Development and validation of two artificial intelligence models for diagnosing benign, pigmented facial skin lesions

Abstract

Objective This study used deep learning for diagnosing common, benign hyperpigmentation. Method In this study, two convolutional neural networks were used to identify six pigmentary diseases, and a disease diagnosis model was established. Because the distribution of lesions in the original training picture is very complex, we cropped the image around the lesions, trained the network on the extracted lesion images, and fused the verification results of the overall picture and the extracted picture to assess the model performance in identifying hyperpigmented dermatitis pictures. Finally, we evaluated the image recognition performance of the two convolutional neural networks and the converged networks in the test set through a comparison of the converged network and the physicians’ assessments. Results The AUC of DenseNet‐96 for the overall picture was 0.98, whereas the AUC of ResNet‐152 was 0.96; therefore, we concluded that DenseNet‐96 performed better than ResNet‐152. From the AUC, the converged network has the best performance. The converged network model achieved a comprehensive classification performance comparable to that of the doctors. Conclusions The diagnostic model for benign, pigmented skin lesions based on convolutional neural networks had a slightly higher overall performance than the skin specialists.

Deep Learning in Skin Disease Image Recognition: A Review

“…However, Non-neoplastic skin diseases are scarce, and only three articles on deep learning of eczema and psoriasis identification [46], [51], and [74] are collected in this study. [44], [45], [47], [48], [50], [52], [53], [54], [55], [56], [57], [60], [61], [64], [65], [76], [77], [78], [79], [80], [82], and [84] Seborrheic keratosis [44], [45], [47], [48], [50], [52], [53], [55], [60], [61], [62], [71], [75], [79], [80], [83], and [84] Malignant neoplasms Basal cell carcinoma [38], [39], [41], [42], [54], [56], [57...…”

“…The HAM10000 dataset is a dermoscopic image collected from the Dermatology Department of Vienna Medical University in Austria and the dermatology practice of Cliff Rosendahl in Queensland, Australia [29]. In this study, 18 datasets were collected, including 14 public datasets and 4 self-collected datasets [38], [49], [72], [75]. See Table II for details.…”

“…Table III shows the details. [47], [50], [51], [52], [53], [58], and [65] Clipping [38], [44], [49], [53], [55], [56], [57], and [75] - [41], [42], [59], and [70] Normalization Data normalization [41], [47], [48], [52], [53], [54], [56], [58], [65], [70], and [75] This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ This article has been accepted for publication in a future issue of this journal, but has not been fully edited.…”

“…Eight articles are about AlexNet ( [41], [42], [47], [52], [59], [60], [74], and [80]), seven articles used VGG ( [47], [52], [54], [55], [60], [61], [72], and [73]), and seven articles utilized the Inception ( [38], [43], [54], [57], [60], [78], and [85]). Among the collected literature, at most 18 papers focused on skin disease identification by ResNet ( [38], [42], [47], [49], [50], [52], [53], [57], [60], [64], [65], [75], [76], [77], [79], [82], [84], and [85]). At least four papers used DenseNet ( [56], [57], [75], and [85]).…”

“…Among the collected literature, at most 18 papers focused on skin disease identification by ResNet ( [38], [42], [47], [49], [50], [52], [53], [57], [60], [64], [65], [75], [76], [77], [79], [82], [84], and [85]). At least four papers used DenseNet ( [56], [57], [75], and [85]). Certain papers are related to some relatively new networks, such as Zhuang et al using SENet ( [42] and [91]) and Hameed et al utilizing SqueezeNet ([51] and [92]).…”