Neuron segmentation using 3D wavelet integrated encoder–decoder network

Li, Qiufu; Shen, Linlin

doi:10.1093/bioinformatics/btab716

Cited by 11 publications

(5 citation statements)

References 46 publications

(89 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this paper, we use UNet as the basic model and the reasons are as follows: (1) UNet and its variant’s version have been used in many research areas with good results, and it can provide ideas for our research [ 41 , 44 , 45 , 47 ]. (2) UNet is a kind of deep convolutional neural network [ 46 ], and the convolutional operation [ 46 ] is commonly used in the deep learning model.…”

Section: Methodsmentioning

confidence: 99%

Deep Learning-Based Road Traffic Noise Annoyance Assessment

Wang

Yuan

et al. 2023

IJERPH

View full text Add to dashboard Cite

With the development of urban road traffic, road noise pollution is becoming a public concern. Controlling and reducing the harm caused by traffic noise pollution have been the hot spots of traffic noise management research. The subjective annoyance level of traffic noise has become one of the most important measurements for evaluating road traffic pollution. There are subjective experimental methods and objective prediction methods to assess the annoyance level of traffic noise: the subjective experimental method usually uses social surveys or listening experiments in laboratories to directly assess the subjective annoyance level, which is highly reliable, but often requires a lot of time and effort. The objective method extracts acoustic features and predicts the annoyance level through model mapping. Combining the above two methods, this paper proposes a deep learning model-based objective annoyance evaluation method, which directly constructs the mapping between the noise and annoyance level based on the listening experimental results and realizes the rapid evaluation of the noise annoyance level. The experimental results show that this method has reduced the mean absolute error by 30% more than the regression algorithm and neural network, while its performance is insufficient in the annoyance interval where samples are lacking. To solve this problem, the algorithm adopts transfer learning to further improve the robustness with a 30% mean absolute error reduction and a 5% improvement in the correlation coefficient between the true results and predicted results. Although the model trained on college students’ data has some limitations, it is still a useful attempt to apply deep learning to noise assessment.

show abstract

Section: Methodsmentioning

confidence: 99%

Deep Learning-Based Road Traffic Noise Annoyance Assessment

Wang

Yuan

et al. 2023

IJERPH

View full text Add to dashboard Cite

show abstract

“…Neuron segmentation is a key step for neuronal digital reconstruction, which is eLife 9 of 33 essential for exploring brain circuits and understanding brain function (Li and Shen, 2022;Sugino et al, 2019). Distinct neuronal cell types acquire and maintain their identity by expressing different genes (Nelson et al, 2006).…”

Section: Scrna-seq Revealed That T Pallidum Infection Inhibited the D...mentioning

confidence: 99%

Single-cell RNA sequencing of iPSC-derived brain organoids revealsTreponema palliduminfection inhibiting neurodevelopment

Xu,

Wang,

et al. 2024

Preprint

View full text Add to dashboard Cite

Congenital syphilis is a vertically transmitted bacterial infection caused byTreponema pallidum, often causing multidomain neurodevelopmental disabilities. However, little is known about the pathogenesis of this disease. Brain organoids platform derived from the induced pluripotent stem cell (iPSC) is employed and exposed toT. palliduminfection for modelling congenital neurodevelopmental impairment. Single-cell RNA sequencing is used for identifying the subpopulations of differentially expressed genes and cellular heterogeneity and reconstructing differentiation trajectories followingT. palliduminfection. The results reveal thatT. palliduminfection influences the formation of neural rosette structures, reduces the cell number of the neural progenitor cell subcluster 1B (subNPC1B) and hindbrain neurons, and affects the neurodevelopment of the brain organoid. Moreover, it is speculated thatT. palliduminhibit the hindbrain neuron cell number through the suppression of subNPC1B subgroup in the organoids and inhibits transcription factor 3 activity in the subNPC1B-hindbrain neuronal axis. This is the first report on the inhibited effects ofT. pallidumon the neurodevelopment of the iPSC-derived brain organoid model. It is concluded thatT. pallidumcould inhibit the differentiation of subNPC1B in brain organoids, thereby reducing the differentiation from subNPC1B to hindbrain neurons, and ultimately affecting the development and maturation of hindbrain neurons.

show abstract

“…Moreover, the MWCNN technique was also applied on a U-Net architecture and the inverse wavelet transform (IWT) for high-resolution image restoration. Moreover, Qiufu Li et al [10] proposed a 3D waveletbased neuron segmentation method-3D WaveUNet. Their proposed model was proven effective to deal with the fine structured neurons that spread over a large area while avoiding the high computational cost that is inherent to the segmentation process of such structures.…”

Section: B Wavelet Poolingmentioning

confidence: 99%

“…Wavelet pooling was first introduced in the literature to overcome spatial information loss which is an inherent drawback in traditional pooling methods. It has been implemented with various deep architectures, in classification and segmentation applications, and solely as well as in hybrid pipelines with traditional pooling methods [8], [9], [10]. Also, the incorporation of all sub-bands, approximation sub-band only, as well as matching the input images to a specific combination of wavelet sub-bands [11] has been investigated in the literature.…”

Section: Introductionmentioning

confidence: 99%

Evaluating the Potential of Wavelet Pooling on Improving the Data Efficiency of Light-Weight CNNs

et al. 2023

View full text Add to dashboard Cite

Wavelet pooling (WP) in neural network architectures has recently demonstrated more discriminative power than traditional pooling methods. This is mainly because the latter suffer from spatial information loss while wavelet pooling harnesses the power of spectral information. However, the potential of WP in increasing the data efficiency and the extent of this potential have not been investigated yet. Data efficiency refers to the volume of training data required to attain a certain performance level during inference, e.g., recognition accuracy. In this research, we are concerned with evaluating the data efficiency of WP in light-weight architectures-MobileNets. Across a wide variety of seven datasets/applications including object recognition (CIFAR-10, STL-10, CINIC-10, and Intel Image Classification datasets) and diagnostic imaging (colon diseases, brain tumors, and malaria cell images datasets), and while considering classification accuracy as a performance metric, we show that WP achieves an average data saving that exceeds 30% compared to traditional pooling techniques. For other performance measures, namely, precision, recall, and F1-score, we report an average of 30% data saving for object recognition datasets and 22% saving for diagnostic imaging datasets. By focusing on a light-weight architecture, this research further emphasizes the significance of wavelet pooling in training and testing resources-challenged settings such as the applications of edge computing and green deep learning.

show abstract

Neuron segmentation using 3D wavelet integrated encoder–decoder network

Cited by 11 publications

References 46 publications

Deep Learning-Based Road Traffic Noise Annoyance Assessment

Deep Learning-Based Road Traffic Noise Annoyance Assessment

Single-cell RNA sequencing of iPSC-derived brain organoids revealsTreponema palliduminfection inhibiting neurodevelopment

Evaluating the Potential of Wavelet Pooling on Improving the Data Efficiency of Light-Weight CNNs

Contact Info

Product

Resources

About