Abstract:Background and Objective: Mitotic activity is a crucial biomarker for diagnosing and predicting outcomes for different types of cancers, particularly breast cancer. However, manual mitosis counting is challenging and time-consuming for pathologists, with moderate reproducibility due to biopsy slide size, low mitotic cell density, and pattern heterogeneity. In recent years, deep learning methods based on convolutional neural networks (CNNs) have been proposed to address these limitations. Nonetheless, these methods have been hampered by the available data labels, which usually consist only of the centroids of mitosis, and by the incoming noise from annotated hard negatives. As a result, complex algorithms with multiple stages are often required to refine the labels at the pixel level and reduce the number of false positives. Methods: This article presents a novel weakly supervised approach for mitosis detection that utilizes only image-level labels on histological hematoxylin and eosin (H&E) images, avoiding the need for complex labeling scenarios. Also, an Uninformed Teacher-Student (UTS) pipeline is introduced to detect and distill hard samples by comparing weakly supervised localizations and the annotated centroids, using strong augmentations to enhance uncertainty. Additionally, an automatic proliferation score is proposed that mimicks the pathologist-annotated mitotic activity index (MAI). The proposed approach is evaluated on three publicly available datasets for mitosis detection on breast histology samples, and two datasets for mitotic activity counting in whole-slide images. Results: The proposed framework achieves competitive performance with relevant prior literature in all the datasets used for evaluation without explicitly using the mitosis location information during training. This approach challenges previous methods that rely on strong mitosis location information and multiple stages to refine false positives. Furthermore, the proposed pipeline for hard-sample distillation demonstrates promising dataset-specific improvements. Concretely, when the annotation has not been thoroughly refined by multiple pathologists, the UTS model offers improvements of up to ∼4% in mitosis localization, thanks to the detection and distillation of uncertain cases. Concerning the mitosis counting task, the proposed automatic proliferation score shows a moderate positive correlation with the MAI annotated by pathologists at the biopsy level on two external datasets. Conclusions: The proposed Uninformed Teacher-Student pipeline leverages strong augmentations to distill uncertain samples and measure dissimilarities between predicted and annotated mitosis. Results demonstrate the feasibility of the weakly supervised approach and highlight its potential as an objective evaluation tool for tumor proliferation.

Leveraging Unlabeled Whole-Slide-Images for Mitosis Detection

PartMitosis: A Partially Supervised Deep Learning Framework for Mitosis Detection in Breast Cancer Histopathology Images

DeepMitosis: Mitosis Detection Via Deep Detection, Verification and Segmentation Networks.

Strongly Supervised Mitosis Detection in Breast Histopathology Images Using Weak Labels.

Uninformed Teacher-Student for hard-samples distillation in weakly supervised mitosis localization

A Fuzzy Segmentation Method to Learn Classification of Mitosis

Weakly Supervised Mitosis Detection in Breast Histopathology Images Using Concentric Loss.

MaskMitosis: a Deep Learning Framework for Fully Supervised, Weakly Supervised, and Unsupervised Mitosis Detection in Histopathology Images

Mitosis Detection, Fast and Slow: Robust and Efficient Detection of Mitotic Figures

A Novel Two-Stage Deep Method for Mitosis Detection in Breast Cancer Histology Images

Assessment of algorithms for mitosis detection in breast cancer histopathology images

Multi-Scale Deep Neural Network for Mitosis Detection in Histological Images

Mitosis Detection Under Limited Annotation: A Joint Learning Approach

SmallMitosis: Small Size Mitotic Cells Detection in Breast Histopathology Images

A multi-phase deep CNN based mitosis detection framework for breast cancer histopathological images

A generalizable and robust deep learning algorithm for mitosis detection in multicenter breast histopathological images

Toward Automatic Mitotic Cell Detection and Segmentation in Multispectral Histopathological Images

OMG-Net: A Deep Learning Framework Deploying Segment Anything to Detect Pan-Cancer Mitotic Figures from Haematoxylin and Eosin-Stained Slides

Low-Cost Histopathological Mitosis Detection for Microscope-acquired Images

Detecting Cancer Metastases on Gigapixel Pathology Images

MITNET: a novel dataset and a two-stage deep learning approach for mitosis recognition in whole slide images of breast cancer tissue