Juyeon Park
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Juyeon Park has a total of 10 co-authors. Here are the top co-authors by number of shared publications:
- Tal Lifshitz
Department of Physics Korea Advanced Institute of Science and Technology (KAIST) Daejeon South Korea
2 shared publications - YongKeun Park
Department of Physics Korea Advanced Institute of Science and Technology (KAIST) Daejeon Republic of Korea
2 shared publications - Eun Kyung Lee
Department of Internal Medicine, National Cancer Center, Goyang, Republic of Korea
1 shared publication Geon Kim
Department of Physics Korea Advanced Institute of Science and Technology (KAIST) Daejeon South Korea
1 shared publicationHayoung Lee
Department of Pathology National Cancer Center Goyang South Korea
1 shared publication- Kwang Suk Lee
Department of Urology, Gangnam Severance Hospital, Yonsei University College of Medicine, Seoul 06273, Republic of Korea
1 shared publication
Latest Publications
- Open Access
- 17 June 2026
- Published by Wiley in Cytometry Part A
- Vol. 109 (7) , 500-512
Refractive Index–Correlated Pseudocoloring for Adaptive Color Fusion in Holotomographic Cytology
Conventional bright‐field (BF) cytology of thyroid fine‐needle aspiration biopsy (FNAB) suffers from staining variability and limited subcellular contrast. Here, we present a refractive index–correlated pseudocoloring (RICP) framework that integrates quantitative refractive index (RI) maps obtained by holotomography (HT) with color BF images to enhance subcellular visualization and image interpretability. A digital micromirror device (DMD)–based HT system combined with an RGB LED illumination module enabled sequential acquisition of RI tomograms and BF images from Papanicolaou (PAP)‐stained thyroid samples over the same field of view. The RICP algorithm adaptively embeds RI‐derived structural information into a complementary hue channel while preserving the overall BF appearance. Applied to benign and malignant thyroid clusters, RICP highlighted subcellular features such as nucleoli, lipid droplets, and nuclear irregularities. Hue–saturation analysis revealed differences between benign and malignant clusters, while complementary image‐fusion metrics and delentropy analysis both indicated systematic changes in structural image content after fusion. An ablation analysis further showed that the RI‐derived input contributes structural content beyond that obtained when the RI map is replaced by an inverted Value channel of the BF image in HSV space within the same pseudocoloring framework. These results indicate that RICP provides a perceptually grounded approach for correlative multimodal visualization that bridges conventional color cytology and quantitative optical imaging.
- 17 June 2026
- Published by Wiley in Cytometry Part A
- Vol. 109 (7) , 500-512
Open AccessRefractive Index–Correlated Pseudocoloring for Adaptive Color Fusion in Holotomographic Cytology
Conventional bright‐field (BF) cytology of thyroid fine‐needle aspiration biopsy (FNAB) suffers from staining variability and limited subcellular contrast. Here, we present a refractive index–correlated pseudocoloring (RICP) framework that integrates quantitative refractive index (RI) maps obtained by holotomography (HT) with color BF images to enhance subcellular visualization and image interpretability. A digital micromirror device (DMD)–based HT system combined with an RGB LED illumination module enabled sequential acquisition of RI tomograms and BF images from Papanicolaou (PAP)‐stained thyroid samples over the same field of view. The RICP algorithm adaptively embeds RI‐derived structural information into a complementary hue channel while preserving the overall BF appearance. Applied to benign and malignant thyroid clusters, RICP highlighted subcellular features such as nucleoli, lipid droplets, and nuclear irregularities. Hue–saturation analysis revealed differences between benign and malignant clusters, while complementary image‐fusion metrics and delentropy analysis both indicated systematic changes in structural image content after fusion. An ablation analysis further showed that the RI‐derived input contributes structural content beyond that obtained when the RI map is replaced by an inverted Value channel of the BF image in HSV space within the same pseudocoloring framework. These results indicate that RICP provides a perceptually grounded approach for correlative multimodal visualization that bridges conventional color cytology and quantitative optical imaging.
- Open Access
- 27 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-8
Label-Free 3D Virtual Staining of Urine Cytology Using Holotomography and GAN-Based Deep Learning
Chemical staining is essential for the enhancement of cellular contrast in urine cytology, but it is labor-intensive and incompatible with live-cell imaging. Virtual staining offers a digital alternative, yet prior research has focused mainly on histology, with little attention to cytology and none to 3D Papanicolaou (Pap) staining. We present an unsupervised deep-learning framework that generates 2D and 3D Pap stains directly from label-free holotomography (HT). The model, based on CycleGAN augmented with a structural similarity (SSIM) loss, was trained on unpaired 2D HT and Pap-stained whole-slide images. It achieved a mean SSIM of 0.714 ± 0.097 while preserving cellular morphology. We extend virtual staining to volumetric urine cytology, generating 3D virtual Pap images that reduce cell overlap and improve structural clarity. Additionally, the model performed inference in 1.01 seconds per whole-slide image and 65.5 seconds per 3D volume on a consumer GPU, demonstrating real-time feasibility. This establishes a label-free and scalable framework for volumetric digital cytopathology, with potential for high-throughput workflows and automation.
- 27 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 1-8
Open AccessLabel-Free 3D Virtual Staining of Urine Cytology Using Holotomography and GAN-Based Deep Learning
Chemical staining is essential for the enhancement of cellular contrast in urine cytology, but it is labor-intensive and incompatible with live-cell imaging. Virtual staining offers a digital alternative, yet prior research has focused mainly on histology, with little attention to cytology and none to 3D Papanicolaou (Pap) staining. We present an unsupervised deep-learning framework that generates 2D and 3D Pap stains directly from label-free holotomography (HT). The model, based on CycleGAN augmented with a structural similarity (SSIM) loss, was trained on unpaired 2D HT and Pap-stained whole-slide images. It achieved a mean SSIM of 0.714 ± 0.097 while preserving cellular morphology. We extend virtual staining to volumetric urine cytology, generating 3D virtual Pap images that reduce cell overlap and improve structural clarity. Additionally, the model performed inference in 1.01 seconds per whole-slide image and 65.5 seconds per 3D volume on a consumer GPU, demonstrating real-time feasibility. This establishes a label-free and scalable framework for volumetric digital cytopathology, with potential for high-throughput workflows and automation.