Investigation of Convolutional Neural Network Architectures for Image-based Feature Learning and Classification

Abstract

Convolutional Neural Networks (CNN) are useful methods for identification of previously unknown embedded patterns in images. Several object and facial recognition along with image segmentation tasks have benefited from the non-linear abstraction of hybrid features using CNN. This work presents a novel CNN model parametrization work-flow developed on the cloud-computing platform of Microsoft Azure Machine Learning Studio (MAMLS) that is capable of learning from the feature maps and classifying multi-modal images with different variabilities using one common flow. This two-step work-flow trains CNN models by splitting the data into training and testing data sets. First, the CNN layers are fixed and the best kernel and normalization parameters that maximize classification accuracy on the test data are identified using grid search. Next, using the best kernel and normalization parameters, the best CNN architecture that maximizes classification accuracy is detected. Finally, the activated feature maps (AFMs) from the parameterized CNN model so far are analyzed to learn new features that can enhance image-based classification accuracies. The proposed flow achieves classification accuracies in the range of 92.5-99.2% that can be further enhanced by doubling the samples based on the features learned from the AFMs. The proposed non-deep CNN models in the MAMLS platform are capable of processing image data sets with 400-4 million samples using a common flow without exponential increase in the computation time. Thus, parametrized non-deep CNN models using the proposed method are capable of identifying novel features that may enhance image-based classification accuracies. For computed tomography (CT) images, quantitative image assessment can allow for benchmarking image processing methods and optimization of image acquisition parameters. Large volumes of CT images from phantoms and patients are analyzed using the CNN models compare to a baseline model that vary in their implementation time complexities. The goal here is to model the data set variability for prediction of CT image quality (CTIQ). We observe that for 70% of data samples in training and 30% data sample in test set, respectively, the average multi-class classification accuracies for CTIQ prediction vary significantly as the data sets are switched from the phantom to patient images. The CNN model is found to be more suitable for CT image texture classification in the absence of structural variabilities. Our analysis demonstrates that CNN models are consistent identifiers of structural similarities for CT image data sets. Future work on multi-objective CNN modeling and 3D CNN modeling may lead to new insights for classification tasks.