U-Det: A Modified U-Net architecture with bidirectional feature network for lung nodule segmentation

Keetha, Nikhil Varma; P, Samson Anosh Babu; Annavarapu, Chandra Sekhara Rao

doi:10.48550/arxiv.2003.09293

Cited by 7 publications

(5 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Effective contextual information can help the model to segment the target better. Keetha et al (2020) came up with a U-Det model that adapts to multiple nodule segmentations. The model uses a bidirectional feature network to fuse multi-scale contextual features, and a new Mish activation function is designed to promote the training efficiency.…”

Section: Methods Based On Deep Learningmentioning

confidence: 99%

A semantic segmentation of the lung nodules using a shape attention-guided contextual residual network

2023

View full text Add to dashboard Cite

Objective. The early diagnosis of lung cancer depends on the precise segmentation of lung nodules. However, the variable size, uneven intensity, and blurred borders of lung nodules bring many challenges to the precise segmentation of lung nodules. Approach. We propose a shape attention-guided contextual residual network to address the difficult problem in lung nodule segmentation. Firstly, we establish a selective kernel convolution residual module to replace the original encoder and decoder. This module incorporates selective kernel convolution, which automatically selects convolutions with different receptive fields to acquire multi-scale spatial features. Secondly, we construct a multi-scale contextual attention module to assist the network in extracting multi-scale contextual feature of local feature maps. Finally, we develop a shape attention-guided module to assist the network to restore details such as the boundary and shape of lung nodules during the upsampling phase. Main Results. The proposed network is comprehensively analyzed using the publicly available LUNA16 dataset, and an ablation experiment is designed to validate the effectiveness of each individual component. Ultimately, the dice similarity coefficient of the experimental results reaches 87.39% on the test set. Compared to other state-of-the-art segmentation methods, the proposed network achieves superior performance in lung nodule segmentation. Significance. Our proposed network improves the accuracy of lung nodule segmentation and provides an important basis for subsequent physicians to develop treatment plans.

show abstract

Section: Methods Based On Deep Learningmentioning

confidence: 99%

A semantic segmentation of the lung nodules using a shape attention-guided contextual residual network

2023

View full text Add to dashboard Cite

show abstract

“…The mIOU score is 0.821 MultiResUNet [21] 2019 2D/3D 5-Fold Cross Validation was 78.1936 ± 0.7868 Dense network FD-UNet [22] 2018 2D AVERAGE SSIM scores was 0.82 ± 0.07 MDU-Net [23] 2018 2D Dice score was 0.928 H-Dense UNet [24] 2018 2D/3D Dice scores were 0.937 ± 0.02 [26] 2019 2D Dice scores were 0.97 ± 0.13 U-DET [27] 2020 2D Dice scores was 0.8282…”

Section: Commonly Used Deep Learning Modelsmentioning

confidence: 99%

A Review of the Application of Artificial Intelligence in Orthopedic Diseases

Diao,

Wang,

Qin

et al. 2024

CMC

View full text Add to dashboard Cite

In recent years, Artificial Intelligence (AI) has revolutionized people's lives. AI has long made breakthrough progress in the field of surgery. However, the research on the application of AI in orthopedics is still in the exploratory stage. The paper first introduces the background of AI and orthopedic diseases, addresses the shortcomings of traditional methods in the detection of fractures and orthopedic diseases, draws out the advantages of deep learning and machine learning in image detection, and reviews the latest results of deep learning and machine learning applied to orthopedic image detection in recent years, describing the contributions, strengths and weaknesses, and the direction of the future improvements that can be made in each study. Next, the paper also introduces the difficulties of traditional orthopedic surgery and the roles played by AI in preoperative, intraoperative, and postoperative orthopedic surgery, scientifically discussing the advantages and prospects of AI in orthopedic surgery. Finally, the article discusses the limitations of current research and technology in clinical applications, proposes solutions to the problems, and summarizes and outlines possible future research directions. The main objective of this review is to inform future research and development of AI in orthopedics.

show abstract

“…Additionally, MSAU-Net [26] attends to features across multiple scales simultaneously, while SegR-Net [27] reweights features spatially and channel-wise via attention. U-NetPlus [28] incorporates attention gates and dense skip connections for enhanced information flow. Moreover, U-Det [29] jointly performs detection and segmentation by integrating attention.…”

Section: Medical Segmentation Studiesmentioning

confidence: 99%

Automatic Medical Image Segmentation with Vision Transformer

Zhang,

Li,

Zhang

et al. 2024

Applied Sciences

View full text Add to dashboard Cite

Automatic image segmentation is vital for the computer-aided determination of treatment directions, particularly in terms of labelling lesions or infected areas. However, the manual labelling of disease regions is inconsistent and a time-consuming assignment. Meanwhile, radiologists’ comments are exceedingly subjective, regularly impacted by personal clinical encounters. To address these issues, we proposed a transformer learning strategy to automatically recognize infected areas in medical images. We firstly utilize a parallel partial decoder to aggregate high-level features and then generate a global feature map. Explicit edge attention and implicit reverse attention are applied to demonstrate boundaries and enhance their expression. Additionally, to alleviate the need for extensive labeled data, we propose a segmentation network combining propagation and transformer architectures that requires only a small amount of labeled data while leveraging fundamentally unlabeled images. The attention mechanisms are integrated within convolutional networks, keeping their global structures intact. Standalone transformers connected straightforwardly and receiving image patches can also achieve impressive segmentation performance. Our network enhanced the learning ability and attained a higher quality execution. We conducted a variety of ablation studies to demonstrate the adequacy of each modelling component. Experiments conducted across various medical imaging modalities illustrate that our model beats the most popular segmentation models. The comprehensive results also show that our transformer architecture surpasses established frameworks in accuracy while better preserving the natural variations in anatomy. Both quantitatively and qualitatively, our model achieves a higher overlap with ground truth segmentations and improved boundary adhesion.

show abstract

U-Det: A Modified U-Net architecture with bidirectional feature network for lung nodule segmentation

Cited by 7 publications

References 41 publications

A semantic segmentation of the lung nodules using a shape attention-guided contextual residual network

A semantic segmentation of the lung nodules using a shape attention-guided contextual residual network

A Review of the Application of Artificial Intelligence in Orthopedic Diseases

Automatic Medical Image Segmentation with Vision Transformer

Contact Info

Product

Resources

About