Georg Wimmer

Scale invariant texture descriptors for classifying celiac disease

Graphical abstract Highlights ► We test several approaches for the computer assisted diagnosis of celiac disease. ► Only scale invariant techniques are considered. ► The scale invariance of the approaches is explicitly assessed. ► Some of the methods improve the state of the art in detecting celiac disease. ► The approaches are distinctly less scale invariant than expected.

Exploring Deep Learning and Transfer Learning for Colonic Polyp Classification

Recently, Deep Learning, especially through Convolutional Neural Networks (CNNs) has been widely used to enable the extraction of highly representative features. This is done among the network layers by filtering, selecting, and using these features in the last fully connected layers for pattern classification. However, CNN training for automated endoscopic image classification still provides a challenge due to the lack of large and publicly available annotated databases. In this work we explore Deep Learning for the automated classification of colonic polyps using different configurations for training CNNs from scratch (or full training) and distinct architectures of pretrained CNNs tested on 8-HD-endoscopic image databases acquired using different modalities. We compare our results with some commonly used features for colonic polyp classification and the good results suggest that features learned by CNNs trained from scratch and the “off-the-shelf” CNNs features can be highly relevant for automated classification of colonic polyps. Moreover, we also show that the combination of classical features and “off-the-shelf” CNNs features can be a good approach to further improve the results.

A Novel Shape Feature Descriptor for the Classification of Polyps in HD Colonoscopy

This work proposes a new method analyzing the shape of connected components (blobs) from segmented images for the classification of colonic polyps. The segmentation algorithm is a novel variation of the fast level lines transform and the resultant blobs are ideal to model the pit pattern structure of the mucosa. The shape of the blobs is described by a mixture of new features (convex hull, skeletonization and perimeter) as well as already proven features (contrast feature). We show that shape features of blobs extracted by segmenting an image are particularly suitable for mucosal texture classification and outperforming commonly used feature extraction methods.

Directional wavelet based features for colonic polyp classification

In this work, various wavelet based methods like the discrete wavelet transform, the dual-tree complex wavelet transform, the Gabor wavelet transform, curvelets, contourlets and shearlets are applied for the automated classification of colonic polyps. The methods are tested on 8 HD-endoscopic image databases, where each database is acquired using different imaging modalities (Pentaxs i-Scan technology combined with or without staining the mucosa), 2 NBI high-magnification databases and one database with chromoscopy high-magnification images. To evaluate the suitability of the wavelet based methods with respect to the classification of colonic polyps, the classification performances of 3 wavelet transforms and the more recent curvelets, contourlets and shearlets are compared using a common framework. Wavelet transforms were already often and successfully applied to the classification of colonic polyps, whereas curvelets, contourlets and shearlets have not been used for this purpose so far. We apply different feature extraction techniques to extract the information of the subbands of the wavelet based methods. Most of the in total 25 approaches were already published in different texture classification contexts. Thus, the aim is also to assess and compare their classification performance using a common framework. Three of the 25 approaches are novel. These three approaches extract Weibull features from the subbands of curvelets, contourlets and shearlets. Additionally, 5 state-of-the-art non wavelet based methods are applied to our databases so that we can compare their results with those of the wavelet based methods. It turned out that extracting Weibull distribution parameters from the subband coefficients generally leads to high classification results, especially for the dual-tree complex wavelet transform, the Gabor wavelet transform and the Shearlet transform. These three wavelet based transforms in combination with Weibull features even outperform the state-of-the-art methods on most of the databases. We will also show that the Weibull distribution is better suited to model the subband coefficient distribution than other commonly used probability distributions like the Gaussian distribution and the generalized Gaussian distribution. So this work gives a reasonable summary of wavelet based methods for colonic polyp classification and the huge amount of endoscopic polyp databases used for our experiments assures a high significance of the achieved results.

Local fractal dimension based approaches for colonic polyp classification

This work introduces texture analysis methods that are based on computing the local fractal dimension (LFD; or also called the local density function) and applies them for colonic polyp classification. The methods are tested on 8 HD-endoscopic image databases, where each database is acquired using different imaging modalities (Pentaxs i-Scan technology combined with or without staining the mucosa) and on a zoom-endoscopic image database using narrow band imaging. In this paper, we present three novel extensions to a LFD based approach. These extensions additionally extract shape and/or gradient information of the image to enhance the discriminativity of the original approach. To compare the results of the LFD based approaches with the results of other approaches, five state of the art approaches for colonic polyp classification are applied to the employed databases. Experiments show that LFD based approaches are well suited for colonic polyp classification, especially the three proposed extensions. The three proposed extensions are the best performing methods or at least among the best performing methods for each of the employed databases. The methods are additionally tested by means of a public texture image database, the UIUCtex database. With this database, the viewpoint invariance of the methods is assessed, an important features for the employed endoscopic image databases. Results imply that most of the LFD based methods are more viewpoint invariant than the other methods. However, the shape, size and orientation adapted LFD approaches (which are especially designed to enhance the viewpoint invariance) are in general not more viewpoint invariant than the other LFD based approaches.

Shape and size adapted local fractal dimension for the classification of polyps in HD colonoscopy

This work proposes a new method for computing the local fractal dimension for the classification of colonic polyps. First an image is segmented by an algorithm based on the idea of the watershed transform. The resultant connected components (blobs) show the local mucosal structure at local minima and maxima in the image and model the pit pattern structure of the mucosa. The local fractal dimension is computed using two different filter masks, an anisotropic Gaussian filter mask and an elliptic binary filter mask, which are especially adapted to the shapes and sizes of the blobs. By specifically fitting shapes and sizes of the filter masks for each blob, our feature is scale, orientation and viewpoint invariant. The proposed method outperforms other methods commonly used for mucosal texture classification.

Complex Wavelet Transform variants and Discrete Cosine Transform for scale invariance in magnification-endoscopy image classification

In this paper, scale invariant features are extracted from different variants of the Dual-Tree Complex Wavelet Transform (DT-CWT) in order to classify high-magnification colon endoscopy imagery with respect to the pit pattern scheme. To enhance the scale invariance, the Discrete Cosine Transform is applied to the feature vectors, that are achieved from a DT-CWT variant. The feature vectors either consist of the means and standard deviations of the subbands from a DTC-WT variant or of the Weibull parameter of these subbands. Superior results as compared to techniques described previously in literature are reported.

CNN transfer learning for the automated diagnosis of celiac disease

In this work, four well known convolutional neural networks (CNNs) that were pretrained on the ImageNet database are applied for the computer assisted diagnosis of celiac disease based on endoscopic images of the duodenum. The images are classified using three different transfer learning strategies and a experimental setup specifically adapted for the classification of endoscopic imagery. The CNNs are either used as fixed feature extractors without any fine-tuning to our endoscopic celiac disease image database or they are fine-tuned by training either all layers of the CNN or by fine-tuning only the fully connected layers. Classification is performed by the CNN SoftMax classifier as well as linear support vector machines. The CNN results are compared with the results of four state-of-the-art image representations. We will show that fine-tuning all the layers of the nets achieves the best results and outperforms the comparison approaches.

On the effects of de-interlacing on the classification accuracy of interlaced endoscopic videos with indication for celiac disease

Interlaced scanning is a technique that has been widely in use to double the perceived frame rate without increasing the used bandwidth. Interlaced scanning is still in use by endoscopic video hardware today. Towards the development of an automated decision support system we focus on the evaluation of the impact of de-interlacing techniques on the accuracy of automated classification of endoscopic video data with indication for celiac disease. In a large experimental setup a variety of de-interlacing methods are evaluated using a set of feature extraction methods from the fields of pattern recognition and medical image analysis.

Customised frequency pre-filtering in a local binary pattern-based classification of gastrointestinal images

Local Binary Patterns (LBP) is a widely used approach for medical image analysis. Limitations of the LBP operator are its sensitivity to noise and its boundedness to first derivative information. These limitations are usually balanced by extensions of the classical LBP operator (e.g. the Local Ternary Pattern operator (LTP) or the Extended LBP (ELBP) operator). In this paper we present a generic framework that is able to overcome this limitations by frequency filtering the images as pre-processing stage to the classical LBP. The advantage of this approach is its easier adaption and optimization to different application scenarios and data sets as compared to other LBP variants. Experiments are carried out employing two endoscopic data sets, the first from the duodenum used for diagnosis of celiac disease, the second from the colon used for polyp malignity assessment. It turned out that high pass filtering combined with LBP outperforms classical LBP and most of its extensions, whereas low pass filtering effects the results only to a small extent.