GlandVision: A Novel Polar Space Random Field Model for Glandular Biological Structure Detection

Abstract

In this paper, we propose a novel method for detecting glandular structures in microscopic images of human tissue. We first transform the image from Cartesian space to polar space and introduce a novel random field model with an efficient inference strategy that uses two simple chain graphs to approximate a circular graph to infer possible boundary of a gland. We then develop a visual feature based support vector regressor (SVR) to verify if the inferred contour corresponds to a true gland. And finally, we combine the outputs of the random field and the regressor to form the GlandVision algorithm for the detection of glandular structures. In the experiments, we treat the task of detecting glandular structures as object (gland) proposal, detection and segmentation problems respectively and show that our new technique outperforms state of the art computer vision algorithms developed for generic objects.

A Novel Polar Space Random Field Model for the Detection of Glandular Structures

“…14 clearly demonstrates that the sum of node potential outperforms the original random field output, and the learned SVM score obtains the best result. It achieves an ALC (srea to the left of the curve) of 0.68 which is much higher than its counterparts, and this number is also much higher than our previously reported result of 0.59 in [29]. From the figure, we can also see that the random field model achieves a high detection rate of 0.96, which means only 21 out of 522 glands could not be detected by this model.…”

A Comprehensive Review for MRF and CRF Approaches in Pathology Image Analysis

“…In [79], a novel method is presented for detecting glandular structures in microscopic images of human colon tissues, where the images are transformed from Cartesian space to polar space first, then a CRF model (shown in Fig. 11) is introduced to infer possible boundary of a gland and a visual feature based support vector regressor (SVR) is developed to verify whether the inferred contour corresponds to a true gland, finally the outputs of these two methods in the second step are combined to form the GlandVision algorithm ranking all the potential contours, and this generates the final results.…”

Dual Adaptive Pyramid Network for Cross-Stain Histopathology Image Segmentation

“…In our experiments, the dataset is separated into training and test sets with 85 and 80 images respectively. GlandVision dataset [1] contains 20 DAB-H stained colon images with size of 1280 × 1024, which were captured with 10× optical magnification. We randomly select 14 images for training and the rest for test.…”