Skip to content

Deep neural networks show a superior performance to all 42 dermatologists in onychomycosis diagnosis

 

We created datasets of standardized nail images using a region-based convolutional neural network (R-CNN) trained to distinguish a nail from a background. We used R-CNN to generate training datasets of 49,567 images, which were then used to fine-tune the ResNet-152 and VGG-19 models.

The validation datasets comprised 100 and 194 images from Inje University (B1 and B2 datasets, respectively), 125 images from Hallym University (C dataset), and 939 images from Seoul National University (D dataset).


The results from the AI ensemble model (ResNet-152 + VGG-19 + feedforward neural networks) showed the test sensitivity/specificity/ area under the curve values of (96.0 / 94.7 / 0.98), (82.7 / 96.7 / 0.95), (92.3 / 79.3 / 0.93), (87.7 / 69.3 / 0.82) for the B1, B2, C, and D datasets, respectively.

With the combination of the B1 and C datasets, the AI Youden index was significantly higher than that of 42 dermatologists who performed the same assessment manually (p = 0.01). For the combinations of B1+C and B2+D dataset, almost none of the dermatologists performed as well as the AI.


The deep learning model trained with the dataset comprising 49,567 images, achieved a diagnostic accuracy for the classification of onychomycosis, which was superior to that of most of the dermatologists who participated in this study.

 

번호 제목 글쓴이 날짜 조회 수
1686 ubuntu cuda nvidia-smi WHRIA 2020.08.29 849
1685 concat network [3] WHRIA 2020.08.27 147
1684 GPT2 [1] WHRIA 2020.08.03 250
1683 scopus [1] WHRIA 2020.08.02 344
1682 melafind WHRIA 2020.08.01 771
1681 nvidia dali [1] WHRIA 2020.08.01 265
1680 pytorch optimize WHRIA 2020.08.01 126
1679 startup , 미국 WHRIA 2020.07.30 134
1678 consort , stard WHRIA 2020.07.30 81
1677 model split [1] WHRIA 2020.07.30 142
1676 FTC [1] WHRIA 2020.07.28 238
1675 암호화 WHRIA 2020.07.26 1000
1674 raid 6 WHRIA 2020.07.23 43
1673 asyncio WHRIA 2020.07.23 286
1672 amp distributed pytorch [1] WHRIA 2020.07.14 51

Powered by Xpress Engine / Designed by Sketchbook

sketchbook5, 스케치북5

sketchbook5, 스케치북5

나눔글꼴 설치 안내


이 PC에는 나눔글꼴이 설치되어 있지 않습니다.

이 사이트를 나눔글꼴로 보기 위해서는
나눔글꼴을 설치해야 합니다.

설치 취소