Detecting Melanoma Fairly: Skin Tone Detection and Debiasing for Skin Lesion Classification (2022-02-06T00:00:00.000000Z)

TL;DR

An efficient yet effective algorithm for automatically labelling the skin tone of lesion images is proposed, and this is used to annotate the benchmark ISIC dataset and to use these automated labels as the target for two leading bias unlearning techniques towards mitigating skin tone bias.

Abstract

Convolutional Neural Networks have demonstrated human-level performance in the classification of melanoma and other skin lesions, but evident performance disparities between differing skin tones should be addressed before widespread deployment. In this work, we propose an efficient yet effective algorithm for automatically labelling the skin tone of lesion images, and use this to annotate the benchmark ISIC dataset. We subsequently use these automated labels as the target for two leading bias unlearning techniques towards mitigating skin tone bias. Our experimental results provide evidence that our skin tone detection algorithm outperforms existing solutions and that unlearning skin tone may improve generalisation and can reduce the performance disparity between melanoma detection in lighter and darker skin tones.

Authors

Amir Atapour-Abarghouei

2 papers

Peter J. Bevan

1 papers

Detecting Melanoma Fairly: Skin Tone Detection and Debiasing for Skin Lesion Classification

TL;DR

Abstract

Authors

References26 items

Evaluating Deep Neural Networks Trained on Clinical Images in Dermatology with the Fitzpatrick 17k Dataset

A patient-centric dataset of images and metadata for identifying melanomas using clinical context

PyTorch: An Imperative Style, High-Performance Deep Learning Library

Beyond Synthetic Noise: Deep Learning on Controlled Noisy Labels

Estimating Skin Tone and Effects on Classification Performance in Dermatology Datasets

EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks

Comparing artificial intelligence algorithms to 157 German dermatologists: the melanoma classification benchmark.

A convolutional neural network trained with dermoscopic images performed on par with 145 dermatologists in a clinical melanoma image classification task.

Uncovering and Mitigating Algorithmic Bias through Learned Latent Structure

Learning Not to Learn: Training Deep Neural Networks With Biased Data

Turning a Blind Eye: Explicit Removal of Biases and Variation from Deep Neural Network Embeddings

Man against machine: diagnostic performance of a deep learning convolutional neural network for dermoscopic melanoma recognition in comparison to 58 dermatologists

Classification of the Clinical Images for Benign and Malignant Cutaneous Tumors Using a Deep Learning Algorithm.

Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification

Aggregated Residual Transformations for Deep Neural Networks

Densely Connected Convolutional Networks

Skin lesion analysis toward melanoma detection: A challenge at the 2017 International symposium on biomedical imaging (ISBI), hosted by the international skin imaging collaboration (ISIC)

Deep Residual Learning for Image Recognition

Rethinking the Inception Architecture for Computer Vision

Simultaneous Deep Transfer Across Domains and Tasks

Domain-Adversarial Training of Neural Networks

Racial Differences in Survival after Surgical Treatment for Melanoma

Interactive Atlas of Dermoscopy

The validity and practicality of sun-reactive skin types I through VI.

Peter A. Lio and Paul Nghiem

Detecting Melanoma Fairly

Field of Study