Gland segmentation in colon histology images using hand-crafted features and convolutional neural networks

Abstract

We investigate glandular structure segmentation in colon histology images as a window-based classification problem. We compare and combine methods based on fine-tuned convolutional neural networks (CNN) and hand-crafted features with support vector machines (HC-SVM). On 85 images of H&E-stained tissue, we find that fine-tuned CNN outperforms HC-SVM in gland segmentation measured by pixel-wise Jaccard and Dice indices. For HC-SVM we further observe that training a second-level window classifier on the posterior probabilities -as an output refinement -can substantially improve the segmentation performance. The final performance of HC-SVM with refinement is comparable to that of CNN. Furthermore, we show that by combining and refining the posterior probability outputs of CNN and HC-SVM together, a further performance boost is obtained.

Gland Segmentation in Histopathology Images Using Deep Networks and Handcrafted Features

“…A pre-processing module extracts appropriate input channels, namely the red channel and Hematoxylin component, as well as some handcrafted features. In some works RGB space is used for gland segmentation [3,4,6]. However, in this work we show that only red channel and the Hematoxylin component are more informative and lead to better segmentation.…”

“…Since structures of glands are different in various types of diseases, hence in histopathology images, structural information is insufficient for segmentation [3]. Some recent works use neural networks and deep features alongside handcrafted features for segmentation of the glands [4], [5], [6]. Chen et al [3] proposed a deep contour-aware network to segment contours and gland objects simultaneously.…”

Gland Instance Segmentation Using Deep Multichannel Neural Networks

“…These methods perform well in benign images but are comparatively unsatisfactory when used on malignant images, which has been the impetus for creating methods based on deep learning [27]. Li et al [30] train a window-based binary classifier to segment glands using both CNN features and hand-crafted features. Kainz et al [26] train two separated networks to recognize glands and glandseparating structures respectively.…”

Gland Segmentation in Histopathological Images by Deep Neural Network

“…Monivannan et al [4] compute some hand-crafted features such as multi-resolution local binary patterns (LBP) and SIFT for sliding windows of the image, which are then clustered into 30 groups by K-means and classified by SVM. In a subsequent work [5], authors add deep features to achieve improved results. Deep features are extracted from a pre-trained fully convolutional neural network.…”