Automatic generation of the ground truth for tumor budding using H&amp;E stained slides

Tavolara, Thomas E.; Dutta, Arijit; Burks, Martin V.; Chen, Wei; Frankel, Wendy L.; Gürcan, Metin N.; Niazi, M. Khalid Khan

doi:10.1117/12.2612892

Cited by 5 publications

(13 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…AUC measurements were unavailable for Faster R-CNN, as a negative ROI or background class did not exist. As a comparison, pathologist precision and recall after the washout period is around 0.6754 ± 0.3253 and 0.3060 ± 0.2014 based on a previous study [7]. We also visualized attention weights from TB and bags to determine whether the regions AB-MIL attends to corresponded to tumor buds.…”

Section: Resultsmentioning

confidence: 99%

“…Image analysis of TB has been reported in [24][25][26][27][28][29]. In our previous method [7], adjacent AE1/3 and H&E WSIs were registered in order to provide a reliable ground truth for tumor regions and TB. Swin transformers were then utilized to segment malignant tumor regions on H&E images, similar to this study.…”

Section: Discussionmentioning

confidence: 99%

“…Bokhorst et al [5] trained convolutional neural networks (CNNs) to classify image patches from digitized CRC biopsies into three classes: TBs, containing glands, or background. Similar to our previous work, their method for generating ground truth involved first annotating TBs on AE1/3 and then visually transferring said annotations onto adjacent H&E. However, there was still considerable interobserver variability in identifying TBs on H&E and AE1/3 [6][7][8] and a high rate of false positives. Therefore, we seek a method that does not rely on detailed TB annotations and can tolerate false positives.…”

Section: Introductionmentioning

confidence: 94%

“…Initially, tumor was segmented using a previously published method [7], and removed from WSIs. As mentioned earlier, TBs were annotated by pathologists.…”

Section: Bag Creationmentioning

confidence: 99%

See 3 more Smart Citations

Minimizing the intra-pathologist disagreement for tumor bud detection on H and E images using weakly supervised learning

Tavolara

Chen

Frankel

et al. 2023

Medical Imaging 2023: Digital and Computational Pathology

View full text Add to dashboard Cite

Tumor budding (TB) is defined as a cluster of one to four tumor cells at the tumor invasive front. Though promising as a prognostic factor for colorectal cancer, its routine clinical use is hampered by high inter-and intra-observer disagreement on routine H&E staining. Pan-cytokeratin immunohistochemical staining increases agreement but is costly, non-routine, and may yield false tumor buds (false positives). This makes the development of automatic algorithms to identify TB difficult. Therefore, we propose a weakly-supervised method that does not require strictly accurate tissuelevel annotations and is resilient to false positives. Our database consists of 29 H&E whole slide images. TB and nontumor ROIs were generated by cropping 512x512 regions around annotated tumor buds and within annotated non-tumor regions, respectively. Attention-based multiple instance learning was applied to identify ROIs containing tumor buds. This resulted in a precision of 0.9477 ± 0.0516, recall of 0.9131 ± 0.0568, and AUC of 0.9482 ± 0.0679 on an external dataset. These results provide preliminary evidence for the feasibility of our method to identify tumor buds accurately.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 94%

“…Initially, tumor was segmented using a previously published method [7], and removed from WSIs. As mentioned earlier, TBs were annotated by pathologists.…”

Section: Bag Creationmentioning

confidence: 99%

See 2 more Smart Citations

Minimizing the intra-pathologist disagreement for tumor bud detection on H and E images using weakly supervised learning

Tavolara

Chen

Frankel

et al. 2023

Medical Imaging 2023: Digital and Computational Pathology

View full text Add to dashboard Cite

show abstract

“…Firstly, there is no agreed-upon method for TB identification [7], making their detection by pathologists subject to bias, which would lead to poor results when applying to test cases. Secondly, TB annotation by pathologists is time-consuming and costly, creating difficulty in gathering annotated datasets on a large-scale to train segmentation or detection models effectively [16]. Multiple previous studies tried to address this challenge by automatically detecting TBs from pan-cytokeratin stained slides [17][18][19][20] or registered H&E and pancytokeratin stained slides [21].…”

Section: Introductionmentioning

confidence: 99%

Few-shot tumor bud segmentation using generative model in colorectal carcinoma

Su,

Chen,

Leigh

et al. 2024

Medical Imaging 2024: Digital and Computational Pathology

View full text Add to dashboard Cite

Current deep learning methods in histopathology are limited by the small amount of available data and time consumption in labeling the data. Colorectal cancer (CRC) tumor budding quantification performed using H&E-stained slides is crucial for cancer staging and prognosis but is subject to labor-intensive annotation and human bias. Thus, acquiring a large-scale, fully annotated dataset for training a tumor budding (TB) segmentation/detection system is difficult. Here, we present a DatasetGAN-based approach that can generate essentially an unlimited number of images with TB masks from a moderate number of unlabeled images and a few annotated images. The images generated by our model closely resemble the real colon tissue on H&E-stained slides. We test the performance of this model by training a downstream segmentation model, UNet++, on the generated images and masks. Our results show that the trained UNet++ model can achieve reasonable TB segmentation performance, especially at the instance level. This study demonstrates the potential of developing an annotation-efficient segmentation model for automatic TB detection and quantification.

show abstract