Wei-Wen Hsu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Wei-Wen Hsu has a total of 18 co-authors. Here are the top co-authors by number of shared publications:
- Jing-Ming Guo
National Taiwan University of Science and Technology
3 shared publications - Sankarasrinivasan Seshathiri
Department of Electrical Engineering, National Taiwan University of Science and Technology, Taipei 106335, Taiwan
2 shared publications Chang Hao
Institute of Big Data Technology, Bright Oceans Corporation, Beijing, China
1 shared publication- Chien-Yu Chen
Department of Electrical Engineering, National Taiwan University of Science and Technology, Taipei 106335, Taiwan
1 shared publication Chung-Hao Chen
Department of Electrical and Computer Engineering, Old Dominion University, Norfolk, USA
1 shared publication- Jia-Hao Liu
Department of Electrical Engineering, National Taiwan University of Science and Technology, Taipei 106335, Taiwan
1 shared publication
Latest Publications
- 16 September 2023
- Published by Springer Nature in Service Oriented Computing and Applications
- Vol. 18 (2) , 183-193
A computer-aided diagnosis system for breast pathology: a deep learning approach with model interpretability from pathological perspective
- 16 September 2023
- Published by Springer Nature in Service Oriented Computing and Applications
- Vol. 18 (2) , 183-193
A computer-aided diagnosis system for breast pathology: a deep learning approach with model interpretability from pathological perspective
- Open Access
- 26 April 2023
- Published by MDPI AG in Electronics
- Vol. 12 (9) , 2013
A Study of the Interpretability of Fundus Analysis with Deep Learning-Based Approaches for Glaucoma Assessment
Earlier studies focused on training ResNet50 deep learning models on a dataset of fundus images from the National Taiwan University Hospital HsinChu Branch. The study aimed to identify class-specific discriminative areas related to various conditions of ganglion cell complex (GCC) thickness, center focus areas, cropped patches from the fundus, and dataset partitions. The study utilized two visualization methods to evaluate and explain the areas of interest of the network model and determine if they aligned with clinical diagnostic knowledge. The results of the experiments demonstrated that incorporating GCC thickness information improved the accuracy of glaucoma determination. The deep learning models primarily focused on the optic nerve head (ONH) for glaucoma diagnosis, which was consistent with clinical rules. Nonetheless, the models achieved high prediction accuracy in detecting glaucomatous cases using only cropped images of macular areas. Moreover, the model’s focus on regions with GCC impairment in some cases indicates that deep learning models can identify morphologically detailed alterations in fundus photographs that may be beyond the scope of visual diagnosis by experts. This highlights the significant contribution of deep learning models in the diagnosis of glaucoma.
- 26 April 2023
- Published by MDPI AG in Electronics
- Vol. 12 (9) , 2013
Open AccessA Study of the Interpretability of Fundus Analysis with Deep Learning-Based Approaches for Glaucoma Assessment
Earlier studies focused on training ResNet50 deep learning models on a dataset of fundus images from the National Taiwan University Hospital HsinChu Branch. The study aimed to identify class-specific discriminative areas related to various conditions of ganglion cell complex (GCC) thickness, center focus areas, cropped patches from the fundus, and dataset partitions. The study utilized two visualization methods to evaluate and explain the areas of interest of the network model and determine if they aligned with clinical diagnostic knowledge. The results of the experiments demonstrated that incorporating GCC thickness information improved the accuracy of glaucoma determination. The deep learning models primarily focused on the optic nerve head (ONH) for glaucoma diagnosis, which was consistent with clinical rules. Nonetheless, the models achieved high prediction accuracy in detecting glaucomatous cases using only cropped images of macular areas. Moreover, the model’s focus on regions with GCC impairment in some cases indicates that deep learning models can identify morphologically detailed alterations in fundus photographs that may be beyond the scope of visual diagnosis by experts. This highlights the significant contribution of deep learning models in the diagnosis of glaucoma.
- Open Access
- 13 February 2023
- Published by MDPI AG in Electronics
- Vol. 12 (4) , 928
Two-Stage Cascaded CNN Model for 3D Mitochondria EM Segmentation
Mitochondria are the organelles that generate energy for the cells. Many studies have suggested that mitochondrial dysfunction or impairment may be related to cancer and other neurodegenerative disorders such as Alzheimer’s and Parkinson’s diseases. Therefore, morphologically detailed alterations in mitochondria and 3D reconstruction of mitochondria are highly demanded research problems in the performance of clinical diagnosis. Nevertheless, manual mitochondria segmentation over 3D electron microscopy volumes is not a trivial task. This study proposes a two-stage cascaded CNN architecture to achieve automated 3D mitochondria segmentation, combining the merits of top-down and bottom-up approaches. For top-down approaches, the segmentation is conducted on objects’ localization so that the delineations of objects’ contours can be more precise. However, the combinations of 2D segmentation from the top-down approaches are inadequate to perform proper 3D segmentation without the information on connectivity among frames. On the other hand, the bottom-up approach finds coherent groups of pixels and takes the information of 3D connectivity into account in segmentation to avoid the drawbacks of the 2D top-down approach. However, many small areas that share similar pixel properties with mitochondria become false positives due to insufficient information on objects’ localization. In the proposed method, the detection of mitochondria is carried out with multi-slice fusion in the first stage, forming the segmentation cues. Subsequently, the second stage is to perform 3D CNN segmentation that learns the pixel properties and the information of 3D connectivity under the supervision of cues from the detection stage. Experimental results show that the proposed structure alleviates the problems in both the top-down and bottom-up approaches, which significantly accomplishes better performance in segmentation and expedites clinical analysis.
- 13 February 2023
- Published by MDPI AG in Electronics
- Vol. 12 (4) , 928
Open AccessTwo-Stage Cascaded CNN Model for 3D Mitochondria EM Segmentation
Mitochondria are the organelles that generate energy for the cells. Many studies have suggested that mitochondrial dysfunction or impairment may be related to cancer and other neurodegenerative disorders such as Alzheimer’s and Parkinson’s diseases. Therefore, morphologically detailed alterations in mitochondria and 3D reconstruction of mitochondria are highly demanded research problems in the performance of clinical diagnosis. Nevertheless, manual mitochondria segmentation over 3D electron microscopy volumes is not a trivial task. This study proposes a two-stage cascaded CNN architecture to achieve automated 3D mitochondria segmentation, combining the merits of top-down and bottom-up approaches. For top-down approaches, the segmentation is conducted on objects’ localization so that the delineations of objects’ contours can be more precise. However, the combinations of 2D segmentation from the top-down approaches are inadequate to perform proper 3D segmentation without the information on connectivity among frames. On the other hand, the bottom-up approach finds coherent groups of pixels and takes the information of 3D connectivity into account in segmentation to avoid the drawbacks of the 2D top-down approach. However, many small areas that share similar pixel properties with mitochondria become false positives due to insufficient information on objects’ localization. In the proposed method, the detection of mitochondria is carried out with multi-slice fusion in the first stage, forming the segmentation cues. Subsequently, the second stage is to perform 3D CNN segmentation that learns the pixel properties and the information of 3D connectivity under the supervision of cues from the detection stage. Experimental results show that the proposed structure alleviates the problems in both the top-down and bottom-up approaches, which significantly accomplishes better performance in segmentation and expedites clinical analysis.
- Open Access
Fall Detection with the Spatial-Temporal Correlation Encoded by a Sequence-to-Sequence Denoised GAN
Falling is a major cause of personal injury and accidental death worldwide, in particular for the elderly. For aged care, a falling alarm system is highly demanded so that medical aid can be obtained immediately when the fall accidents happen. Previous studies on fall detection lacked practical considerations to deal with real-world situations, including the camera’s mounting angle, lighting differences between day and night, and the privacy protection for users. In our experiments, IR-depth images and thermal images were used as the input source for fall detection; as a result, detailed facial information is not captured by the system for privacy reasons, and it is invariant to the lighting conditions. Due to the different occurrence rates between fall accidents and other normal activities, supervised learning approaches may suffer from the problem of data imbalance in the training phase. Accordingly, in this study, anomaly detection is performed using unsupervised learning approaches so that the models were trained only with the normal cases while the fall accident was defined as an anomaly event. The proposed system takes sequential frames as the inputs to predict future frames based on a GAN structure, and it provides (1) multi-subject detection, (2) real-time fall detection triggered by motion, (3) a solution to the situation that subjects were occluded after falling, and (4) a denoising scheme for depth images. The experimental results show that the proposed system achieves the state-of-the-art performance and copes with the real-world cases successfully.
Open AccessFall Detection with the Spatial-Temporal Correlation Encoded by a Sequence-to-Sequence Denoised GAN
Falling is a major cause of personal injury and accidental death worldwide, in particular for the elderly. For aged care, a falling alarm system is highly demanded so that medical aid can be obtained immediately when the fall accidents happen. Previous studies on fall detection lacked practical considerations to deal with real-world situations, including the camera’s mounting angle, lighting differences between day and night, and the privacy protection for users. In our experiments, IR-depth images and thermal images were used as the input source for fall detection; as a result, detailed facial information is not captured by the system for privacy reasons, and it is invariant to the lighting conditions. Due to the different occurrence rates between fall accidents and other normal activities, supervised learning approaches may suffer from the problem of data imbalance in the training phase. Accordingly, in this study, anomaly detection is performed using unsupervised learning approaches so that the models were trained only with the normal cases while the fall accident was defined as an anomaly event. The proposed system takes sequential frames as the inputs to predict future frames based on a GAN structure, and it provides (1) multi-subject detection, (2) real-time fall detection triggered by motion, (3) a solution to the situation that subjects were occluded after falling, and (4) a denoising scheme for depth images. The experimental results show that the proposed system achieves the state-of-the-art performance and copes with the real-world cases successfully.