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

Abstract

Detection of mitotic tumor cells per tissue area is one of the critical markers of breast cancer prognosis. The aim of this paper is to develop a method for the automatic detection of mitotic figures from breast cancer histological slides using a partially supervised deep learning framework. Unlike the previous literature, which has focused on solving the problem of mitosis detection in the weakly annotated datasets using centroid pixel labels (weak labels) only without taking advantage of the available pixel-level labels (strong labels) of other datasets, in this paper, we design a novel partially supervised framework based on two parallel deep fully convolutional networks. One of them is trained using weak labels and the other is trained using strong labels, together with a weight transfer function. In the detection phase, we fuse the segmentation maps produced by the two networks to obtain the final mitosis detections. Our system exploits the available large sets of mitosis detection samples with mitosis centroid annotation, such as the 2014 ICPR dataset and the AMIDA13 dataset, and only a small set of samples with the annotation of all mitosis pixels, such as the 2012 ICPR dataset, to perform a more accurate mitosis detection on weakly labeled data. This enables us to outperform all previous mitosis detection systems by achieving F-scores of 0.575 and 0.698 on the 2014 ICPR dataset and the AMIDA13 dataset respectively.

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

“…This annotation scheme overcomes the downsides of thresholding based technique used by different researchers 37 , which resulted in many disconnected false-positive blobs and overlooked true positives. Likewise, circular labelling scheme 15 , 16 , 53 is unable to provide complete morphological information. In our work, the label-refiner improves the labels by exploiting the prediction maps generated by deep instance based segmentation model Mask R-CNN.…”

“…The defined scheme uses concentric circular labels to represent the mitotic region and proposed a concentric loss function that only considers the region inside the circle whereby it excludes the chance of non-mitotic region overlap with the mitotic region. Similarly, Sebai et al (2020) adapted the semantic segmentation model for the mitosis detection problem 16 , 54 , 53 . They handled the issue of weak labels by integrating two deep networks in an end to end manner.…”

Breast Cancer Detection Using Deep Belief Network by Applying Feature Extraction on Various Classifiers

“…In the thickly associated obstruct, all the appearance of the previous level were result as input to the last cover. The limitation is, the system taught for identifying exact abnormality since the feebly administer localization chore could only identify the lesion with a elevated growth likelihood [12].…”

“…It is the likelihood of a classification function which is forecast to outcome as true positive rate at the presence of disease. It is also recognized as true positive(TP) amount and can be compute as:( ) = +(12)…”

SmallMitosis: Small Size Mitotic Cells Detection in Breast Histopathology Images

“…In third stage, score level fusion of ResNet-50 and DenseNet model is performed to further reduce the false positives. Similarly, Sebai et al [10] proposed a partially supervised mitosis detection model using two parallel deep fully convolutional networks: one for weakly labeled datasets and another for strongly labeled datasets. In parallel connection, semantic information is transformed from weak segmentation model to strong segmentation model by a weight transfer function.…”