Solution to overcome the sparsity issue of annotated data in medical domain

CAAI Transactions on Intelligence Technology

article

Appan K. Pujitha¹  Jayanthi Sivaswamy¹ 

[1] Center for Visual Information Technology

关键词: learning (artificial intelligence);    image colour analysis;    neural nets;    image classification;    image segmentation;    medical image processing;    diseases;    annotated data;    medical domain;    machine learning;    developing computer;    diagnosis algorithms;    CAD;    good performance;    medical data;    image level;    data-driven approaches;    deep learning;    data augmentation;    popular solution;    synthetic image generation;    crowdsourced annotations;    interest markings;    pixel-level markings;    generative adversarial network-based solution;    severity level;    crowdsourced region;    synthetically generated data;    colour fundus images;    processed/refined crowdsourced data/synthetic images;    detection performance;    (B6135) Optical;    image and video signal processing;    (C5260B) Computer vision and image processing techniques;    (C6170K) Knowledge engineering techniques;    (C7330) Biology and medical computing;   

DOI  :  10.1049/trit.2018.1010

学科分类：数学（综合）

来源: Wiley

【 摘 要 】

Annotations are critical for machine learning and developing computer aided diagnosis (CAD) algorithms. Good performance of CAD is critical to their adoption, which generally rely on training with a wide variety of annotated data. However, a vast amount of medical data is either unlabeled or annotated only at the image-level. This poses a problem for exploring data driven approaches like deep learning for CAD. In this paper, we propose a novel crowdsourcing and synthetic image generation for training deep neural net-based lesion detection. The noisy nature of crowdsourced annotations is overcome by assigning a reliability factor for crowd subjects based on their performance and requiring region of interest markings from the crowd. A generative adversarial network-based solution is proposed to generate synthetic images with lesions to control the overall severity level of the disease. We demonstrate the reliability of the crowdsourced annotations and synthetic images by presenting a solution for training the deep neural network (DNN) with data drawn from a heterogeneous mixture of annotations. Experimental results obtained for hard exudate detection from retinal images show that training with refined crowdsourced data/synthetic images is effective as detection performance in terms of sensitivity improves by 25%/27% over training with just expert-markings.

【 授权许可】

CC BY|CC BY-ND|CC BY-NC|CC BY-NC-ND