内容紹介
人工知能を実現する方法の一つとして、深層学習が注目を集めている。近年になって、深層学習の手法の一つであるconvolutional neural network(CNN)の放射線画像への応用も研究報告が見られるようになってきた。本記事では、CNNの実装や学習法、CNNの性能評価などについて概観する。また、CNNモデルを用いたダイナミックCT画像における肝腫瘤分類についても紹介する。
Recently, deep learning with convolutional neural networks (CNNs) is gaining attention for its high performance in image recognition tasks. Most recently, application of this technique to radiological images is beginning to be investigated. In this article, overview of the deep learning technique (features of CNNs, overfitting problems, data collection for CNNs, CNN implementation, training of CNNs, and testing with CNNs) is described. We also described the overview of its usefulness in categorizing liver masses on dynamic contrastenhanced computed tomography images.
目次
Recently, deep learning with convolutional neural networks (CNNs) is gaining attention for its high performance in image recognition tasks. Most recently, application of this technique to radiological images is beginning to be investigated. In this article, overview of the deep learning technique (features of CNNs, overfitting problems, data collection for CNNs, CNN implementation, training of CNNs, and testing with CNNs) is described. We also described the overview of its usefulness in categorizing liver masses on dynamic contrastenhanced computed tomography images.