Data, Depth, and Design: Learning Reliable Models for Skin Lesion Analysis

Valle, Eduardo; Fornaciali, Michel; Menegola, Afonso; Tavares, Julia Valentim; Bittencourt, Flávia Vasques; Li, Lin Tzy; Avila, Sandra

doi:10.48550/arxiv.1711.00441

Cited by 7 publications

(15 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In previous work, we showed that the training set size responds by almost 50% of the variation on the prediction power of the classifier [5]. The freedom to use external sources enabled us to gather more data to boost our models.…”

Section: B Datamentioning

confidence: 94%

“…We aimed, from the start, at deep learning solutions for all tasks. We know from experience that the success factors for a competitive deep learning approach are data availability and model depth [5,6]. To improve our chances, we also introduced two original contributions -synthetic lesions generation and stronger data augmentation approaches -to boost the models training.…”

Section: A Strategymentioning

confidence: 99%

“…Our starting point was our last work on how to design powerful deep-learning classifiers for skin lesions [5]. We evaluated the main factors that vary on the approaches found in literature: use of transfer learning, model architecture, train dataset, image resolution, type of data augmentation, input normalization, use of segmentation, duration of training, additional use of SVM, and test data augmentation.…”

Section: Experimental Design Tacticsmentioning

confidence: 99%

“…Using our participation on ISIC Challenge 2017 (with Inception-v4) with as baseline, we observed similar performance using the new data augmentation methods, but without using external data. That is, the image processing data augmentation methods were equal to the performance of the model trained with external data (which we know that has a huge impact on the classifier prediction power [5]). Among all experiments and scenarios, scenario J (please refer to [12] for details) leads to better performance and was the one introduced in the experiments of the competition (only in Task 3).…”

Section: Notable Noveltiesmentioning

confidence: 99%

“…to the performance of a skin lesion classifier was not evident. We cleared such question by extensively analyzing several combinations of architectures, dataset sizes, and other eight relevant aspects [5].…”

mentioning

confidence: 99%

See 4 more Smart Citations

Deep-Learning Ensembles for Skin-Lesion Segmentation, Analysis, Classification: RECOD Titans at ISIC Challenge 2018

Bissoto¹,

Perez²,

Ribeiro³

et al. 2018

Preprint

Self Cite

View full text Add to dashboard Cite

Section: B Datamentioning

confidence: 94%

Section: A Strategymentioning

confidence: 99%

Section: Experimental Design Tacticsmentioning

confidence: 99%

Section: Notable Noveltiesmentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations

Deep-Learning Ensembles for Skin-Lesion Segmentation, Analysis, Classification: RECOD Titans at ISIC Challenge 2018

Bissoto¹,

Perez²,

Ribeiro³

et al. 2018

Preprint

Self Cite

View full text Add to dashboard Cite

Skin Lesion Synthesis with Generative Adversarial Networks

Bissoto

Perez

Valle

et al. 2018

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Skin cancer is by far the most common type of cancer. Early detection is the key to increase the chances for successful treatment significantly. Currently, Deep Neural Networks are the state-of-the-art results on automated skin cancer classification. To push the results further, we need to address the lack of annotated data, which is expensive and require much effort from specialists. To bypass this problem, we propose using Generative Adversarial Networks for generating realistic synthetic skin lesion images. To the best of our knowledge, our results are the first to show visually-appealing synthetic images that comprise clinically-meaningful information.

show abstract

Data Augmentation for Skin Lesion Analysis

Perez

Vasconcelos

Avila

et al. 2018

Lecture Notes in Computer Science

Self Cite

132

View full text Add to dashboard Cite

Deep learning models show remarkable results in automated skin lesion analysis. However, these models demand considerable amounts of data, while the availability of annotated skin lesion images is often limited. Data augmentation can expand the training dataset by transforming input images. In this work, we investigate the impact of 13 data augmentation scenarios for melanoma classification trained on three CNNs (Inception-v4, ResNet, and DenseNet). Scenarios include traditional color and geometric transforms, and more unusual augmentations such as elastic transforms, random erasing and a novel augmentation that mixes different lesions. We also explore the use of data augmentation at test-time and the impact of data augmentation on various dataset sizes. Our results confirm the importance of data augmentation in both training and testing and show that it can lead to more performance gains than obtaining new images. The best scenario results in an AUC of 0.882 for melanoma classification without using external data, outperforming the top-ranked submission (0.874) for the ISIC Challenge 2017, which was trained with additional data.

show abstract

Data, Depth, and Design: Learning Reliable Models for Skin Lesion Analysis

Cited by 7 publications

References 0 publications

Deep-Learning Ensembles for Skin-Lesion Segmentation, Analysis, Classification: RECOD Titans at ISIC Challenge 2018

Deep-Learning Ensembles for Skin-Lesion Segmentation, Analysis, Classification: RECOD Titans at ISIC Challenge 2018

Skin Lesion Synthesis with Generative Adversarial Networks

Data Augmentation for Skin Lesion Analysis

Contact Info

Product

Resources

About