Deep Convolutional Neural Networks for Computer-Aided Detection: CNN Architectures, Dataset Characteristics and Transfer Learning
Studies CNN architectures, dataset scale, and ImageNet transfer learning for medical computer-aided detection of lymph nodes and lung disease.
Deep CNNs excel at image recognition given large annotated datasets, but such data is scarce in medical imaging. This paper examines three understudied factors for applying CNNs to computer-aided detection: architecture, dataset scale and spatial context, and when ImageNet transfer learning helps. Models range from 5 thousand to 160 million parameters. Studying lymph node detection and interstitial lung disease classification, the authors reach state-of-the-art mediastinal lymph node detection and report first five-fold cross-validation results for ILD.
Based on: Deep Convolutional Neural Networks for Computer-Aided Detection: CNN Architectures, Dataset Characteristics and Transfer Learning · IEEE Transactions on Medical Imaging
Curated by Aramai Editorial
Read summary →