Learning histopathological patterns

Abstract

Aims:The aim was to demonstrate a method for automated image analysis of immunohistochemically stained tissue samples for extracting features that correlate with patient disease. We address the problem of quantifying tumor tissue and segmenting and counting cell nuclei.Materials and Methods:Our method utilizes a flexible segmentation method based on sparse coding trained from representative image samples. Nuclei counting is based on a nucleus model that takes size, shape, and nucleus probability into account. Nuclei clustering and overlays are resolved using a gray-weighted distance transform. We obtain a probability measure for pixels belonging to a nucleus from our segmentation procedure. Experiments are carried out on two sets of immunohistochemically stained images – one set based on the estrogen receptor (ER) and the other on antigen KI-67. For the nuclei separation we have selected 207 ER image samples from 58 tissue micro array-cores corresponding to 58 patients and 136 KI-67 image samples also from 58 cores. The images are hand-annotated by marking the center position of each nucleus. For the ER data we have a total of 1006 nuclei and for the KI-67 we have 796 nuclei. Segmentation performance was evaluated in terms of missing nuclei, falsely detected nuclei, and multiple detections. The proposed method is compared to state-of-the-art Bayesian classification.Statistical analysis used:The performance of the proposed method and a state-of-the-art algorithm including variations thereof is compared using the Wilcoxon rank sum test.Results:For both the ER experiment and the KI-67 experiment the proposed method exhibits lower error rates than the state-of-the-art method. Total error rates were 4.8 % and 7.7 % in the two experiments, corresponding to an average of 0.23 and 0.45 errors per image, respectively. The Wilcoxon rank sum tests show statistically significant improvements over the state-of-the-art method.Conclusions:We have demonstrated a method and obtained good performance compared to state-of-the-art nuclei separation. The segmentation procedure is simple, highly flexible, and we demonstrate how it, in addition to the nuclei separation, can perform precise segmentation of cancerous tissue. The complexity of the segmentation procedure is linear in the image size and the nuclei separation is linear in the number of nuclei. Additionally the method can be parallelized to obtain high-speed computations.

Skin Blood Perfusion and Cellular Response to Insertion of Insulin Pen Needles With Different Diameters

“…19 The magnitude of tissue trauma can be evaluated based on manual stereology of histological slides, through semiquantitative methods, or by automatic analysis of digital images [20][21][22] where stains are used to detect specific cells of interest. For example, analyzing the presence of white blood cells (WBCs) in the tissue is one commonly used inflammation marker.…”

IHC Color Histograms for Unsupervised Ki67 Proliferation Index Calculation

“…Color deconvolution (CD) proposed in Ruifrok and Johnston (2001) has dominated digital pathology image analysis as a pre-processing step for automated stain separation. It has been applied to Ki67 analysis for stain deconvolution and PI estimation (Kårsnäs et al, 2011; Shi et al, 2016; Mungle et al, 2017). CD is dependent on the Beer-Lambert (BL) law of absorption (Ruifrok and Johnston, 2001; Macenko et al, 2009) which characterizes each pure stain by an optical density (OD) vector of light in the red, green, and blue (RGB) intensity channels (Kårsnäs et al, 2011).…”

“…Using 10-fold cross validation with 50–50 training and test splits, the highest accuracy of 98% was reported. In Kårsnäs et al (2011) an iterative dictionary learning approach was used for classification of tissue into probability maps of each stain type. The learning algorithm automatically updated the dictionary using the training images for final implementation on the test images.…”

High-throughput histopathological image analysis via robust cell segmentation and hashing

“…This method can handle touching cases by selecting the best matched model parameters. Another learning based nucleus segmentation is presented in (Kårsnäs et al, 2011), where intensity and label dictionaries are constructed to separate the foreground from the background and then touching nuclei are split by combining region merging with a marker-controlled watershed. Probabilistic models have also attracted research interests.…”