Bioimage classification with subcategory discriminant transform of high dimensional visual descriptors

Feature transforms for image data augmentation

A problem with convolutional neural networks (CNNs) is that they require large datasets to obtain adequate robustness; on small datasets, they are prone to overfitting. Many methods have been proposed to overcome this shortcoming with CNNs. In cases where additional samples cannot easily be collected, a common approach is to generate more data points from existing data using an augmentation technique. In image classification, many augmentation approaches utilize simple image manipulation algorithms. In this work, we propose some new methods for data augmentation based on several image transformations: the Fourier transform (FT), the Radon transform (RT), and the discrete cosine transform (DCT). These and other data augmentation methods are considered in order to quantify their effectiveness in creating ensembles of neural networks. The novelty of this research is to consider different strategies for data augmentation to generate training sets from which to train several classifiers which are combined into an ensemble. Specifically, the idea is to create an ensemble based on a kind of bagging of the training set, where each model is trained on a different training set obtained by augmenting the original training set with different approaches. We build ensembles on the data level by adding images generated by combining fourteen augmentation approaches, with three based on FT, RT, and DCT, proposed here for the first time. Pretrained ResNet50 networks are finetuned on training sets that include images derived from each augmentation method. These networks and several fusions are evaluated and compared across eleven benchmarks. Results show that building ensembles on the data level by combining different data augmentation methods produce classifiers that not only compete competitively against the state-of-the-art but often surpass the best approaches reported in the literature.

Data representation using robust nonnegative matrix factorization for edge computing

As a popular data representation technique, Nonnegative matrix factorization (NMF) has been widely applied in edge computing, information retrieval and pattern recognition. Although it can learn parts-based data representations, existing NMF-based algorithms fail to integrate local and global structures of data to steer matrix factorization. Meanwhile, semi-supervised ones ignore the important role of instances from different classes in learning the representation. To solve such an issue, we propose a novel semi-supervised NMF approach via joint graph regularization and constraint propagation for edge computing, called robust constrained nonnegative matrix factorization (RCNMF), which learns robust discriminative representations by leveraging the power of both L2, 1-norm NMF and constraint propagation. Specifically, RCNMF explicitly exploits global and local structures of data to make latent representations of instances involved by the same class closer and those of instances involved by different classes farther. Furthermore, RCNMF introduces the L2, 1-norm cost function for addressing the problems of noise and outliers. Moreover, L2, 1-norm constraints on the factorial matrix are used to ensure the new representation sparse in rows. Finally, we exploit an optimization algorithm to solve the proposed framework. The convergence of such an optimization algorithm has been proven theoretically and empirically. Empirical experiments show that the proposed RCNMF is superior to other state-of-the-art algorithms.

Efficient Regularization Framework for Histopathological Image Classification Using Convolutional Neural Networks

Deep learning methods are characterized by their capacity to learn data representation compared to the traditional machine learning algorithms. However, these methods are prone to overfitting on small volumes of data. The objective of this research is to overcome this limitation by improving the generalization in the proposed deep learning framework based on various techniques: data augmentation, small models, optimizer selection, and ensemble learning. For ensembling, the authors used selected models from different checkpoints and both voting and unweighted average methods for combination. The experimental study on the lymphomas histopathological dataset highlights the efficiency of the MobileNet2 network combined with the stochastic gradient descent (SGD) optimizer in terms of generalization. The best results have been achieved by the combination of the best three checkpoint models (98.67% of accuracy). These findings provide important insights into the efficiency of the checkpoint ensemble learning method for histopathological image classification.

Ensemble of convolutional neural networks for bioimage classification

This work presents a system based on an ensemble of Convolutional Neural Networks (CNNs) and descriptors for bioimage classification that has been validated on different datasets of color images. The proposed system represents a very simple yet effective way of boosting the performance of trained CNNs by composing multiple CNNs into an ensemble and combining scores by sum rule. Several types of ensembles are considered, with different CNN topologies along with different learning parameter sets. The proposed system not only exhibits strong discriminative power but also generalizes well over multiple datasets thanks to the combination of multiple descriptors based on different feature types, both learned and handcrafted. Separate classifiers are trained for each descriptor, and the entire set of classifiers is combined by sum rule. Results show that the proposed system obtains state-of-the-art performance across four different bioimage and medical datasets. The MATLAB code of the descriptors will be available at https://github.com/LorisNanni.

NHL Pathological Image Classification Based on Hierarchical Local Information and GoogLeNet-Based Representations

Background

Accurate classification for different non-Hodgkin lymphomas (NHL) is one of the main challenges in clinical pathological diagnosis due to its intrinsic complexity. Therefore, this paper proposes an effective classification model for three types of NHL pathological images, including mantle cell lymphoma (MCL), follicular lymphoma (FL), and chronic lymphocytic leukemia (CLL).

Methods

There are three main parts with respect to our model. First, NHL pathological images stained by hematoxylin and eosin (H&E) are transferred into blue ratio (BR) and Lab spaces, respectively. Then specific patch-level textural and statistical features are extracted from BR images and color features are obtained from Lab images both using a hierarchical way, yielding a set of hand-crafted representations corresponding to different image spaces. A random forest classifier is subsequently trained for patch-level classification. Second, H&E images are cropped and fed into a pretrained google inception net (GoogLeNet) for learning high-level representations and a softmax classifier is used for patch-level classification. Finally, three image-level classification strategies based on patch-level results are discussed including a novel method for calculating the weighted sum of patch results. Different classification results are fused at both feature 1 and image levels to obtain a more satisfactory result.

Results

The proposed model is evaluated on a public IICBU Malignant Lymphoma Dataset and achieves an improved overall accuracy of 0.991 and area under the receiver operating characteristic curve of 0.998.

Conclusion

The experimentations demonstrate the significantly increased classification performance of the proposed model, indicating that it is a suitable classification approach for NHL pathological images.

Lymphoma images analysis using morphological and non-morphological descriptors for classification

Mantle cell lymphoma, follicular lymphoma and chronic lymphocytic leukemia are the principle subtypes of the non-Hodgkin lymphomas. The diversity of clinical presentations and the histopathological features have made diagnosis of such lymphomas a complex task for specialists. Computer aided diagnosis systems employ computational vision and image processing techniques, which contribute toward the detection, diagnosis and prognosis of digitised images of histological samples. Studies aimed at improving the understanding of morphological and non-morphological features for classification of lymphoma remain a challenge in this area. This work presents a new approach for the classification of information extracted from morphological and non-morphological features of nuclei of lymphoma images. We extracted data of the RGB model of the image and employed contrast limited adaptive histogram equalisation and 2D order-statistics filter methods to enhance the contrast and remove noise. The regions of interest were identified by the global thresholding method. The flood-fill and watershed techniques were used to remove the small false positive regions. The area, extent, perimeter, convex area, solidity, eccentricity, equivalent diameter, minor axis and major axis measurements were computed for the regions detected in the nuclei. In the feature selection stage, we applied the ANOVA, Ansari-Bradley and Wilcoxon rank sum methods. Finally, we employed the polynomial, support vector machine, random forest and decision tree classifiers in order to evaluate the performance of the proposed approach. The non-morphological features achieved higher AUC and AC values for all cases: the obtained rates were between 95% and 100%. These results are relevant, especially when we consider the difficulties of clinical practice in distinguishing the studied groups. The proposed approach is useful as an automated protocol for the diagnosis of lymphoma histological tissue.

Bioimage Classification with Handcrafted and Learned Features

Bioimage classification is increasingly becoming more important in many biological studies including those that require accurate cell phenotype recognition, subcellular localization, and histopathological classification. In this paper, we present a new General Purpose (GenP) bioimage classification method that can be applied to a large range of classification problems. The GenP system we propose is an ensemble that combines multiple texture features (both handcrafted and learned descriptors) for superior and generalizable discriminative power. Our ensemble obtains a boosting of performance by combining local features, dense sampling features, and deep learning features. Each descriptor is used to train a different Support Vector Machine that is then combined by sum rule. We evaluate our method on a diverse set of bioimage classification tasks each represented by a benchmark database, including some of those available in the IICBU 2008 database. Each bioimage classification task represents a typical subcellular, cellular, and tissue level classification problem. Our evaluation on these datasets demonstrates that the proposed GenP bioimage ensemble obtains state-of-the-art performance without any ad-hoc dataset tuning of the parameters (thereby avoiding any risk of overfitting/overtraining). To reproduce the experiments reported in this paper, the MATLAB code of all the descriptors is available at https://github.com/LorisNanni and https://www.dropbox.com/s/bguw035yrqz0pwp/ElencoCode.docx?dl=0.