A fast and effective detection framework for whole-slide histopathology image analysis

Abstract

Pathologists generally pan, focus, zoom and scan tissue biopsies either under microscopes or on digital images for diagnosis. With the rapid development of whole-slide digital scanners for histopathology, computer-assisted digital pathology image analysis has attracted increasing clinical attention. Thus, the working style of pathologists is also beginning to change. Computer-assisted image analysis systems have been developed to help pathologists perform basic examinations. This paper presents a novel lightweight detection framework for automatic tumor detection in whole-slide histopathology images. We develop the Double Magnification Combination (DMC) classifier, which is a modified DenseNet-40 to make patch-level predictions with only 0.3 million parameters. To improve the detection performance of multiple instances, we propose an improved adaptive sampling method with superpixel segmentation and introduce a new heuristic factor, local sampling density, as the convergence condition of iterations. In postprocessing, we use a CNN model with 4 convolutional layers to regulate the patch-level predictions based on the predictions of adjacent sampling points and use linear interpolation to generate a tumor probability heatmap. The entire framework was trained and validated using the dataset from the Camelyon16 Grand Challenge and Hubei Cancer Hospital. In our experiments, the average AUC was 0.95 in the test set for pixel-level detection.

Whole Slide Image Understanding in Pathology: What Is the Salient Scale of Analysis?

“…Ruan et al [28] first used a fixed-level threshold segmentation method to remove background from the WSIs. Patches were sampled using a novel adaptive sampling method at both the 20× and 40× magnification levels and were chosen to be 256 × 256 pixels.…”

Region of Interest Identification in the Cervical Digital Histology Images

“…Although RoI localization and identification is a well-known problem in analyzing histological images, there is still no general approach to all kinds of histology images. Much recent progress on an interpretation of the histology slide images has benefited a lot from the deep learning modes [10,15,17,18,19,20,21], including adversarial neural networks [17], trained on labeled data for basal membrane segmentation to detect cancer micro invasions; DenseNet [18] for tumor metastasis detection as the RoIs.…”

“…The framework [18] comprises a patch-based classifier, an improved adaptive sampling method and a postprocessing filter on annotated data. A graph convolution network that admits a graph-based RoI representation [10] to incorporate local inter-patch context and, as in previous cases, RoI annotated data were used for RoI image classification.…”

Deep-learning based breast cancer detection for cross-staining histopathology images

“…DL is an extremely data-demanding technology, and the performance of a DL-based algorithm usually requires qualified ground truth data for model construction. Thousands of images could be required, especially for weakly supervised or unsupervised learning DL-model development [ 16 , 17 ]. Such data requirements are not difficult for H&E model development because a digitalized pathology lab usually generates abundant H&E images in routine diagnostic processes.…”