Segmentation of glioblastomas via 3D FusionNet DOI Creative Commons
Xiangyu Guo, Botao Zhang, Peng Yue

и другие.

Frontiers in Oncology, Год журнала: 2024, Номер 14

Опубликована: Ноя. 15, 2024

Introduction This study presented an end-to-end 3D deep learning model for the automatic segmentation of brain tumors. Methods The MRI data used in this were obtained from a cohort 630 GBM patients University Pennsylvania Health System (UPENN-GBM). Data augmentation techniques such as flip and rotations employed to further increase sample size training set. performance models was evaluated by recall, precision, dice score, Lesion False Positive Rate (LFPR), Average Volume Difference (AVD) Symmetric Surface Distance (ASSD). Results When applying FLAIR, T1, ceT1, T2 modalities, FusionNet-A FusionNet-C best-performing overall, with particularly excelling enhancing tumor areas, while demonstrates strong necrotic core peritumoral edema regions. excels areas across all metrics (0.75 0.83 precision 0.74 scores) also performs well regions (0.77 0.77 0.75 scores). Combinations including FLAIR ceT1 tend have better performance, especially Using only achieves recall 0.73 Visualization results indicate that our generally similar ground truth. Discussion FusionNet combines benefits U-Net SegNet, outperforming both. Although effectively segments tumors competitive accuracy, we plan extend framework achieve even performance.

Язык: Английский

A systematic review of trending technologies in non-invasive automatic brain tumor detection DOI

Jyoti Jyoti -,

Anuj Kumar

Multimedia Tools and Applications, Год журнала: 2024, Номер unknown

Опубликована: Окт. 17, 2024

Язык: Английский

Процитировано

0
Phenotypic approaches for CNS drugs DOI Creative Commons

Rakesh Sharma,

Caitlin R. M. Oyagawa, Hamid Abbasi

и другие.

Trends in Pharmacological Sciences, Год журнала: 2024, Номер 45(11), С. 997 - 1017

Опубликована: Окт. 21, 2024

Central nervous system (CNS) drug development is plagued by high clinical failure rate. Phenotypic assays promote translation of drugs reducing complex brain diseases to measurable, clinically valid phenotypes. We critique recent platforms integrating patient-derived cells, which most accurately recapitulate CNS disease phenotypes, with higher throughput models, including immortalized balance validity and scalability. These were screened conventional commercial chemogenomic compound libraries. explore emerging library curation strategies improve hit rate quality, screening novel fragment libraries as alternatives, for more tractable target deconvolution. The relevant models used in these could harbor important, unidentified targets, so we review evolving agnostic deconvolution approaches, chemical proteomics artificial intelligence (AI), aid phenotypic mechanism elucidation, thereby facilitating rational hit-to-drug optimization.

Язык: Английский

Процитировано

0
Artificial Intelligence Applications to Detect Pediatric Brain Tumor Biomarkers DOI
Parniyan Sadeghi, Yalda Ghazizadeh,

Soltani Arabshahi

и другие.

Interdisciplinary cancer research, Год журнала: 2024, Номер unknown

Опубликована: Янв. 1, 2024

Язык: Английский

Процитировано

0
Efficient Brain Tumor Detection Based on Channel Shuffling DOI Creative Commons
Pei Li, Rong Zhang,

Zhongjie Zhu

и другие.

Research Square (Research Square), Год журнала: 2024, Номер unknown

Опубликована: Ноя. 12, 2024

Abstract Brain tumor detection is crucial for early diagnosis and treatment planning, as it involves automatically identifying localizing brain tumors. However, existing methods often lack accuracy in detecting highly heterogeneous tumors struggle to balance speed. To alleviate these issues, a novel method termed channel shuffling YOLO (CS-YOLO) has been proposed, which optimizes both First, depthwise separable convolution with RepVGG module designed. This combines efficient parameter computation robust feature extraction. It extracts deep features from images, thereby enhancing speed of detection. Second, enhance the network's performance perceiving complex targets, convolutional multi-head self-attention constructed. learns long-range dependencies at lower resolutions, improving model's capability recognize Finally, lightweight designed used construct residual module. dramatically reduces number parameters computational complexity model by splitting channels, thus learning generalization performance. Experimental results demonstrate that proposed surpasses YOLOv6-L, YOLOv7, YOLOv8-L, latest RCS-YOLO terms on Br35H dataset. Compared state-of-the-art methods, CS-YOLO significantly enhances Specifically, network GFLOPs reduced 41%, FPS increased 14%, AP improved 0.8%, achieving advanced

Язык: Английский

Процитировано

0
Segmentation of glioblastomas via 3D FusionNet DOI Creative Commons
Xiangyu Guo, Botao Zhang, Peng Yue

и другие.

Frontiers in Oncology, Год журнала: 2024, Номер 14

Опубликована: Ноя. 15, 2024

Introduction This study presented an end-to-end 3D deep learning model for the automatic segmentation of brain tumors. Methods The MRI data used in this were obtained from a cohort 630 GBM patients University Pennsylvania Health System (UPENN-GBM). Data augmentation techniques such as flip and rotations employed to further increase sample size training set. performance models was evaluated by recall, precision, dice score, Lesion False Positive Rate (LFPR), Average Volume Difference (AVD) Symmetric Surface Distance (ASSD). Results When applying FLAIR, T1, ceT1, T2 modalities, FusionNet-A FusionNet-C best-performing overall, with particularly excelling enhancing tumor areas, while demonstrates strong necrotic core peritumoral edema regions. excels areas across all metrics (0.75 0.83 precision 0.74 scores) also performs well regions (0.77 0.77 0.75 scores). Combinations including FLAIR ceT1 tend have better performance, especially Using only achieves recall 0.73 Visualization results indicate that our generally similar ground truth. Discussion FusionNet combines benefits U-Net SegNet, outperforming both. Although effectively segments tumors competitive accuracy, we plan extend framework achieve even performance.

Язык: Английский

Процитировано

0