Few-shot Breast Cancer Metastases Classification Via Unsupervised Cell Ranking

Tumor metastases detection is of great importance for the treatment of breast cancer patients.Especially,breast cancer has the highest mortality rate in women.A large number of patients each year are treated for cancer that has metastasized from breast tissue.It is very valuable to analyze the prognostic factors of breast cancer.In the past,due to the professionalism of medical diagnosis and the strictness of diagnostic accuracy,the detection of cancer metastasis is mainly determined by the pathologists by browsing a large number of pathological sections,and there is a huge gap in this regard.With the development of machine learning,some artificial intelligence methods and natural image classification techniques have been applied to image classification and detection tasks of cancers such as breast cancer with good results,such as logistic regression,support vector machines,random forest and so on.Besides,with the development of digital pathology,image analysis shows its advantages for histopathological image analysis,which can not only speed up the diagnosis but also effectively avoid the misdiagnosis caused by the fatigue of doctors working for a long time.Recently,various CNN(Convolutional Neural Network)based methods get excellent performance in object detection/segmentation.However,the detection of metastases in hematoxylin and eosin(H&E)stained whole-slide images(WSI)is still challenging mainly due to two aspects:(1)The resolution of the image is too large;(2)lacking labeled training data.Whole-slide images generally stored in a multi-resolution structure with multiple downsampling tiles.It is difficult to feed the whole image into memory without compression.Moreover,labeling images for the pathologists are time-consuming and expensive.To deal with the data-hungry nature of supervised learning,unsupervised learning and semi-supervised learning gradually emerged.They use little or no labeled data.This alleviates the lack of data to some extent,but generally,the result is not as effective as supervised learning.In this paper,we study the problem of detecting breast cancer metastases in the pathological image on the “patch” level.To address the abovementioned challenges,we propose a few-shot learning method to classify whether an image patch contains tumor cells.Specifically,we propose a patch-level unsupervised cell ranking approach,which only relies on images with limited labels.The main idea of the proposed method is that when cropping a patch A from the WSI and further cropping a sub-patch B from A,the cell number of A is always larger than that of B.Based on this observation,we make use of the unlabeled images to learn the ranking information of cell counting to extract the abstract features.Secondly,in the image pre-processing stage,we compared with different color normalization operations.Also,different machine learning classification methods are compared to classify the features extracted from the heatmaps.Experimental results show that our method is effective to improve the patch-level classification accuracy,compared to the traditional supervised method.