"Can AI read a skin test?" is one of the most common questions allergists ask about AI, and the answers online range from hype to dismissal. This page sets out what reading a skin test involves, what the published research on automated and AI-assisted reading shows for skin prick, intradermal and patch tests, what still needs a person, and how Medora handles it today.
What "reading" a skin test actually involves
Per the 2008 AAAAI/ACAAI diagnostic testing parameter:[1]
- Skin prick tests peak at 15 to 20 minutes, and intradermal tests are read at 10 to 15 minutes. Both wheal and erythema (flare) are recorded in millimeters and compared with the positive (histamine) and negative (saline or diluent) controls.
- Qualitative 0 to 4+ grades are not recommended; millimeter measurements are.
- Size can be reported as the mean of the longest diameter and the longest diameter perpendicular to it, by planimetry, or relative to histamine.
- For skin prick tests, a response at least 3 mm larger than the negative control, with equivalent erythema, is the parameter's evidence of allergen-specific IgE. For drug skin tests, the 2022 drug allergy parameter requires a wheal at least 3 mm larger than the negative control plus a flare of at least 5 mm.[2]
Patch tests are different: they are read at about 48 hours and again between days 3 and 7, and graded on the descriptive ICDRG scale from negative through doubtful, weak, strong and extreme positive, plus irritant.[3]
So a skin test "reading" combines several tasks: find each site, identify which allergen it is, measure the wheal (and flare), compare with the controls, apply a criterion, and then interpret the result against the history. AI research has made progress on the first three. The last one belongs to the allergist.
Why measurement is hard to standardize
The 2008 parameter is candid about variability. No single prick device has a clear advantage, and wheal size varies significantly from one device to another. Reliability depends on the tester's skill, the instrument, skin color, skin reactivity that day, age, and the potency of the extracts. Intradermal test reproducibility has been reported as poor.[1] Histamine wheals are significantly larger in darkly pigmented skin.[1] Any automated method has to cope with all of this, and has to be evaluated across skin tones.
Skin prick tests: what published research shows
The most rigorous published work pairs a standardized prick device with automated reading. In a 2025 Nature Communications study, an AI-assisted readout was trained on 7,812 manually labeled wheals from 651 patients and validated against physician measurement on 2,604 wheals from 217 patients. In a separate test cohort, physicians adjusted 5.8% of the AI measurements, and the AI-assisted readout reduced inter- and intra-observer variability and readout time. Several authors are employees or shareholders of the device maker.[4]
Two features of that study are worth noticing. The AI worked inside a controlled device workflow, not on arbitrary phone photos. And it was designed to support physicians, who still reviewed and adjusted the measurements.
Intradermal tests: earlier and harder
Intradermal testing is harder to automate: blebs, smaller differences, and poorer reproducibility to begin with. A company-sponsored study of a 3D-laser reader performed intradermal tests in 194 participants using histamine and saline controls as stand-ins for true positives and negatives. Agreement between the device and manual measurement was moderate to low, with intraclass correlations from 0.40 to 0.65.[5] That is useful progress, and a reminder that intradermal reading is not a solved problem.
Patch tests: grading from images is the hard part
Patch testing seems a natural fit for photos, because readings happen days apart. The difficulty is the grading. In a study where five experienced dermatologists graded patch-test photographs, their ICDRG grades differed widely; they agreed only on a simplified negative, positive or irritant reading.[6] A 2025 review of AI in patch testing found 10 qualifying studies, mostly convolutional neural networks, and identified small samples, variable image capture and a lack of standardized reporting on skin types as the key limitations.[7]
What still needs a person
- Interpretation. A positive test shows sensitization, not allergy. The NIAID guidelines say the skin prick test alone cannot be considered diagnostic of food allergy.[8]
- Negatives. Recording "no reaction" is a clinical statement. A missed wheal recorded as negative is worse than a wheal flagged for review.
- Flare and controls. Erythema matters in several criteria[1][2] and is harder to see on some skin; in Medora the nurse records it.
- Irritant versus allergic. Especially in patch testing, where pustular and "angry back" reactions can mislead.[3]
- Sign-off. The clinician signs the result.
How Medora handles skin testing today
Medora Vision: investigational, not FDA-cleared
Medora Vision is clinician-assisted documentation for skin prick and intradermal testing. It is investigational and in development, and it is not FDA-cleared. A person confirms every result, and the allergist signs.
- The nurse photographs the skin prick or intradermal test panel. The histamine and saline controls come first, and the draft asks the nurse for them first.
- Medora finds each test site, named by the number written beside it, and drafts each wheal size from the photo, using the printed ruler card for scale.
- When it can't confirm the scale, or isn't sure about a site, it flags that site for the nurse instead of guessing. It also says when a site wasn't photographed.
- The nurse checks the values Medora flags, corrects the draft and confirms the rest. A person confirms every value before it is recorded, and nothing is recorded as "no reaction" until a person confirms it. The nurse records flare.
- The practice's own grading rules classify the results. The review screen shows how each result was made (drafted by Medora, confirmed by the nurse, signed by the clinician), with an audit trail. The clinician signs.
Patch testing is live in Medora as a workflow: ordering, 48-hour and Day 3–4 readings scored on the ICDRG scale, photos, and clinician sign-off. Photo-assisted reading of patch tests is in development.
Because all Medora modules share one patient context, the clinician sees the wheal sizes, controls and grades the nurse recorded alongside the drafted note. See the skin testing workflow.
In a 60-day pilot at Allergy Affiliates, provider sign-off per skin-test visit went from 18 minutes to 4. For the clinical details of each test, see our allergy skin testing guide; for the bigger picture, see AI in allergy and immunology.
Questions to ask any skin testing AI vendor
- What is the regulatory status of the image feature, in writing?
- Does it ever record a negative without a person confirming it?
- How does it get scale: a reference object, a device, or an assumption?
- What does it do when it isn't sure: guess, or flag?
- Has it been evaluated across skin tones, and on intradermal as well as prick tests?
- Does it apply my practice's criteria, and does it measure flare or leave it to staff?
Frequently asked questions
Can AI read allergy skin tests?
Published research devices can measure wheals with AI support, mostly inside controlled device workflows and with physicians reviewing the results. Interpretation still needs the allergist. Medora Vision is investigational and not FDA-cleared: it drafts wheal sizes for the nurse to confirm, and the clinician signs.
Does Medora read or interpret skin tests?
No. Medora drafts wheal sizes from the photo using the printed ruler card for scale and flags sites it isn't sure of. The nurse confirms every value, the practice's grading rules classify the results, and the clinician signs.
Is Medora Vision FDA-cleared?
No. Medora Vision is investigational and in development, and it is not FDA-cleared.
Does Medora measure flare?
No. The nurse records flare.
Does Medora support patch testing?
Yes, as a workflow: ordering, 48-hour and Day 3-4 readings scored on the ICDRG scale, photos and clinician sign-off. Photo-assisted reading of patch tests is in development.
Why can't AI just record negatives automatically?
Because a missed reaction recorded as negative is a clinical error that nobody would look at again. In Medora, nothing is recorded as no reaction until a person confirms it.
Related guides
Sources
Last reviewed October 8, 2026. Summaries paraphrase the cited documents; read the originals before making clinical decisions.
- Bernstein IL, Li JT, Bernstein DI, et al. Allergy diagnostic testing: an updated practice parameter. Ann Allergy Asthma Immunol. 2008;100(3 Suppl 3):S1–S148. www.aaaai.org
- Khan DA, Banerji A, Blumenthal KG, et al. Drug allergy: a 2022 practice parameter update. J Allergy Clin Immunol. 2022;150(6):1333–1393. www.aaaai.org
- Fonacier L, Bernstein DI, Pacheco K, et al. Contact dermatitis: a practice parameter—update 2015. J Allergy Clin Immunol Pract. 2015;3(3 Suppl):S1–S39. www.aaaai.org
- Seys SF, Hox V, Chaker AM, et al. Artificial intelligence (AI)-assisted readout method for the evaluation of skin prick automated test results. Nat Commun. 2025;16:8637. (Several authors are employees or shareholders of the device maker.) doi.org
- Morales-Palacios MP, Núñez-Córdoba JM, Tejero E, et al. Evaluation of a novel automated allergy intradermal skin test reader: a diagnostic accuracy study. Clin Exp Allergy. 2024;54(12):1006–1009. (Company-sponsored.) doi.org
- Ivens U, Serup J, O’goshi K. Allergy patch test reading from photographic images: disagreement on ICDRG grading but agreement on simplified tripartite reading. Skin Res Technol. 2007;13(1):110–113. pubmed.ncbi.nlm.nih.gov
- Tang HS, Ebriani J, Yan MJ, Wongvibulsin S, Farshchian M. Artificial intelligence in patch testing: comprehensive review of current applications and future prospects in dermatology. JMIR Dermatol. 2025;8:e67154. doi.org
- NIAID-Sponsored Expert Panel (Boyce JA, et al.). Guidelines for the diagnosis and management of food allergy in the United States. J Allergy Clin Immunol. 2010;126(6 Suppl):S1–S58. pmc.ncbi.nlm.nih.gov