Detection of Ki67 Hot-Spots of Invasive Breast Cancer Based on Convolutional Neural Networks Applied to Mutual Information of H&E and Ki67 Whole Slide Images

Abstract

Ki67 hot-spot detection and its evaluation in invasive breast cancer regions play a significant role in routine medical practice. The quantification of cellular proliferation assessed by Ki67 immunohistochemistry is an established prognostic and predictive biomarker that determines the choice of therapeutic protocols. In this paper, we present three deep learning-based approaches to automatically detect and quantify Ki67 hot-spot areas by means of the Ki67 labeling index. To this end, a dataset composed of 100 whole slide images (WSIs) belonging to 50 breast cancer cases (Ki67 and H&E WSI pairs) was used. Three methods based on CNN classification were proposed and compared to create the tumor proliferation map. The best results were obtained by applying the CNN to the mutual information acquired from the color deconvolution of both the Ki67 marker and the H&E WSIs. The overall accuracy of this approach was 95%. The agreement between the automatic Ki67 scoring and the manual analysis is promising with a Spearman’s ρ correlation of 0.92. The results illustrate the suitability of this CNN-based approach for detecting hot-spots areas of invasive breast cancer in WSI.

LYSTO: The Lymphocyte Assessment Hackathon and Benchmark Dataset

“…According to the averaged sensitivities, four out of five methods achieved better performance than the pathologist panel. The performance of TU/e and HUST is even comparable with fully-supervised method [6], [27].…”

“…This task suffers from implicit variability and tediousness when performed by pathologists, suggesting the potential value of a computer-aided system. However, despite the simple nature of this task, it has been shown recently [6] that detecting lymphocytes in IHC goes beyond simply "counting dark-brown spots". Moreover, IHC slides in daily practice contain challenging regions such as dense clusters, possibly background staining, and presence of artifacts such as ink (see examples in Figure 1).…”

Heimdall: an AI-based infrastructure for traffic monitoring and anomalies detection

“…These methods have undermined canonical feature extraction techniques, which, although they do not require much data to work acceptably, do not always guarantee high reliability and accuracy at the generalisation stage. Deep Learning is now widely used in video and image analysis in a variety of research areas, ranging from medical diagnostics [19], [20], financial forecasting [21], [22] and cybersecurity [23], and over the years has also had a positive impact on road safety [24]. The state of the art presents considerable scientific solutions to the problem of road safety, providing inspiration for the development of methods for pedestrian detection [25], vehicle tracking [26], and accident detection [27].…”

The Approach of Artificial Intelligence in Neuroendocrine Carcinomas of the Breast: A Next Step towards Precision Pathology?—A Case Report and Review of the Literature

“…To establish a final diagnosis of breast carcinoma with neuroendocrine differentiation, it is essential to perform a neuroendocrine immunohistochemical profile of the tumor using antibodies against synaptophysin or chromogranin, as well as the MIB-1 (mindbomb E3 ubiquitin protein ligase 1) proliferation index, one of the most controversial markers in breast pathology with regard to its methodology in current clinical practice. For the majority of this index, it is imperative to precisely measure an indispensable nuclear antigen for cell proliferation [4], as the methods of digitally analyzing the images are of great use [5]. Routinely, pathologists visually evaluate MIB-1 by manually counting cells using an optical microscope, even though this method lacks repeatability among observers [4,[6][7][8] and leads to limitations in its clinical application [4,9].…”