Man against machine: diagnostic performance of a deep learning convolutional neural network for dermoscopic melanoma recognition in comparison to 58 dermatologists

Haenssle, Holger A.; Schneiderbauer, Roland; Toberer, Ferdinand; Buhl, Timo; Kalloo, Aadi; Hassen, A.; Thomas, Luc; Enk, Alexander; Uhlmann, Lorenz; Alt, Christina; Arenbergerová, Monika; Bakos, Renato Marchiori; Baltzer, Anne; Bertlich, Ines; Blum, Andreas; Bokor-Billmann, Therezia; Bowling, Jonathan; Braghiroli, Naira; Braun, Ralph; Buder‐Bakhaya, Kristina; Cabo, Horacio; Čabrijan, Leo; Cevic, Naciye; Claßen, Anna; Deltgen, David; Fink, Christine; Georgieva, Ivelina; Hakim‐Meibodi, Lara‐Elena; Hanner, Susanne; Hartmann, Fritz; Hartmann, Julia; Haus, Georg; Hoxha, Elti; Karls, Raimonds; Koga, Hiroshi; Kreusch, Jürgen; Lallas, Aimilios; Majenka, Pawel; Marghoob, Ashfaq; Massone, Cesare; Mekokishvili, Lali; Mestel, Dominik; Meyer, V.; Neuberger, Anna; Nielsen, Kari; Oliviero, Margaret; Pampena, Riccardo; Paoli, John; Pawlik, Erika; Rao, B. K. Nagaraja; Rendón, Adriana; Russo, Teresa; Sadek, Ahmed; Samhaber, Kinga T.; Schweizer, Anissa; Trennheuser, Lukas; Vlahova, Lyobomira; Wald, Alexander; Winkler, Julia; Wölbing, Priscila; Zalaudek, Iris

doi:10.1093/annonc/mdy166

Cited by 991 publications

(601 citation statements)

References 20 publications

Supporting

Mentioning

546

Contrasting

Unclassified

Order By: Relevance

“…In this study, we show that the new algorithm for risk assessment has a high accuracy for self‐assessment of skin lesions for skin cancer. This result is comparable to some of the recently published, best skin cancer disease classification algorithms …”

Section: Discussionsupporting

confidence: 86%

“…In the first study, 21 dermatologists scored, on average, a sensitivity and specificity higher than 90% and 70%, respectively. In the second study, the average sensitivity of 58 dermatologists is 87%, for a 71% specificity. Though there are some differences in the setup, this is comparable to the level shown by the risk assessment algorithm evaluated here.…”

Section: Discussionmentioning

confidence: 89%

“…Access to medical care is being transformed with the advent of mobile applications which can monitor, diagnose and provide health information. In the field of dermatology, there is considerable interest in developing algorithms for image analysis of skin lesions to evaluate the risk of malignant melanoma (MM) and keratinocyte carcinomas (KC) . Early detection of skin cancer is important since the prognosis is highly dependent on whether the cancer is detected at an early stage, when it is easier to treat and its prognosis is more favourable…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Accuracy of a smartphone application for triage of skin lesions based on machine learning algorithms

Udrea

Mitra

Costea

et al. 2019

Acad Dermatol Venereol

View full text Add to dashboard Cite

Background Machine learning algorithms achieve expert‐level accuracy in skin lesion classification based on clinical images. However, it is not yet shown whether these algorithms could have high accuracy when embedded in a smartphone app, where image quality is lower and there is high variability in image taking scenarios by users. In the past, these applications were criticized due to lack of accuracy. Objective In this study, we evaluate the accuracy of the newest version of a smartphone application (SA) for risk assessment of skin lesions. Methods This SA uses a machine learning algorithm to compute a risk rating. The algorithm is trained on 131 873 images taken by 31 449 users in multiple countries between January 2016 and August 2018 and rated for risk by dermatologists. To evaluate the sensitivity of the algorithm, we use 285 histopathologically validated skin cancer cases (including 138 malignant melanomas), from two previously published clinical studies (195 cases) and from the SA user database (90 cases). We calculate the specificity on a separate set from the SA user database containing 6000 clinically validated benign cases. Results The algorithm scored a 95.1% (95% CI, 91.9–97.3%) sensitivity in detecting (pre)malignant conditions (93% for malignant melanoma and 97% for keratinocyte carcinomas and precursors). This level of sensitivity was achieved with a 78.3% (95% CI, 77.2–79.3%) specificity. Conclusions This SA provides a high sensitivity to detect skin cancer; however, there is still room for improvement in terms of specificity. Future studies are needed to assess the impact of this SA on the health systems and its users.

show abstract

Section: Discussionsupporting

confidence: 86%

Section: Discussionmentioning

confidence: 89%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Accuracy of a smartphone application for triage of skin lesions based on machine learning algorithms

Udrea

Mitra

Costea

et al. 2019

Acad Dermatol Venereol

View full text Add to dashboard Cite

show abstract

“…Algorithms have been shown to outperform the dermatologists in the detection of benign lesions [20]. Although we did not specifically address the value of artificial intelligence in dermatoscopy, the majority of patients welcomed the digital progress in medicine.…”

Section: Discussionmentioning

confidence: 99%

Patient Attitude towards Videodermatoscopy for the Detection of Skin Cancer: A Cross-Sectional Study

et al. 2019

View full text Add to dashboard Cite

Background: Videodermatoscopy (VD) is a useful device for supporting dermatologists in the distinction between benign and malignant lesions. However, only few patients have access to VD in daily practice. Objectives: To investigate patient attitudes towards VD. Method: A cross-sectional study was conducted between May and June 2018. Patients were asked to complete a self-administered questionnaire on the popularity of VD. Descriptive analysis was performed including contingency tables and χ² tests to investigate associations between sociodemographic data and the popularity of VD. Results: A total of 61.2% (123/201) of the patients had not heard of VD at the time of assessment or were unsure. Of the 38.8% of patients (78/201) who already knew of VD, 64.1% (50/78) reported that they had already been investigated by VD; 57.5% (111/193) were willing to pay an extra fee for VD. A high level of education and private insurance status had a statistically significant association with the popularity of VD (p = 0.036 and p = 0.026, respectively). Conclusions: There was a strong information deficit, especially in patients with lower education and statutory health insurance. Nevertheless, the willingness to pay an extra fee for a VD-assisted skin examination was high. Dermatologists should actively offer and inform their patients about VD when performing skin cancer screening.

show abstract

“…The difficulty in producing meaningful results may be due to the inexperience of human interpreters on reflectance confocal microscopic images. In a recent study by Haenssle et al [8], dermoscopic images of suspected melanoma were correctly categorized (benign, in situ, or invasive) by AI at levels equal to and greater than experienced dermatologists. This suggests that AI trained with data labeled by expert physicians may have utility as a screening tool in patient populations that have low access to dermatological care.…”

mentioning

confidence: 99%

Augmented Intelligence in Dermatology: Fantasy or Future?

2019

View full text Add to dashboard Cite

Man against machine: diagnostic performance of a deep learning convolutional neural network for dermoscopic melanoma recognition in comparison to 58 dermatologists

Abstract: This study was registered at the German Clinical Trial Register (DRKS-Study-ID: DRKS00013570; https://www.drks.de/drks_web/).

Cited by 991 publications

References 20 publications

Accuracy of a smartphone application for triage of skin lesions based on machine learning algorithms

Accuracy of a smartphone application for triage of skin lesions based on machine learning algorithms

Patient Attitude towards Videodermatoscopy for the Detection of Skin Cancer: A Cross-Sectional Study

Augmented Intelligence in Dermatology: Fantasy or Future?

Contact Info

Product

Resources

About