Fangfang Dong
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
Fangfang Dong has a total of 28 co-authors. Here are the top co-authors by number of shared publications:
Bailin Yang
School of Computer and Information Engineering, Zhejiang Gongshang University, Hangzhou, 310018, China
2 shared publicationsDan Long
Zhejiang Cancer Hospital, Hangzhou Institute of Medicine (HIM), Chinese Academy of Sciences, Hangzhou, China
2 shared publicationsChanghuo Cen
State Key Laboratory of Stem Cell and Reproductive Biology, Institute of Zoology, Chinese Academy of Sciences
1 shared publicationChangrui Zhu
School of Computer and Information Engineering, Zhejiang Gongshang University, Hangzhou, Zhejiang, China
1 shared publicationChen Zhu
Affiliation School of Statistics and Data Science, Zhejiang Gongshang University, Hangzhou, Zhejiang, China
1 shared publicationChenying Guo
School of Statistics and Mathematics, Zhejiang Gongshang University, Hangzhou, 310018, China
1 shared publication
Latest Publications
- Open Access
- 19 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 79536-79556
Nuclei Instance Segmentation Based on Generative Data Augmentation by an Improved CycleGAN
The segmentation of cell nuclei in pathological images is a critical yet highly challenging task. Accurate segmentation enables the extraction of key nuclear features, such as morphological characteristics, which are essential for cancer diagnosis. Although fully supervised cell nucleus segmentation methods have achieved impressive results, the requirement on a large-scale instance labeled data for model training has limited their further development. Due to the complexity of pathological images, manual annotation of cell nuclei is both time-consuming and labor-intensive, making the acquisition of extensive labeled datasets impractical. To address this issue, we propose an effective generative model for pathological images based on the CycleGAN framework to enhance the quantity and diversity of labeled data. Specifically, we introduce the spatially adaptive de-normalization(SPADE) module and a multi-component discriminator to ensure better alignment between the generation conditions and the resulted pathological images. Additionally, we incorporate a cross-attention mechanism to integrate random noise into the generative process, thereby increasing the diversity of generated images. To further improve image quality, we employ a correction network. Beyond data generation, we also improve the classic HoVer-Net segmentation model by adding a new branch for nuclei center detection and nuclei counting. Experimental results demonstrate that our generative model, trained on limited annotated data, not only achieves superior generation performance but also significantly improves segmentation accuracy. Furthermore, the improved segmentation network outperforms some existing methods in nucleus segmentation tasks.
- 19 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 79536-79556
Open AccessNuclei Instance Segmentation Based on Generative Data Augmentation by an Improved CycleGAN
The segmentation of cell nuclei in pathological images is a critical yet highly challenging task. Accurate segmentation enables the extraction of key nuclear features, such as morphological characteristics, which are essential for cancer diagnosis. Although fully supervised cell nucleus segmentation methods have achieved impressive results, the requirement on a large-scale instance labeled data for model training has limited their further development. Due to the complexity of pathological images, manual annotation of cell nuclei is both time-consuming and labor-intensive, making the acquisition of extensive labeled datasets impractical. To address this issue, we propose an effective generative model for pathological images based on the CycleGAN framework to enhance the quantity and diversity of labeled data. Specifically, we introduce the spatially adaptive de-normalization(SPADE) module and a multi-component discriminator to ensure better alignment between the generation conditions and the resulted pathological images. Additionally, we incorporate a cross-attention mechanism to integrate random noise into the generative process, thereby increasing the diversity of generated images. To further improve image quality, we employ a correction network. Beyond data generation, we also improve the classic HoVer-Net segmentation model by adding a new branch for nuclei center detection and nuclei counting. Experimental results demonstrate that our generative model, trained on limited annotated data, not only achieves superior generation performance but also significantly improves segmentation accuracy. Furthermore, the improved segmentation network outperforms some existing methods in nucleus segmentation tasks.
- Open Access
- 5 May 2026
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 21 (5) , e0347870
Intelligent diagnosis of major depressive disorder with edge convolution and contrastive learning
Based on functional connectivity (FC) matrices derived from resting-state functional magnetic resonance imaging (rs-fMRI) data, graph neural networks (GNNs), as an advanced deep learning technique, have been widely applied in major depressive disorder (MDD) diagnosis. However, conventional GNNs suffer from a critical limitation in preserving the spatial specificity of brain regions, which is attributed to their intrinsic node permutation invariance that ignores the unique order and specific roles of brain regions in neural circuits. To address this limitation, in this paper, we propose a novel deep learning framework, Graph Contrastive Learning based on Edge Convolution (EC-GCL), to analyze resting-state fMRI data from 1,160 participants, including 597 patients with MDD and 563 healthy controls. This framework integrates an edge convolution encoder, specifically designed to preserve the spatial specificity of brain regions, with a learnable graph augmentation module into an adversarial graph contrastive learning, thereby enhancing the extraction of discriminative MDD-related FC features and improving diagnostic classification accuracy. Compared with conventional machine learning and GNN models, our proposed EC-GCL achieved superior performance (AUC = 71.2%) and improved interpretability. In particular, the framework identified several key brain regions, including the dorsolateral superior frontal gyrus, thalamus, and insula, that are closely linked to the pathophysiology of MDD, which is consistent with the findings of prior neuroimaging studies. This study demonstrates that combining edge convolution with contrastive learning provides a robust and explainable method for MDD diagnosis. This provides new insights into depression and may support improvements in clinical practice.
- 5 May 2026
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 21 (5) , e0347870
Open AccessIntelligent diagnosis of major depressive disorder with edge convolution and contrastive learning
Based on functional connectivity (FC) matrices derived from resting-state functional magnetic resonance imaging (rs-fMRI) data, graph neural networks (GNNs), as an advanced deep learning technique, have been widely applied in major depressive disorder (MDD) diagnosis. However, conventional GNNs suffer from a critical limitation in preserving the spatial specificity of brain regions, which is attributed to their intrinsic node permutation invariance that ignores the unique order and specific roles of brain regions in neural circuits. To address this limitation, in this paper, we propose a novel deep learning framework, Graph Contrastive Learning based on Edge Convolution (EC-GCL), to analyze resting-state fMRI data from 1,160 participants, including 597 patients with MDD and 563 healthy controls. This framework integrates an edge convolution encoder, specifically designed to preserve the spatial specificity of brain regions, with a learnable graph augmentation module into an adversarial graph contrastive learning, thereby enhancing the extraction of discriminative MDD-related FC features and improving diagnostic classification accuracy. Compared with conventional machine learning and GNN models, our proposed EC-GCL achieved superior performance (AUC = 71.2%) and improved interpretability. In particular, the framework identified several key brain regions, including the dorsolateral superior frontal gyrus, thalamus, and insula, that are closely linked to the pathophysiology of MDD, which is consistent with the findings of prior neuroimaging studies. This study demonstrates that combining edge convolution with contrastive learning provides a robust and explainable method for MDD diagnosis. This provides new insights into depression and may support improvements in clinical practice.
- 20 May 2022
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 147, 105651
CRAUNet: A cascaded residual attention U-Net for retinal vessel segmentation
- 20 May 2022
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 147, 105651
CRAUNet: A cascaded residual attention U-Net for retinal vessel segmentation
- Open Access
- 27 August 2021
- Published by eLife Sciences Publications, Ltd in eLife
- Vol. 10
PRMT5 regulates ovarian follicle development by facilitating Wt1 translation
Protein arginine methyltransferase 5 (Prmt5) is the major type II enzyme responsible for symmetric dimethylation of arginine. Here, we found that PRMT5 was expressed at high level in ovarian granulosa cells of growing follicles. Inactivation of Prmt5 in granulosa cells resulted in aberrant follicle development and female infertility. In Prmt5-knockout mice, follicle development was arrested with disorganized granulosa cells in which WT1 expression was dramatically reduced and the expression of steroidogenesis-related genes was significantly increased. The premature differentiated granulosa cells were detached from oocytes and follicle structure was disrupted. Mechanism studies revealed that Wt1 expression was regulated by PRMT5 at the protein level. PRMT5 facilitated IRES-dependent translation of Wt1 mRNA by methylating HnRNPA1. Moreover, the upregulation of steroidogenic genes in Prmt5-deficient granulosa cells was repressed by Wt1 overexpression. These results demonstrate that PRMT5 participates in granulosa cell lineage maintenance by inducing Wt1 expression. Our study uncovers a new role of post-translational arginine methylation in granulosa cell differentiation and follicle development.
- 27 August 2021
- Published by eLife Sciences Publications, Ltd in eLife
- Vol. 10
Open AccessPRMT5 regulates ovarian follicle development by facilitating Wt1 translation
Protein arginine methyltransferase 5 (Prmt5) is the major type II enzyme responsible for symmetric dimethylation of arginine. Here, we found that PRMT5 was expressed at high level in ovarian granulosa cells of growing follicles. Inactivation of Prmt5 in granulosa cells resulted in aberrant follicle development and female infertility. In Prmt5-knockout mice, follicle development was arrested with disorganized granulosa cells in which WT1 expression was dramatically reduced and the expression of steroidogenesis-related genes was significantly increased. The premature differentiated granulosa cells were detached from oocytes and follicle structure was disrupted. Mechanism studies revealed that Wt1 expression was regulated by PRMT5 at the protein level. PRMT5 facilitated IRES-dependent translation of Wt1 mRNA by methylating HnRNPA1. Moreover, the upregulation of steroidogenic genes in Prmt5-deficient granulosa cells was repressed by Wt1 overexpression. These results demonstrate that PRMT5 participates in granulosa cell lineage maintenance by inducing Wt1 expression. Our study uncovers a new role of post-translational arginine methylation in granulosa cell differentiation and follicle development.
- 27 May 2021
- Published by Springer Nature in Machine Vision and Applications
- Vol. 32 (4) , 1-15
Multiplicative noise removal and blind inpainting of ultrasound images based on a new variational framework
- 27 May 2021
- Published by Springer Nature in Machine Vision and Applications
- Vol. 32 (4) , 1-15
Multiplicative noise removal and blind inpainting of ultrasound images based on a new variational framework
- 11 December 2020
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 65 (24) , 245040
Segmentation and classification of thyroid follicular neoplasm using cascaded convolutional neural network
In this paper, we present a segmentation and classification method for thyroid follicular neoplasms based on a combination of the prior-based level set method and deep convolutional neural network. The proposed method aims to discriminate thyroid follicular adenoma (TFA) and follicular thyroid carcinoma (FTC) in ultrasound images. In their appearance, these two kinds of tumours have similar shapes, sizes and contrasts. Therefore, it is difficult for even ultrasound specialists to distinguish them. Because of the complex background in thyroid ultrasound images, before distinguishing TFA and FTC, we need to segment the lesions from the whole image for each patient. The main challenge of segmentation is that the images often have weak edges and heterogeneous regions. The main issue of classification is that the accuracy depends on the features extracted from the segmentation results. To solve these problems, we conduct the two tasks, i.e. segmentation and classification, by a cascaded learning architecture. For segmentation, to obtain more accurate results, we exploit the Res-U-net framework and the prior-based level set method to enhance their respective abilities. Then, the classification network is trained by sharing shallow layers of the segmentation network. Testing the proposed method on real patient data shows that it is able to segment the lesion areas in thyroid ultrasound images with a Dice score of 92.65% and to distinguish TFA and FTC with a classification accuracy of 96.00%.
- 11 December 2020
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 65 (24) , 245040
Segmentation and classification of thyroid follicular neoplasm using cascaded convolutional neural network
In this paper, we present a segmentation and classification method for thyroid follicular neoplasms based on a combination of the prior-based level set method and deep convolutional neural network. The proposed method aims to discriminate thyroid follicular adenoma (TFA) and follicular thyroid carcinoma (FTC) in ultrasound images. In their appearance, these two kinds of tumours have similar shapes, sizes and contrasts. Therefore, it is difficult for even ultrasound specialists to distinguish them. Because of the complex background in thyroid ultrasound images, before distinguishing TFA and FTC, we need to segment the lesions from the whole image for each patient. The main challenge of segmentation is that the images often have weak edges and heterogeneous regions. The main issue of classification is that the accuracy depends on the features extracted from the segmentation results. To solve these problems, we conduct the two tasks, i.e. segmentation and classification, by a cascaded learning architecture. For segmentation, to obtain more accurate results, we exploit the Res-U-net framework and the prior-based level set method to enhance their respective abilities. Then, the classification network is trained by sharing shallow layers of the segmentation network. Testing the proposed method on real patient data shows that it is able to segment the lesion areas in thyroid ultrasound images with a Dice score of 92.65% and to distinguish TFA and FTC with a classification accuracy of 96.00%.
- 23 October 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 158-162
A Total Variation Model for Multiplicative Noise Removal and Blind Inpainting
Multiplicative noise and artifacts removal are two important tasks in the field of image processing. In some applications, multiplicative noise and intensity missing always exist at the same time. In this paper, we introduce a new variational model based on total variation and l 0 norm for simultaneously removing the multiplicative noise, estimating the location of missing pixels, and filling in them. To be specific, we use the total variation to regularize the estimated image, and use the l 0 norm to make the missing pixels to be sparse. Moreover, based on the Gamma noise assumption, the data fidelity term is given by a new forward description of the degraded process. Then, for solving the proposed model, we develop an algorithm by exploiting the alternating direction method of multipliers (ADMM). The experiments validate that the proposed method can effectively restore the synthetic and real images damaged by the multiplicative Gamma noise and artifacts, and meanwhile estimate the location of missing pixels.
- 23 October 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 158-162
A Total Variation Model for Multiplicative Noise Removal and Blind Inpainting
Multiplicative noise and artifacts removal are two important tasks in the field of image processing. In some applications, multiplicative noise and intensity missing always exist at the same time. In this paper, we introduce a new variational model based on total variation and l 0 norm for simultaneously removing the multiplicative noise, estimating the location of missing pixels, and filling in them. To be specific, we use the total variation to regularize the estimated image, and use the l 0 norm to make the missing pixels to be sparse. Moreover, based on the Gamma noise assumption, the data fidelity term is given by a new forward description of the degraded process. Then, for solving the proposed model, we develop an algorithm by exploiting the alternating direction method of multipliers (ADMM). The experiments validate that the proposed method can effectively restore the synthetic and real images damaged by the multiplicative Gamma noise and artifacts, and meanwhile estimate the location of missing pixels.
- 22 June 2017
- Published by Springer Nature in Machine Vision and Applications
- Vol. 28 (5-6) , 635-647
A fractional differential fidelity-based PDE model for image denoising
- 22 June 2017
- Published by Springer Nature in Machine Vision and Applications
- Vol. 28 (5-6) , 635-647
A fractional differential fidelity-based PDE model for image denoising