Sanghoon Lee
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Sanghoon Lee has a total of 82 co-authors. Here are the top co-authors by number of shared publications:
Brian P. Pollack
9 shared publications- David A. Gutman
Emory University, Department of Pathology, GA, USA
9 shared publications Habiba Elfandy
9 shared publicationsHagar Hussein
9 shared publications- Lee Alex Donald Cooper
Department of Pathology, Northwestern University Feinberg School of Medicine, Chicago, IL, USA
9 shared publications Matt McCormick
9 shared publications
Latest Publications
- Open Access
- 23 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 305-306
A Study of EEG-to-Text Translation Using Convolutional Encoding and Multi-View Transformer Modeling
- 23 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 305-306
Open AccessA Study of EEG-to-Text Translation Using Convolutional Encoding and Multi-View Transformer Modeling
- Open Access
- 15 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 125275-125286
Task-Ready PanNuke and NuCLS Datasets: Reorganization, Synthetic Data Generation, and Experimental Evaluation
Automating nuclei analysis in histopathology is essential for enhancing disease diagnosis; however, training reliable models necessitates well-structured datasets. This paper addresses the gap in standardized data preparation workflows for two critical histopathology datasets: PanNuke and NuCLS. First, we organize histopathology images and masks into training-validation splits, extract subsets specific to cell types, and generate multi-scale patches to enable robust model training across various resolutions using the PanNuke dataset. Second, we curate task-specific subsets for object detection and semantic segmentation, ensuring consistency across splits while addressing annotation inconsistencies with the NuCLS dataset. Third, we conduct experiments on two reorganized datasets, including cell-type-specific binary classification, multi-task evaluation, and extension to synthetic datasets. Our workflows address common histopathology data challenges, including fragmented annotations, class imbalance, and mismatched metadata. The processed datasets are shared in standardized formats, allowing researchers to train models directly for critical tasks such as detecting cancerous nuclei or segmenting inflammatory cells in histopathology images.
- 15 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 125275-125286
Open AccessTask-Ready PanNuke and NuCLS Datasets: Reorganization, Synthetic Data Generation, and Experimental Evaluation
Automating nuclei analysis in histopathology is essential for enhancing disease diagnosis; however, training reliable models necessitates well-structured datasets. This paper addresses the gap in standardized data preparation workflows for two critical histopathology datasets: PanNuke and NuCLS. First, we organize histopathology images and masks into training-validation splits, extract subsets specific to cell types, and generate multi-scale patches to enable robust model training across various resolutions using the PanNuke dataset. Second, we curate task-specific subsets for object detection and semantic segmentation, ensuring consistency across splits while addressing annotation inconsistencies with the NuCLS dataset. Third, we conduct experiments on two reorganized datasets, including cell-type-specific binary classification, multi-task evaluation, and extension to synthetic datasets. Our workflows address common histopathology data challenges, including fragmented annotations, class imbalance, and mismatched metadata. The processed datasets are shared in standardized formats, allowing researchers to train models directly for critical tasks such as detecting cancerous nuclei or segmenting inflammatory cells in histopathology images.
- Open Access
- 4 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 92764-92773
SN-FPN: Self-Attention Nested Feature Pyramid Network for Digital Pathology Image Segmentation
Digital pathology has played a key role in replacing glass slides with digital images, enhancing various pathology workflows. Whole slide images are digitized pathological images improving the capabilities of digital pathology and contributing to the overall turnaround time for diagnoses. The digitized images have been successfully integrated with artificial intelligence algorithms assisting pathologists in many tasks, but there are still demands to develop a new algorithm for a better diagnosis process. In this paper, we propose a new deep convolutional neural network model integrating a feature pyramid network with a self-attention mechanism in three pathways: encoder, decoder, and self-attention nested for providing accurate tumor region segmentation on whole slide images. The encoder pathway adopts ResNet50 architecture for the bottom-up network. The decoder pathway adopts the feature pyramid network for the top-down network. The self-attention nested pathway forms the attention map represented by the distribution of attention scores focusing on localizing tumor regions and avoiding irrelevant information. The results of our experiment show that the proposed model outperforms the state-of-the-art deep convolutional neural network models in terms of tumor and stromal region segmentation. Moreover, various encoder networks were equipped with the proposed model and compared with each other. The results indicate that the ResNet series using the proposed model outperforms other encoder networks.
- 4 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 92764-92773
Open AccessSN-FPN: Self-Attention Nested Feature Pyramid Network for Digital Pathology Image Segmentation
Digital pathology has played a key role in replacing glass slides with digital images, enhancing various pathology workflows. Whole slide images are digitized pathological images improving the capabilities of digital pathology and contributing to the overall turnaround time for diagnoses. The digitized images have been successfully integrated with artificial intelligence algorithms assisting pathologists in many tasks, but there are still demands to develop a new algorithm for a better diagnosis process. In this paper, we propose a new deep convolutional neural network model integrating a feature pyramid network with a self-attention mechanism in three pathways: encoder, decoder, and self-attention nested for providing accurate tumor region segmentation on whole slide images. The encoder pathway adopts ResNet50 architecture for the bottom-up network. The decoder pathway adopts the feature pyramid network for the top-down network. The self-attention nested pathway forms the attention map represented by the distribution of attention scores focusing on localizing tumor regions and avoiding irrelevant information. The results of our experiment show that the proposed model outperforms the state-of-the-art deep convolutional neural network models in terms of tumor and stromal region segmentation. Moreover, various encoder networks were equipped with the proposed model and compared with each other. The results indicate that the ResNet series using the proposed model outperforms other encoder networks.
- Open Access
SN-FPN: Self-attention Nested Feature Pyramid Network for Digital Pathology Image Segmentation
Digital pathology has played a key role in replacing glass slides with digital images, enhancing various pathology workflows. Whole slide images are digitized pathological images improving the capabilities of digital pathology and contributing to the overall turn-around time for diagnoses. The digitized images have been successfully integrated with artificial intelligence algorithms assisting pathologists in many tasks, but there are still demands to develop a new algorithm for a better diagnosis process. In this paper, we propose a new deep convolutional neural network model integrating a feature pyramid network with a self-attention mechanism in three pathways: encoder, decoder, and sel-attention nested for providing accurate tumor region segmentation on whole slide images. The encoder pathway adopts ResNet50 architecture for the bottom-up network. The decoder pathway adopts the feature pyramid network for the top-down network. The self-attention nested pathway forms the attention map represented by the distribution of attention scores focusing on localizing tumor regions and avoiding irrelevant information. The experiment results show that the proposed model outperforms the state-of-the-art deep convolutional neural network models in terms of tumor and stromal region segmentation. Moreover, various encoder networks were equipped with the proposed model and compared with each other. The results indicate that the ResNet series using the proposed model outperforms other encoder networks. CCS CONCEPTS Computing methodologies → Image segmentation. ACM Reference Format Sanghoon Lee. 2024. SN-FPN: Self-attention Nested Feature Pyramid Network for Digital Pathology Image Segmentation. In 2024 ACM Southeast Conference (ACMSE 2024), April 18–20, 2024, Marietta, GA, USA. ACM, New York, NY, USA, 8 pages. https://doi.org/10.1145/3564746.35YYYYY
Open AccessSN-FPN: Self-attention Nested Feature Pyramid Network for Digital Pathology Image Segmentation
Digital pathology has played a key role in replacing glass slides with digital images, enhancing various pathology workflows. Whole slide images are digitized pathological images improving the capabilities of digital pathology and contributing to the overall turn-around time for diagnoses. The digitized images have been successfully integrated with artificial intelligence algorithms assisting pathologists in many tasks, but there are still demands to develop a new algorithm for a better diagnosis process. In this paper, we propose a new deep convolutional neural network model integrating a feature pyramid network with a self-attention mechanism in three pathways: encoder, decoder, and sel-attention nested for providing accurate tumor region segmentation on whole slide images. The encoder pathway adopts ResNet50 architecture for the bottom-up network. The decoder pathway adopts the feature pyramid network for the top-down network. The self-attention nested pathway forms the attention map represented by the distribution of attention scores focusing on localizing tumor regions and avoiding irrelevant information. The experiment results show that the proposed model outperforms the state-of-the-art deep convolutional neural network models in terms of tumor and stromal region segmentation. Moreover, various encoder networks were equipped with the proposed model and compared with each other. The results indicate that the ResNet series using the proposed model outperforms other encoder networks. CCS CONCEPTS Computing methodologies → Image segmentation. ACM Reference Format Sanghoon Lee. 2024. SN-FPN: Self-attention Nested Feature Pyramid Network for Digital Pathology Image Segmentation. In 2024 ACM Southeast Conference (ACMSE 2024), April 18–20, 2024, Marietta, GA, USA. ACM, New York, NY, USA, 8 pages. https://doi.org/10.1145/3564746.35YYYYY
- Open Access
- 18 April 2024
- Published by Association for Computing Machinery (ACM)
- p. 121-128
Comprehensive Experiments on Breast Cancer Hematoxylin and Eosin-stained Images Using UNet
The current challenge in breast cancer segmentation on hematoxylin and eosin-stained images lies in accurately capturing histologic phenotypes associated with cancer biomarkers. Many researchers have studied histologic patterns of cancerous regions of breast cancer images for a better understanding of diagnosis and treatment, yet it is not fully investigated because of its variability, complexity, and large data volumes. Comprehensive experiments on breast cancer segmentation can address this challenge by identifying heterogeneous cell regions in the tumor microenvironment and investigating a new methodology of segmenting histologic images by using advanced deep learning architectures. In this paper, we present findings in three experiments on breast cancer segmentation, exploring the effectiveness of a convolutional neural network called UNet identifying multiple heterogeneous cells such as blood, blood vessels, fat, glandular secretions, necrosis, and plasma cells in a tumor microenvironment, investigating the performance of UNet architecture in different sizes of cancerous regions such as a tumor, and tumor-infiltrating lymphocytes, and proposing a new methodology of a neural-style guided data augmentation focusing on the image segmentation of breast tumor-related nuclei in hematoxylin and eosin-stained images. The experiment results show that a modified UNet performs well on fat identification. Another experiment shows that the smaller input size of the tumor, stroma, and tumor-infiltrating lymphocyte provides better performance than the larger input size. Moreover, we demonstrate that the proposed method outperforms the traditional deep neural network models.
- 18 April 2024
- Published by Association for Computing Machinery (ACM)
- p. 121-128
Open AccessComprehensive Experiments on Breast Cancer Hematoxylin and Eosin-stained Images Using UNet
The current challenge in breast cancer segmentation on hematoxylin and eosin-stained images lies in accurately capturing histologic phenotypes associated with cancer biomarkers. Many researchers have studied histologic patterns of cancerous regions of breast cancer images for a better understanding of diagnosis and treatment, yet it is not fully investigated because of its variability, complexity, and large data volumes. Comprehensive experiments on breast cancer segmentation can address this challenge by identifying heterogeneous cell regions in the tumor microenvironment and investigating a new methodology of segmenting histologic images by using advanced deep learning architectures. In this paper, we present findings in three experiments on breast cancer segmentation, exploring the effectiveness of a convolutional neural network called UNet identifying multiple heterogeneous cells such as blood, blood vessels, fat, glandular secretions, necrosis, and plasma cells in a tumor microenvironment, investigating the performance of UNet architecture in different sizes of cancerous regions such as a tumor, and tumor-infiltrating lymphocytes, and proposing a new methodology of a neural-style guided data augmentation focusing on the image segmentation of breast tumor-related nuclei in hematoxylin and eosin-stained images. The experiment results show that a modified UNet performs well on fat identification. Another experiment shows that the smaller input size of the tumor, stroma, and tumor-infiltrating lymphocyte provides better performance than the larger input size. Moreover, we demonstrate that the proposed method outperforms the traditional deep neural network models.
- Open Access
- 31 March 2023
- Published by American Association for Cancer Research (AACR)
Data from Interactive Classification of Whole-Slide Imaging Data for Cancer Researchers
Whole-slide histology images contain information that is valuable for clinical and basic science investigations of cancer but extracting quantitative measurements from these images is challenging for researchers who are not image analysis specialists. In this article, we describe HistomicsML2, a software tool for learn-by-example training of machine learning classifiers for histologic patterns in whole-slide images. This tool improves training efficiency and classifier performance by guiding users to the most informative training examples for labeling and can be used to develop classifiers for prospective application or as a rapid annotation tool that is adaptable to different cancer types. HistomicsML2 runs as a containerized server application that provides web-based user interfaces for classifier training, validation, exporting inference results, and collaborative review, and that can be deployed on GPU servers or cloud platforms. We demonstrate the utility of this tool by using it to classify tumor-infiltrating lymphocytes in breast carcinoma and cutaneous melanoma.Significance:An interactive machine learning tool for analyzing digital pathology images enables cancer researchers to apply this tool to measure histologic patterns for clinical and basic science studies.
- 31 March 2023
- Published by American Association for Cancer Research (AACR)
Open AccessData from Interactive Classification of Whole-Slide Imaging Data for Cancer Researchers
Whole-slide histology images contain information that is valuable for clinical and basic science investigations of cancer but extracting quantitative measurements from these images is challenging for researchers who are not image analysis specialists. In this article, we describe HistomicsML2, a software tool for learn-by-example training of machine learning classifiers for histologic patterns in whole-slide images. This tool improves training efficiency and classifier performance by guiding users to the most informative training examples for labeling and can be used to develop classifiers for prospective application or as a rapid annotation tool that is adaptable to different cancer types. HistomicsML2 runs as a containerized server application that provides web-based user interfaces for classifier training, validation, exporting inference results, and collaborative review, and that can be deployed on GPU servers or cloud platforms. We demonstrate the utility of this tool by using it to classify tumor-infiltrating lymphocytes in breast carcinoma and cutaneous melanoma.Significance:An interactive machine learning tool for analyzing digital pathology images enables cancer researchers to apply this tool to measure histologic patterns for clinical and basic science studies.
- Open Access
- 31 March 2023
- Published by American Association for Cancer Research (AACR)
Video 1 from Interactive Classification of Whole-Slide Imaging Data for Cancer Researchers
- 31 March 2023
- Published by American Association for Cancer Research (AACR)
Open AccessVideo 1 from Interactive Classification of Whole-Slide Imaging Data for Cancer Researchers
- Open Access
- 31 March 2023
- Published by American Association for Cancer Research (AACR)
Supplementary Tables from Interactive Classification of Whole-Slide Imaging Data for Cancer Researchers
- 31 March 2023
- Published by American Association for Cancer Research (AACR)
Open AccessSupplementary Tables from Interactive Classification of Whole-Slide Imaging Data for Cancer Researchers
- Open Access
- 31 March 2023
- Published by American Association for Cancer Research (AACR)
Supplementary Data from Interactive Classification of Whole-Slide Imaging Data for Cancer Researchers
Supplementary figures and table legends.
- 31 March 2023
- Published by American Association for Cancer Research (AACR)
Open AccessSupplementary Data from Interactive Classification of Whole-Slide Imaging Data for Cancer Researchers
Supplementary figures and table legends.
- Open Access
- 31 March 2023
- Published by American Association for Cancer Research (AACR)
Data from Phase Ib Dose Expansion and Translational Analyses of Olaparib in Combination with Capivasertib in Recurrent Endometrial, Triple-Negative Breast, and Ovarian Cancer
Purpose:On the basis of strong preclinical rationale, we sought to confirm recommended phase II dose (RP2D) for olaparib, a PARP inhibitor, combined with the AKT inhibitor capivasertib and assess molecular markers of response and resistance.Patients and Methods:We performed a safety lead-in followed by expansion in endometrial, triple-negative breast, ovarian, fallopian tube, or peritoneal cancer. Olaparib 300 mg orally twice daily and capivasertib orally twice daily on a 4-day on 3-day off schedule was evaluated. Two dose levels (DL) of capivasertib were planned: 400 mg (DL1) and 320 mg (DL-1). Patients underwent biopsies at baseline and 28 days.Results:A total of 38 patients were enrolled. Seven (18%) had germline BRCA1/2 mutations. The first 2 patients on DL1 experienced dose-limiting toxicities (DLT) of diarrhea and vomiting. No DLTs were observed on DL-1 (n = 6); therefore, DL1 was reexplored (n = 6) with no DLTs, confirming DL1 as RP2D. Most common treatment-related grade 3/4 adverse events were anemia (23.7%) and leukopenia (10.5%). Of 32 evaluable subjects, 6 (19%) had partial response (PR); PR rate was 44.4% in endometrial cancer. Seven (22%) additional patients had stable disease greater than 4 months. Tumor analysis demonstrated strong correlations between response and immune activity, cell-cycle alterations, and DNA damage response. Therapy resistance was associated with receptor tyrosine kinase and RAS-MAPK pathway activity, metabolism, and epigenetics.Conclusions:The combination of olaparib and capivasertib is associated to no serious adverse events and demonstrates durable activity in ovarian, endometrial, and breast cancers, with promising responses in endometrial cancer. Importantly, tumor samples acquired pre- and on-therapy can help predict patient benefit.
- 31 March 2023
- Published by American Association for Cancer Research (AACR)
Open AccessData from Phase Ib Dose Expansion and Translational Analyses of Olaparib in Combination with Capivasertib in Recurrent Endometrial, Triple-Negative Breast, and Ovarian Cancer
Purpose:On the basis of strong preclinical rationale, we sought to confirm recommended phase II dose (RP2D) for olaparib, a PARP inhibitor, combined with the AKT inhibitor capivasertib and assess molecular markers of response and resistance.Patients and Methods:We performed a safety lead-in followed by expansion in endometrial, triple-negative breast, ovarian, fallopian tube, or peritoneal cancer. Olaparib 300 mg orally twice daily and capivasertib orally twice daily on a 4-day on 3-day off schedule was evaluated. Two dose levels (DL) of capivasertib were planned: 400 mg (DL1) and 320 mg (DL-1). Patients underwent biopsies at baseline and 28 days.Results:A total of 38 patients were enrolled. Seven (18%) had germline BRCA1/2 mutations. The first 2 patients on DL1 experienced dose-limiting toxicities (DLT) of diarrhea and vomiting. No DLTs were observed on DL-1 (n = 6); therefore, DL1 was reexplored (n = 6) with no DLTs, confirming DL1 as RP2D. Most common treatment-related grade 3/4 adverse events were anemia (23.7%) and leukopenia (10.5%). Of 32 evaluable subjects, 6 (19%) had partial response (PR); PR rate was 44.4% in endometrial cancer. Seven (22%) additional patients had stable disease greater than 4 months. Tumor analysis demonstrated strong correlations between response and immune activity, cell-cycle alterations, and DNA damage response. Therapy resistance was associated with receptor tyrosine kinase and RAS-MAPK pathway activity, metabolism, and epigenetics.Conclusions:The combination of olaparib and capivasertib is associated to no serious adverse events and demonstrates durable activity in ovarian, endometrial, and breast cancers, with promising responses in endometrial cancer. Importantly, tumor samples acquired pre- and on-therapy can help predict patient benefit.