratosis, wart) is still recognised, but capped at 60 % and referred to the dermatological mode. Livor mortis is never offered as green — that hue belongs to putrefaction. A new examination question: bleeding from the lesion — it excludes flat macules outright. Atlas search in both languages at once, value copying by long-pressing the colour swatch, haptic feedback, a clearer Info tab, and a new icon: the colour fog of the skin brought into focus by the diagnostic loupe. A welcome screen on first launch; the language follows the device — Slovak only on a Slovak device, English everywhere else. SHARING FINGERPRINTS WITH COLLEAGUES. A .ch3lab file now carries who measured each fingerprint and whether the measurement was chart-calibrated. Each import arrives as a named pack you can disable or delete as a whole, and received fingerprints stay in quarantine until you review them - nothing changes your conclusions behind your back. For each waiting fingerprint the app shows what will change once it is accepted: adds nothing, overlaps another finding, will have no effect (it measures like healthy skin), confirms the current conclusion, or shifts recognition. It never judges whether your label is correct - that does not follow from the measured numbers, and the app says so. Fixed Slovak localisation — some headings appeared in English while the app was set to Slovak, and a reference-photograph caption stayed Slovak in the English interface. On iPad the photograph now uses the full height of the display, and the assessment opens as a full page instead of a cramped panel. Verified by 228 regression tests.

Recognition thresholds are now verified on real clinical photographs — not estimated. The verification series exposed seven faults you would otherwise have met in the field: a fresh livid bruise dropped out of recognition, a resorbing haematoma carried no evidence of blood, and solar lentigo was rejected outright — although melanin adds yellow while blood takes it away. Recognition now covers most of the atlas: petechiae, cyanosis, Mongolian spot, second-degree burn blisters, haemangioma, post-inflammatory pigmentation, pustules, scaly dermatoses, exanthems, putrefactive change, electrical marks and ligature furrows. Atlas expanded to 67 findings — a new "Benign skin growths" category: warts, molluscum contagiosum and skin tags, each with its forensic pitfalls. Forensic mode stays focused on injuries: a non-traumatic finding (mole, keratosis, wart) is still recognised, but capped at 60 % and referred to the dermatological mode. A new examination question: bleeding from the lesion — it excludes flat macules outright. Reference fingerprints: measurements are compared against fingerprints taken from verified photographs, and every match shows its source. Fingerprints can be imported and exported (.ch3lab). Atlas search in both languages at once, value copying by long-pressing the colour swatch, haptic feedback, a clearer Info tab, and a new icon: the colour fog of the skin brought into focus by the diagnostic loupe. A welcome screen on first launch; the language follows the device — Slovak only on a Slovak device, English everywhere else. Fixed Slovak localisation — some headings appeared in English while the app was set to Slovak. Verified by 202 regression tests.

Colour chart calibration (24 patches) — a 3×3 correction matrix corrects the camera's own colour rendering, so series are comparable. Scale in millimetres — set it from two points on a ruler; the points can be dragged before you confirm. Atlas expanded to 63 findings — body modifications and traditional practices (tattoo, scarification, piercing, cupping, coining, moxibustion, Mongolian spot) and pattern findings from forensic practice. Erythema — the earliest change after blunt force; diascopy decides between it and a bruise. Tattoo recognition — a hue no skin chromophore can produce means exogenous pigment. Bruise dating corrected to the current evidence: colour tables assigning a number of days are not validated and have been removed. Dark mode — the interface no longer glares around the measured colour. CSV data export for scientific processing. Clinical context redesigned into short rows, with answers that now also establish a finding, not only exclude one. Many fixes to recognition, verified by hundreds of regression tests.