Skip to main content
Tessa NewsLink
Paste a health news link, or browse

Source study found

Story checked

Detecting skin cancer even before it becomes visible with AI-assisted image recognition (opens in a new tab)

medicalxpress.com · 2026-09-22

Short answerEvidenceSource

Short answer

Not supported

Not supported.

One claim goes further than the study. 4 other points were not covered by the paper.

  • 1 overstated
  • 4 not covered

Checked against the study summary. The full text wasn't available, so some details couldn't be settled either way.

Share this check

Follow the evidence trail
1
2

NewsLink checks it

Not supported

One claim overstates the study. Four claims the study doesn't address.

  • 1 overstated
  • 4 not covered
Open claim evidence
3
Then inspect each claim

Evidence layer

Claim by claim

Each claim gets a verdict. Expand it to see the evidence directly below.

5 claims in this story

Showing all 5 claimsChoose a verdict to focus the list.

Claim 1 of 5OverstatedIn a feasibility study published in JAMA Dermatology, the team systematically screened the faces of patients at high risk of developing skin cancer and detected basal cell carcinoma before any visible skin changes developed.View evidenceHide evidence

As statedfor the first time

Why this verdict

The paper profile supports a prospective cross-sectional feasibility study in high-risk inpatients, with macroscopically suggestive lesions excluded and inconspicuous facial skin screened using AI-assisted LC-OCT; subclinical BCCs were identified and some were histologically confirmed. However, the story’s “for the first time” novelty claim is not supported by the abstract-level profile, and the wording presents a broad success claim without the paper’s diagnostic-performance context, including PPV, false positive/nonbiopsied lesions, and absence of sensitivity estimates.

1

Study evidence

AI-assisted LC-OCT identified 18 lesions as BCC; 15 of these were histologically confirmed, giving a PPV of 83.3% (95% CI, 58.1%–96.4%).PPV 83.3% (95% CI, 58.1%–96.4%)

“This prospective cross-sectional feasibility study was conducted from April 2025 to October 2025”
Claim 2 of 5Not coveredResearchers at Friedrich-Alexander-Universität Erlangen-Nürnberg developed a new AI-assisted approach using LC-OCT.View evidenceHide evidence

Why this verdict

The abstract-level profile supports that the paper evaluated an AI-assisted LC-OCT approach for systematic screening of inconspicuous facial skin in high-risk patients. However, it does not fully verify the story’s framing that FAU researchers specifically “developed” a new approach, as opposed to assessing feasibility and diagnostic performance of an AI-assisted LC-OCT screening protocol.

1

Study evidence

AI-assisted LC-OCT identified 18 lesions as BCC; 15 of these were histologically confirmed, giving a PPV of 83.3% (95% CI, 58.1%–96.4%).PPV 83.3% (95% CI, 58.1%–96.4%)

“This prospective cross-sectional feasibility study was conducted from April 2025 to October 2025”
Claim 3 of 5Not coveredDr. Moritz Ronicke said that detecting basal cell carcinoma early allows for less invasive treatment options, and that in some cases a cream may be sufficient.View evidenceHide evidence

As statedless invasive treatment options

Why this verdict

The profile’s introduction states that early BCC diagnosis permits less invasive treatment, which supports the general thrust of the claim. The more specific assertion that in some cases a cream may be sufficient is not present in the abstract-level paper profile, so the full claim cannot be verified at this depth.

1

Study evidence

AI-assisted LC-OCT identified 18 lesions as BCC; 15 of these were histologically confirmed, giving a PPV of 83.3% (95% CI, 58.1%–96.4%).PPV 83.3% (95% CI, 58.1%–96.4%)

“This prospective cross-sectional feasibility study was conducted from April 2025 to October 2025”
Claim 4 of 5Not coveredAI-assisted LC-OCT can provide a color-coded probability value in real time for the presence of basal cell carcinoma, while the physician still makes the final diagnosis.View evidenceHide evidence

As statedin real time

Why this verdict

The profile supports use of AI-assisted LC-OCT with automated lesion classification, but it does not describe a color-coded probability display, real-time probability output, or the workflow point that the physician makes the final diagnosis. Those operational details may exist outside the abstract, but they are not verifiable from the supplied abstract-level profile.

1

Study evidence

AI-assisted LC-OCT identified 18 lesions as BCC; 15 of these were histologically confirmed, giving a PPV of 83.3% (95% CI, 58.1%–96.4%).PPV 83.3% (95% CI, 58.1%–96.4%)

“This prospective cross-sectional feasibility study was conducted from April 2025 to October 2025”
Claim 5 of 5Not coveredThe article says SUBSCAN is not yet ready for routine use because there is still a lack of data on sensitivity and the method remains too time-consuming.View evidenceHide evidence

Why this verdict

The profile supports that routine implementation is not recommended yet and that further prospective studies are needed to assess sensitivity, cost-effectiveness, and clinical benefit. The supplied abstract-level profile does not verify the additional stated reason that the method remains too time-consuming, so the full claim is not verifiable at this depth.

1

Study evidence

AI-assisted LC-OCT identified 18 lesions as BCC; 15 of these were histologically confirmed, giving a PPV of 83.3% (95% CI, 58.1%–96.4%).PPV 83.3% (95% CI, 58.1%–96.4%)

“This prospective cross-sectional feasibility study was conducted from April 2025 to October 2025”
2

Study evidence

AI-assisted LC-OCT subtype assignment matched histologic subtype in 13 of 15 histologically confirmed subclinical BCCs.86.7% (13/15); 95% CI, 59.5%–98.3%

“The distribution of BCC subtypes included 13 cases of superficial BCCs (76.5%), followed by 3 cases of nodular BCCs (17.6%), and 1 case of infiltrative BCC (5.9%).”
Then look for missing context

Context layer

What the story left out

Important study details the story did not include.

  • The primary diagnostic-performance endpoint was positive predictive value against histology, not sensitivity; AI-assisted LC-OCT diagnosed 18 lesions as BCC and 15 were histologically confirmed, yielding PPV 83.3% with a wide 95% CI.

    The story says BCC was detected but does not report the paper’s main PPV result, histologic reference standard, false-positive context, or confidence interval. This omission matters because the paper’s evidence is about PPV among AI-positive lesions, not overall screening sensitivity.

    From Prospective cross-sectional feasibility study

  • Seventeen subclinical BCCs were identified in 14 of 150 patients, corresponding to 9.3% of the screened high-risk cohort.

    The story conveys that subclinical BCC was detected but does not include the magnitude of the finding or the denominator, which are central to interpreting screening yield.

    From Prospective cross-sectional feasibility study

  • Two AI-positive lesions were not biopsied because patients declined, creating incomplete histologic verification; a sensitivity analysis counted those lesions as confirmed by expert analysis and produced a higher PPV estimate.

    The story does not mention incomplete reference-standard verification or the sensitivity analysis. This is a material limitation because some confirmation relied on expert image review rather than histology in the sensitivity analysis.

    From Prospective cross-sectional feasibility study; Sensitivity analysis with alternative verification (expert image adjudica

  • The authors also note the need to assess cost-effectiveness and whether earlier detection improves clinical outcomes before routine implementation.

    The story mentions sensitivity and time burden, but it does not mention the paper-profile caveats about cost-effectiveness or proving clinical benefit from earlier detection.

    From Prospective cross-sectional feasibility study; Prospective cross-sectional feasibility study; Sensitivity analysis with

  • Generalizability is limited by the single-center inpatient cohort of patients with at least two BCC risk factors, and by exclusion of macroscopically suggestive lesions.

    Although the story says the patients were high risk, it does not convey the single-center inpatient setting, the specific enriched-risk cohort, or the restriction to inconspicuous lesions. These limitations affect how broadly the findings can be applied.

    From Prospective cross-sectional feasibility study; Prospective cross-sectional feasibility study

  • Subtype classification was a secondary endpoint: AI-assisted LC-OCT subtype assignment matched histology in 13 of 15 confirmed cases, with small numbers and wide confidence intervals.

    The story does not discuss subtype-classification performance or its small-sample limitations. This is a secondary paper element rather than the main detection claim.

    From Prospective cross-sectional feasibility study

3 things the story did carry across
  • The paper is a prospective cross-sectional feasibility study of systematic AI-assisted LC-OCT screening of inconspicuous facial skin in high-risk inpatients after excluding macroscopically suggestive lesions.
  • The paper notes that sensitivity was not estimated and that further prospective studies are needed before routine implementation.
  • Early BCC diagnosis can permit less invasive treatment.
Then read the study layer

Study layer

Study at a glance

Scan the study first. Expand only the parts you want to inspect.

Pieces of work

3

Evidence read

study summary

Lead result

human in vivo

1Lead resulthuman in vivoAssess feasibility and diagnostic performance (PPV) of systematic screening of inconspicuous facial skin in high-risk patients using AI-assisted LC-OCT to detect subclinical basal cell carcinoma, with histology as reference.Prospective cross-sectional feasibility studyExpand

In plain English

Prospective cross-sectional feasibility study evaluating systematic screening of inconspicuous facial skin with AI-assisted line-field confocal optical coherence tomography (LC-OCT) in consecutive inpatients at high risk for basal cell carcinoma (≥2 risk factors). The prespecified primary outcome was positive predictive value (PPV) of AI-assisted LC-OCT for histologically confirmed subclinical BCC; histologic confirmation was obtained for AI-positive lesions when biopsied.

Key findings

  • AI-assisted LC-OCT identified 18 lesions as BCC; 15 of these were histologically confirmed, giving a PPV of 83.3% (95% CI, 58.1%–96.4%).PPV 83.3% (95% CI, 58.1%–96.4%)
  • Sensitivity analysis that included the 2 nonbiopsied lesions as confirmed by expert analysis yielded a PPV of 94.4% (95% CI, 72.7%–99.9%).PPV 94.4% (95% CI, 72.7%–99.9%)
“This prospective cross-sectional feasibility study was conducted from April 2025 to October 2025”
What this piece can’t prove
  • Two AI-positive lesions were not biopsied because patients declined, resulting in incomplete reference-standard verification for those cases.
  • Feasibility cross-sectional design and single-center recruitment of inpatients with ≥2 BCC risk factors limit ability to estimate sensitivity, specify screening yield in broader populations, or determine clinical benefit of earlier detection.
  • Relatively small number of confirmed BCC cases and lesions for subtype-accuracy estimates, resulting in wide confidence intervals.
  • Macroscopically suggestive lesions were excluded from screening, so findings apply to detection of inconspicuous (subclinical) facial lesions rather than general diagnostic use including obvious lesions.
2human in vivoEvaluate accuracy of AI-assisted LC-OCT in classifying BCC subtype (eg, superficial vs nodular vs infiltrative) compared with histology among detected subclinical BCCs.Prospective cross-sectional feasibility studyExpand

In plain English

In a prospective cross-sectional feasibility study of AI-assisted LC-OCT screening of inconspicuous facial skin in high-risk inpatients, 17 subclinical BCCs were detected; subtype distribution was 13 superficial (76.5%), 3 nodular (17.6%), and 1 infiltrative (5.9%). Among the 15 histologically confirmed subclinical BCCs, AI-assigned subtype matched histology in 13 cases (86.7%; 95% CI, 59.5%–98.3%).

Key findings

  • AI-assisted LC-OCT subtype assignment matched histologic subtype in 13 of 15 histologically confirmed subclinical BCCs.86.7% (13/15); 95% CI, 59.5%–98.3%
  • Distribution of BCC subtypes among detected subclinical BCCs was 13 superficial, 3 nodular, and 1 infiltrative.13/17 (76.5%) superficial; 3/17 (17.6%) nodular; 1/17 (5.9%) infiltrative
“The distribution of BCC subtypes included 13 cases of superficial BCCs (76.5%), followed by 3 cases of nodular BCCs (17.6%), and 1 case of infiltrative BCC (5.9%).”
What this piece can’t prove
  • Small sample size and small numbers of lesions in some subtype categories, yielding wide confidence intervals.
  • Two lesions diagnosed by AI did not undergo histologic confirmation and were excluded from the histology-based accuracy calculation.
  • Single-center, inpatient population of patients with ≥2 BCC risk factors; findings may not generalize to broader or community screening populations.
  • Cross-sectional feasibility design; study does not provide sensitivity estimates for screening or evidence of clinical benefit from earlier detection.
3human in vivoExplore impact of handling nonbiopsied AI-positive lesions via a sensitivity analysis (including expert confirmation) on PPV estimates.Sensitivity analysis with alternative verification (expert image adjudication)Expand

In plain English

Sensitivity analysis evaluated how treating two AI-positive, nonbiopsied facial lesions (patients declined biopsy) as true positives based on expert LC-OCT image review affects the estimated positive predictive value (PPV) for AI-assisted LC-OCT detection of subclinical BCC. Under this alternative verification, PPV was 94.4% (95% CI, 72.7%–99.9%), compared with the primary analysis PPV of 83.3% (95% CI, 58.1%–96.4%) based on histology.

Key findings

  • Including two AI-positive lesions that were not biopsied but were confirmed by expert LC-OCT image review increased the PPV of AI-assisted LC-OCT for subclinical BCC to 94.4% (95% CI, 72.7%–99.9%).PPV 94.4% (95% CI, 72.7%–99.9%)
“2 patients declined biopsy.”
What this piece can’t prove
  • Two AI-positive lesions lacked histologic confirmation because patients declined biopsy; sensitivity analysis instead used expert image confirmation.
  • The sensitivity analysis result is based on a small number of additional cases, resulting in wide confidence intervals.
  • Expert LC-OCT confirmation is an alternative verification method and not equivalent to the histologic gold standard reported in the primary analysis.
Finally, the search trail

Method layer

NewsLink found the paper. Tessa takes you deeper.

NewsLink checks the story. Tessa is where you inspect the paper, authors, evidence, and research context.

Papers considered

The selected paper, plus nearby candidates.

PubMed, Europe PMC, Crossref · 30 candidate papers

And 24 more candidates considered.