Consistency of convolutional neural networks in dermoscopic melanoma recognition: A prospective real-world study about the pitfalls of augmented intelligence

BACKGROUND

Deep-learning convolutional neural networks (CNNs) have outperformed even experienced dermatologists in dermoscopic melanoma detection under controlled conditions. It remains unexplored how real-world dermoscopic image transformations affect CNN robustness.

OBJECTIVES

To investigate the consistency of melanoma risk assessment by two commercially available CNNs to help formulate recommendations for current clinical use.

METHODS

A comparative cohort study was conducted from January to July 2022 at the Department of Dermatology, University Hospital Basel. Five dermoscopic images of 116 different lesions on the torso of 66 patients were captured consecutively by the same operator without deliberate rotation. Classification was performed by two CNNs (CNN-1/CNN-2). Lesions were divided into four subgroups based on their initial risk scoring and clinical dignity assessment. Reliability was assessed by variation and intraclass correlation coefficients. Excisions were performed for melanoma suspicion or two consecutively elevated CNN risk scores, and benign lesions were confirmed by expert consensus (n = 3).

RESULTS

117 repeated image series of 116 melanocytic lesions (2 melanomas, 16 dysplastic naevi, 29 naevi, 1 solar lentigo, 1 suspicious and 67 benign) were classified. CNN-1 demonstrated superior measurement repeatability for clinically benign lesions with an initial malignant risk score (mean variation coefficient (mvc): CNN-1: 49.5(±34.3)%; CNN-2: 71.4(±22.5)%; p = 0.03), while CNN-2 outperformed for clinically benign lesions with benign scoring (mvc: CNN-1: 49.7(±22.7)%; CNN-2: 23.8(±29.3)%; p = 0.002). Both systems exhibited lowest score consistency for lesions with an initial malignant risk score and benign assessment. In this context, averaging three initial risk scores achieved highest sensitivity of dignity assessment (CNN-1: 94%; CNN-2: 89%). Intraclass correlation coefficients indicated 'moderate'-to-'good' reliability for both systems (CNN-1: 0.80, 95% CI:0.71-0.87, p < 0.001; CNN-2: 0.67, 95% CI:0.55-0.77, p < 0.001).

CONCLUSIONS

Potential user-induced image changes can significantly influence CNN classification. For clinical application, we recommend using the average of three initial risk scores. Furthermore, we advocate for CNN robustness optimization by cross-validation with repeated image sets.

TRIAL REGISTRATION

ClinicalTrials.gov (NCT04605822).

Improved diagnosis by automated macro- and micro-anatomical region mapping of skin photographs

BACKGROUND

The exact location of skin lesions is key in clinical dermatology. On one hand, it supports differential diagnosis (DD) since most skin conditions have specific predilection sites. On the other hand, location matters for dermatosurgical interventions. In practice, lesion evaluation is not well standardized and anatomical descriptions vary or lack altogether. Automated determination of anatomical location could benefit both situations.

OBJECTIVE

Establish an automated method to determine anatomical regions in clinical patient pictures and evaluate the gain in DD performance of a deep learning model (DLM) when trained with lesion locations and images.

METHODS

Retrospective study based on three datasets: macro-anatomy for the main body regions with 6000 patient pictures partially labelled by a student, micro-anatomy for the ear region with 182 pictures labelled by a student and DD with 3347 pictures of 16 diseases determined by dermatologists in clinical settings. For each dataset, a DLM was trained and evaluated on an independent test set. The primary outcome measures were the precision and sensitivity with 95% CI. For DD, we compared the performance of a DLM trained with lesion pictures only with a DLM trained with both pictures and locations.

RESULTS

The average precision and sensitivity were 85% (CI 84-86), 84% (CI 83-85) for macro-anatomy, 81% (CI 80-83), 80% (CI 77-83) for micro-anatomy and 82% (CI 78-85), 81% (CI 77-84) for DD. We observed an improvement in DD performance of 6% (McNemar test P-value 0.0009) for both average precision and sensitivity when training with both lesion pictures and locations.

CONCLUSION

Including location can be beneficial for DD DLM performance. The proposed method can generate body region maps from patient pictures and even reach surgery relevant anatomical precision, e.g. the ear region. Our method enables automated search of large clinical databases and make targeted anatomical image retrieval possible.