Past and present of computer-assisted dermoscopic diagnosis: performance of a conventional image analyser versus a convolutional neural network in a prospective data set of 1,981 skin lesions.

Convolutional neural networks (CNNs) have shown a dermatologist-level performance in the classification of skin lesions. We aimed to deliver a head-to-head comparison of a conventional image analyser (CIA), which depends on segmentation and weighting of handcrafted features, to a CNN trained by deep...

Full description

Bibliographic Details
Published in:European Journal of Cancer Vol. 135; pp. 39 - 47
Main Authors: Sies, Katharina, Winkler, Julia K., Fink, Christine, Bardehle, Felicitas, Toberer, Ferdinand, Buhl, Timo, Enk, Alexander, Blum, Andreas, Rosenberger, Albert, Haenssle, Holger A.
Format: research tables/charts Journal Article
Published: Pergamon Press - An Imprint of Elsevier Science Aug2020
Online Access:View this record in EBSCOhost
fields @attributes:
  recordID: 1
pdfLink:
plink: https://search.ebscohost.com/login.aspx?direct=true&db=ccm&AN=144622341&site=ehost-live
header:
  @attributes:
    shortDbName: ccm
    uiTerm: 144622341
    longDbName: CINAHL Complete
    uiTag: AN
  controlInfo:
    bkinfo:
    dissinfo:
    jinfo:
      jid:
        09598049
        1VW
      jtl: European Journal of Cancer
      issn: 09598049
      maglogo: N
    pubinfo:
      dt: Aug2020
      vid: 135
      pid: 2410
      pub: Pergamon Press - An Imprint of Elsevier Science
    artinfo:
      ui:
        144622341
        144622341
        144622341
        10.1016/j.ejca.2020.04.043
        144622341
      ppf: 39
      ppct: 8
      formats:
      tig:
        atl: Past and present of computer-assisted dermoscopic diagnosis: performance of a conventional image analyser versus a convolutional neural network in a prospective data set of 1,981 skin lesions.
      aug:
        au:
          Sies, Katharina
          Winkler, Julia K.
          Fink, Christine
          Bardehle, Felicitas
          Toberer, Ferdinand
          Buhl, Timo
          Enk, Alexander
          Blum, Andreas
          Rosenberger, Albert
          Haenssle, Holger A.
        affil: Department of Dermatology, University of Heidelberg, Heidelberg, Germany
      sug:
        subj:
          Skin Neoplasms Diagnosis
          Image Interpretation, Computer Assisted Methods
          Deep Learning
          Neural Networks (Computer)
          Predictive Value of Tests
          Human
          Descriptive Statistics
          Prospective Studies
          Image Interpretation, Computer Assisted Equipment and Supplies
          Cross Sectional Studies
          Male
          Female
          Middle Age
          Comparative Studies
          Confidence Intervals
          Sensitivity and Specificity
          ROC Curve
          Melanoma
          Algorithms
          Skin Neoplasms Classification
          Middle Aged: 45-64 years
          Male
          Female
      ab: Convolutional neural networks (CNNs) have shown a dermatologist-level performance in the classification of skin lesions. We aimed to deliver a head-to-head comparison of a conventional image analyser (CIA), which depends on segmentation and weighting of handcrafted features, to a CNN trained by deep learning. Cross-sectional study using a real-world, prospectively acquired, dermoscopic dataset of 1981 skin lesions to compare the diagnostic performance of a market-approved CNN (Moleanalyzer-Pro™, developed in 2018) to a CIA (Moleanalyzer-3™/Dynamole™; developed in 2004, all FotoFinder Systems Inc, Germany). As a reference standard, we used histopathological diagnoses (n = 785) or, in non-excised benign lesions (n = 1196), expert consensus plus an uneventful follow-up by sequential digital dermoscopy for at least 2 years. A total of 281 malignant lesions and 1700 benign lesions from 435 patients (62.2% male, mean age: 52 years) were prospectively imaged. The CNN showed a sensitivity of 77.6% (95% confidence interval [CI]: [72.4%–82.1%]), specificity of 95.3% (95% CI: [94.2%–96.2%]), and receiver operating characteristic (ROC)-area under the curve (AUC) of 0.945 (95% CI: [0.930–0.961]). In contrast, the CIA achieved a sensitivity of 53.4% (95% CI: [47.5%–59.1%]), specificity of 86.6% (95% CI: [84.9%–88.1%]) and ROC-AUC of 0.738 (95% CI: [0.701–0.774]). The data set included melanomas originally diagnosed by dynamic changes during sequential digital dermoscopy (52 of 201, 20.6%), which reduced the sensitivities of both classifiers. Pairwise comparisons of sensitivities, specificities, and ROC-AUCs indicated a clear outperformance by the CNN (all p < 0.001). The superior diagnostic performance of the CNN argues against a continued application of former CIAs as an aide to physicians' clinical management decisions. • We compared two market-approved computer-algorithms for skin cancer detection. • A conventional image analyser (CIA) was compared with a deep learning convolutional neural network (CNN). • The CNN significantly outperformed the CIA in sensitivity, specificity, and area under the receiver operating characteristic curve.
      pubtype: Academic Journal
      doctype:
        research
        tables/charts
        Journal Article
      ougenre: Article
    language: English
    refInfo:
    holdings:
      @attributes:
        islocal: N