Deep segmentation networks generally consist of an encoder to extract features from an input image and a decoder to restore them to the original input size to produce segmentation results. In an ideal setting, the trained encoder should possess the semantic embedding capability, which maps a pair of features close to each other when they belong to the same class, and maps them distantly if they correspond to different classes. Recent deep segmentation networks do not directly deal with the embedding behavior of the encoder. Accordingly, we cannot expect that the features embedded by the encoder will have the semantic embedding property. If the model can be trained to have the embedding ability, it will further enhance the performance as restoring from those features is much easier for the decoder. To this end, we propose supervised contrastive embedding, which employs feature-wise contrastive loss for the feature map to enhance the segmentation performance on medical images. We also introduce a boundary-aware sampling strategy, which focuses on the features corresponding to image patches located at the boundary area of the ground-truth annotations. Through extensive experiments on lung segmentation in chest radiographs, liver segmentation in computed tomography, and brain tumor and spinal cord gray matter segmentation in magnetic resonance images, it is demonstrated that the proposed method helps to improve the segmentation performance of popular U-Net, U-Net++, and DeepLabV3+ architectures. Furthermore, it is confirmed that the robustness on domain shifts can be enhanced for segmentation models by the proposed contrastive embedding.

Accurately segmenting cancer lesions is essential for effective personalized treatment and enhanced patient outcomes. We propose a multi-resolution selective segmentation (MurSS) model to accurately segment breast cancer lesions from hematoxylin and eosin (H&E) stained whole-slide images (WSIs). We used The Cancer Genome Atlas breast invasive carcinoma (BRCA) public dataset for training and validation. We used the Korea University Medical Center, Guro Hospital, BRCA dataset for the final test evaluation. MurSS utilizes both low- and high-resolution patches to leverage multi-resolution features using adaptive instance normalization. This enhances segmentation performance while employing a selective segmentation method to automatically reject ambiguous tissue regions, ensuring stable training. MurSS rejects 5% of WSI regions and achieves a pixel-level accuracy of 96.88% (95% confidence interval (CI): 95.97–97.62%) and mean Intersection over Union of 0.7283 (95% CI: 0.6865–0.7640). In our study, MurSS exhibits superior performance over other deep learning models, showcasing its ability to reject ambiguous areas identified by expert annotations while using multi-resolution inputs.

Introduction: To analyze how the histomorphological features of adenocarcinoma lesions are related to the survival of patients, we developed a deep learning based pancreatic adenocarcinoma survival model that predicts death risk through the general adenocarcinoma feature extractor (GAFE). Our pancreatic adenocarcinoma survival model and GAFE was trained with hematoxylin and eosin (H&E) stained whole slide images (WSIs). Also, to measure the generalization performance, we predict the death risk of rectum adenocarcinoma and breast adenocarcinoma using the pancreatic adenocarcinoma survival model. Design: The whole data analyzed in this study was from The Cancer Genome Atlas (TCGA). We used a self-supervised contrastive learning method to train GAFE with randomly chosen 328 adenocarcinoma H&E stained WSIs. These WSIs are consisted of 92 lung (TCGA-LUAD), 100 colon (TCGA-COAD), 97 prostate (TCGA-PRAD) and 39 stomach (TCGA-STAD) adenocarcinoma WSIs. For the pancreatic adenocarcinoma survival model, we used 179 WSIs of TCGA pancreatic adenocarcinoma (TCGA-PAAD) dataset labeled with survival events and periods. There existed 66 uncensored data, and we used 5-fold cross validation. Each WSI was divided into 256*256 pixel-wise patches. Feature vectors were extracted from the patches using the GAFE. Then the attention-based multi-instance learning was applied to train the survival model. The trained model returned a death risk for each patient. We used 122 adenocarcinoma H&E stained WSIs of rectum (TCGA-READ) and 999 of breast (TCGA-BRCA) from TCGA to test the generalization performance for different organs. Result: We used 5-fold cross validation for testing the pancreatic adenocarcinoma survival model. The mean value of C-index from the TCGA-PAAD test datasets was 0.7258 where the highest and the lowest C-index was 0.7784 and 0.6598. Using the model with median performance (C-index: 0.7216), we measured death risk and divided patients into 50% of high-risk and 50% of low-risk groups. From this, the p-value was 0.01986 for the log-rank test.To analyze the performance of the generalization, we ensembled 5 models from the above. We tested our ensembled model on TCGA-READ and TCGA-BRCA. The C-index for TCGA-Read was 0.6941 where the C-index for TCGA-BRCA was 0.5711. Conclusion: We found that adenocarcinoma’s histomorphological features had some correlation with pancreatic adenocarcinoma survival prediction. Moreover, we could see the possibility of transferability of adenocarcinoma histomorphological knowledge and survival analysis to other organs such as rectum adenocarcinoma and breast adenocarcinoma. Citation Format: Joonho Lee, Geongyu Lee, Tae-Yeong Kwak, Sun Woo Kim, Hyeyoon Chang. A deep learning based pancreatic adenocarcinoma survival prediction model applicable to adenocarcinoma of other organs [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 5060.

Accurate prediction of the likelihood of recurrence is important in the selection of postoperative treatment for patients with early-stage breast cancer. In this study, we investigated whether deep learning algorithms can predict patients' risk of recurrence by analyzing the pathology images of their cancer histology.We analyzed 125 hematoxylin and eosin-stained whole slide images (WSIs) from 125 patients across two institutions (National Cancer Center and Korea University Medical Center Guro Hospital) to predict breast cancer recurrence risk using deep learning. Sensitivity reached 0.857, 0.746, and 0.529 for low, intermediate, and high-risk categories, respectively, with specificity of 0.816, 0.803, and 0.972, and a Pearson correlation of 0.61 with histological grade. Class activation maps highlighted features like tubule formation and mitotic rate, suggesting a cost-effective approach to risk stratification, pending broader validation. These findings suggest that deep learning models trained exclusively on hematoxylin and eosin stained whole slide images can approximate genomic assay results, offering a cost-effective and scalable tool for breast cancer recurrence risk assessment. However, further validation using larger and more balanced datasets is needed to confirm the clinical applicability of our approach.

The RS (Recurrence Score) is the results of a 21-gene test assessing the likelihood of breast cancer recurrence for hormone-positive, HER2-negative, and node-negative breast cancer patients. We conducted an experiment to find out whether there is a correlation between the RS value derived through genetic testing and specific morphological characteristics of breast cancer tissues with the expectation that these morphological features could predict RS values. Deep learning was utilized to assess the morphology of the hematoxylin-eosin (H&E) stained whole slide images (WSIs) of breast cancer. In this study, we used 125 breast cancer cases associated with H&E stained WSIs and RS scores which were divided into 3 risk groups based on the RS values: the low-risk group with 49 cases, the intermediate-risk group with 59 cases, and the high-risk group with 17 cases (low-risk for RS < 18, high-risk for RS > 31, and intermediate-risk for RS in between). On each WSI, an expert pathologist has annotated invasive breast tumor regions. We divided WSIs into 512x512 patches and trained a deep CNN model to classify the patches as benign (including DCIS), low-risk, intermediate-risk, or high-risk. In determining the risk group of each WSI, all patches from the WSI were classified by the model first, and then their majority class was assigned to the WSI. A nested cross validation was adopted to evaluate how well the model generalizes. The performance of the algorithm is shown in Table 1. According to the results, our model was able to distinguish between the low-risk group and the high-risk group to some extent. In particular, there was no case of misclassifying the low-risk group as high-risk. Since these high-risk patients require radiological examination, there is an advantage that the 21-gene test is not required. Although our model has yielded modest results, we believe that future studies through the integration of additional data with other clinical and pathological data will yield better results. Citation Format: Geongyu Lee, Chungyeul Kim, Tae-Yeong Kwak, Sun Woo Kim, Hyeyoon Chang. Recurrence risk prediction based on automatic histopathologic analysis of breast cancer using whole slide images [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 5058.

We propose the Multi-resolution Selective Segmentation model (MurSS) for segmenting benign, Ductal Carcinoma In Situ, and Invasive Ductal Carcinoma in breast resection Hematoxylin and Eosin stained Whole Slide Images. MurSS simultaneously trains on context information from a wide area at low resolution and content information from a local area at high resolution, aiming for a more accurate diagnosis. Additionally, through the selection stage, it provides solutions for ambiguous tissue regions. Our proposed MurSS achieves a mean Intersection of Union performance of 91.1%, which is at least 16.8% and at most 19.0% higher than well-known image segmentation models.