当前位置:网站首页>Deep transfer learning (CS CV) for automatic diagnosis of skin lesions from photographs

Deep transfer learning (CS CV) for automatic diagnosis of skin lesions from photographs

2020-12-07 19:23:54 Ling Qian

Melanoma is the most common skin cancer in the world . at present , The disease is diagnosed by a dermatologist , It's expensive , You need to see a doctor in time . Recent advances in deep learning have the potential to improve diagnostic performance , Speed up emergency referral and reduce the burden on clinicians . Through smartphones , This technology can benefit people who don't usually have access to this kind of health care , for example , In remote parts of the world , Due to financial constraints , Or in the 2020 year ,COVID-19 Cancel . So , We have studied various transfer learning methods , utilize ImageNet On the model parameters trained in advance , And the melanoma detection was fine tuned . We compare learning with and without transfer learning EfficientNet、MnasNet、MobileNet、DenseNet、squezenet、ShuffleNet、GoogleNet、ResNet、ResNeXt、VGG And a simple convolution neural network (CNN). We found that , In mobile networks ,EfficientNet( Learning with transfer ) The average performance of the , The area under the receiver operating characteristic curve (AUROC) by 0.931±0.005, The area under the exact recall curve is 0.840±0.010. It's obviously better than a GP (0.83±0.03 AUROC) And dermatologists (0.91±0.02 AUROC).

Original title :Deep Transfer Learning for Automated Diagnosis of Skin Lesions from Photographs

original text :Melanoma is the most common form of skin cancer worldwide. Currently, the disease is diagnosed by expert dermatologists, which is costly and requires timely access to medical treatment. Recent advances in deep learning have the potential to improve diagnostic performance, expedite urgent referrals and reduce burden on clinicians. Through smart phones, the technology could reach people who would not normally have access to such healthcare services, e.g. in remote parts of the world, due to financial constraints or in 2020, COVID-19 cancellations. To this end, we have investigated various transfer learning approaches by leveraging model parameters pre-trained on ImageNet with finetuning on melanoma detection. We compare EfficientNet, MnasNet, MobileNet, DenseNet, SqueezeNet, ShuffleNet, GoogleNet, ResNet, ResNeXt, VGG and a simple CNN with and without transfer learning. We find the mobile network, EfficientNet (with transfer learning) achieves the best mean performance with an area under the receiver operating characteristic curve (AUROC) of 0.931±0.005 and an area under the precision recall curve (AUPRC) of 0.840±0.010. This is significantly better than general practitioners (0.83±0.03 AUROC) and dermatologists (0.91±0.02 AUROC).

Original author :Emma Rocheteau, Doyoon Kim

Original address :https://arxiv.org/abs/2011.04475

Original statement , This article is authorized by the author + Community publication , Unauthorized , Shall not be reproduced .

If there is any infringement , Please contact the yunjia_community@tencent.com Delete .

版权声明
本文为[Ling Qian]所创,转载请带上原文链接,感谢
https://chowdera.com/2020/11/20201119031742521b.html