A Novel Loss Calibration Strategy for Object Detection Networks Training on Sparsely Annotated Pathological Datasets

Li, Hansheng; Han, Xin; Kang, Yuxin; Shi, Xiaoshuang; Yan, Ming; Tong, Zixu; Bu, Qirong; Cui, Lei; Feng, Jun; Yang, Lin

doi:10.1007/978-3-030-59722-1_31

Cited by 17 publications

(17 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This study thus supported the potential applications of SSL to develop medical AI systems. In addition, we have noticed some other recent works 46,47 , which have made a new strategy on the sparse and incomplete annotations to reduce the annotation effort for cell detection. This strategy is also applicable to annotations in our WSIs, and the unlabeled data is useful for SSL.…”

Section: Discussionmentioning

confidence: 98%

Accurate recognition of colorectal cancer with semi-supervised deep learning on pathological images

Sun

et al. 2021

Nat Commun

View full text Add to dashboard Cite

Machine-assisted pathological recognition has been focused on supervised learning (SL) that suffers from a significant annotation bottleneck. We propose a semi-supervised learning (SSL) method based on the mean teacher architecture using 13,111 whole slide images of colorectal cancer from 8803 subjects from 13 independent centers. SSL (~3150 labeled, ~40,950 unlabeled; ~6300 labeled, ~37,800 unlabeled patches) performs significantly better than the SL. No significant difference is found between SSL (~6300 labeled, ~37,800 unlabeled) and SL (~44,100 labeled) at patch-level diagnoses (area under the curve (AUC): 0.980 ± 0.014 vs. 0.987 ± 0.008, P value = 0.134) and patient-level diagnoses (AUC: 0.974 ± 0.013 vs. 0.980 ± 0.010, P value = 0.117), which is close to human pathologists (average AUC: 0.969). The evaluation on 15,000 lung and 294,912 lymph node images also confirm SSL can achieve similar performance as that of SL with massive annotations. SSL dramatically reduces the annotations, which has great potential to effectively build expert-level pathological artificial intelligence platforms in practice.

show abstract

Section: Discussionmentioning

confidence: 98%

Accurate recognition of colorectal cancer with semi-supervised deep learning on pathological images

Sun

et al. 2021

Nat Commun

View full text Add to dashboard Cite

show abstract

“…The localization loss L loc is computed from the predicted location v and the ground truth location b = {X b , Y b , W b , H b } of the positive instances. For example, a typical choice of L loc is the smooth L 1 loss function [9], where…”

Section: Background: Cell Detection With Complete Annotations For Tra...mentioning

confidence: 99%

“…However, due to the large number and diverse morphology of cells in histopathology images, completely annotating all instances becomes very challenging. In practice, experts can prefer to only ensure that all instances labeled as positive are correct [10], and the annotation may even be sparse in the image if a large number of images are to be annotated [9]. In this case, the rest of the instances may not be all true negatives, and there still exist unannotated positive instances with high probability.…”

Section: Introductionmentioning

confidence: 99%

“…Very recently, the problem of incomplete annotations for training CNN-based cell detectors draws attention [9]. In [9], motivated by the significant density difference between the regression boxes of positive and negative instances, Boxes Density Energy (BDE) is proposed to calibrate the training loss. The assumption of BDE is that negative samples are inclined to possess a lower density of regression boxes.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Positive-unlabeled Learning for Cell Detection in Histopathology Images with Incomplete Annotations

Zhao¹,

Pang²,

Zhi-wen³

et al. 2021

Preprint

View full text Add to dashboard Cite

Cell detection in histopathology images is of great value in clinical practice. Convolutional neural networks (CNNs) have been applied to cell detection to improve the detection accuracy, where cell annotations are required for network training. However, due to the variety and large number of cells, complete annotations that include every cell of interest in the training images can be challenging. Usually, incomplete annotations can be achieved, where positive labeling results are carefully examined to ensure their reliability but there can be other positive instances, i.e., cells of interest, that are not included in the annotations. This annotation strategy leads to a lack of knowledge about true negative samples. Most existing methods simply treat instances that are not labeled as positive as truly negative during network training, which can adversely affect the network performance. In this work, to address the problem of incomplete annotations, we formulate the training of detection networks as a positive-unlabeled learning problem. Specifically, the classification loss in network training is revised to take into account incomplete annotations, where the terms corresponding to negative samples are approximated with the true positive samples and the other samples of which the labels are unknown. To evaluate the proposed method, experiments were performed on a publicly available dataset for mitosis detection in breast cancer cells, and the experimental results show that our method improves the performance of cell detection given incomplete annotations for training.

show abstract

“…A preliminary version of this study has been published in a conference study ( 12 ), which is only evaluated on the MITOS-ATYPIA-14 dataset. In this study, we have made significant extensions to generalize our methods on the Ki-67 dataset, aiming to provide a strong and comprehensive theory for relevant research.…”

Section: Introductionmentioning

confidence: 99%

A Robust Training Method for Pathological Cellular Detector via Spatial Loss Calibration

Kang

Yang

et al. 2021

Front. Med.

Self Cite

View full text Add to dashboard Cite

Computer-aided diagnosis of pathological images usually requires detecting and examining all positive cells for accurate diagnosis. However, cellular datasets tend to be sparsely annotated due to the challenge of annotating all the cells. However, training detectors on sparse annotations may be misled by miscalculated losses, limiting the detection performance. Thus, efficient and reliable methods for training cellular detectors on sparse annotations are in higher demand than ever. In this study, we propose a training method that utilizes regression boxes' spatial information to conduct loss calibration to reduce the miscalculated loss. Extensive experimental results show that our method can significantly boost detectors' performance trained on datasets with varying degrees of sparse annotations. Even if 90% of the annotations are missing, the performance of our method is barely affected. Furthermore, we find that the middle layers of the detector are closely related to the generalization performance. More generally, this study could elucidate the link between layers and generalization performance, provide enlightenment for future research, such as designing and applying constraint rules to specific layers according to gradient analysis to achieve “scalpel-level” model training.

show abstract

A Novel Loss Calibration Strategy for Object Detection Networks Training on Sparsely Annotated Pathological Datasets

Cited by 17 publications

References 17 publications

Accurate recognition of colorectal cancer with semi-supervised deep learning on pathological images

Accurate recognition of colorectal cancer with semi-supervised deep learning on pathological images

Positive-unlabeled Learning for Cell Detection in Histopathology Images with Incomplete Annotations

A Robust Training Method for Pathological Cellular Detector via Spatial Loss Calibration

Contact Info

Product

Resources

About