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

• Published in: CAAI Transactions on Intelligence Technology Vol. 3; no. 3; pp. 153 - 160
• Main Authors: Pujitha, Appan K, Sivaswamy, Jayanthi
• Format: Journal Article
• Language: English
• Published: Beijing The Institution of Engineering and Technology 01.09.2018; John Wiley & Sons, Inc; Wiley
• Subjects: (B6135) Optical, image and video signal processing; (C5260B) Computer vision and image processing techniques; (C6170K) Knowledge engineering techniques; (C7330) Biology and medical computing; Algorithms; annotated data; Annotations; Artificial neural networks; CAD; colour fundus images; Computer aided design; crowdsourced annotations; crowdsourced region; Crowdsourcing; data augmentation; data-driven approaches; Deep learning; detection performance; developing computer; diagnosis algorithms; diseases; Exudation; generative adversarial network-based solution; Generative adversarial networks; good performance; image classification; image colour analysis; image level; Image processing; image segmentation; interest markings; learning (artificial intelligence); Lesions; Machine learning; Magnetic resonance imaging; medical data; medical domain; medical image processing; Medical imaging; Network reliability; neural nets; Neural networks; pixel-level markings; popular solution; processed/refined crowdsourced data/synthetic images; Retinal images; severity level; Sparsity; Special Issue: Selected Papers from The 4th Asian Conference on Pattern Recognition (ACPR 2017); Synthetic data; synthetic image generation; synthetically generated data; diagnosis algorithms; image classification; data augmentation; popular solution; detection performance; annotated data; synthetically generated data; deep learning; medical domain; crowdsourced region; image level; interest markings; image segmentation; processed/refined crowdsourced data/synthetic images; image colour analysis; learning (artificial intelligence); medical image processing; generative adversarial network-based solution; CAD; diseases; good performance; colour fundus images; crowdsourced annotations; machine learning; severity level; data-driven approaches; synthetic image generation; medical data; pixel-level markings; neural nets; developing computer

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 annotat...