Segmentation of MR images for brain tumor detection using autoencoder neural network

Discover Artificial Intelligence, Journal Year: 2024, Volume and Issue: 4(1)

Published: Oct. 26, 2024

Language: Английский

---

G-Net: Implementing an enhanced brain tumor segmentation framework using semantic segmentation design

PLoS ONE, Journal Year: 2024, Volume and Issue: 19(8), P. e0308236 - e0308236

Published: Aug. 6, 2024

A fundamental computer vision task called semantic segmentation has significant uses in the understanding of medical pictures, including tumors brain. The G-Shaped Net architecture appears this context as an innovative and promising design that combines components from many models to attain improved accuracy efficiency. In order improve efficiency, synergistically incorporates four components: Self-Attention, Squeeze Excitation, Fusion, Spatial Pyramid Pooling block structures. These factors work together precision effectiveness brain tumor segmentation. a crucial component architecture, gives model ability concentrate on image’s most informative areas, enabling accurate localization boundaries. By adjusting channel-wise feature maps, Excitation completes by improving model’s capacity capture fine-grained information pictures. Since provides multi-scale contextual information, is capable handling various sizes complexity levels. Additionally, Fusion architectures combine characteristics sources, thorough comprehension image outcomes. asset for imaging diagnostics represents substantial development segmentation, which needed more

Language: Английский

---

Parallel-way: Multi-modality-based brain tumor segmentation using parallel capsule network

Electromagnetic Biology and Medicine, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 25

Published: Oct. 29, 2024

Brain tumors present a formidable diagnostic challenge due to their aberrant cell growth. Accurate determination of tumor location and size is paramount for effective diagnosis. Magnetic Resonance Imaging (MRI) Positron Emission Tomography (PET) are pivotal tools in clinical diagnosis, yet segmentation within images remains challenging, particularly at boundary pixels, owing limited sensitivity. Recent endeavors have introduced fusion-based strategies refine accuracy, these methods often prove inadequate. In response, we introduce the Parallel-Way framework surmount obstacles. Our approach integrates MRI PET data holistic analysis. Initially, enhance image quality by employing noise reduction, bias field correction, adaptive thresholding, leveraging Improved Kalman Filter (IKF), Expectation Maximization (EM), Vibe Algorithm (IVib), respectively. Subsequently, conduct multi-modality fusion through Dual-Tree Complex Wavelet Transform (DTWCT) amalgamate from both modalities. Following fusion, extract pertinent features using Advanced Capsule Network (ACN) reduce feature dimensionality via Multi-objective Diverse Evolution-based selection. Tumor then executed utilizing Twin Vision Transformer with dual attention mechanism. Implemented our which exhibits heightened model performance. Evaluation across multiple metrics, including sensitivity, specificity, F1-Score, AUC, underscores its superiority over existing methodologies.

Language: Английский

---

An Approach of SIFT With Fed-VGG16 and Fed-CNN for Identification and Classification of Brain Tumors

Advances in medical technologies and clinical practice book series, Journal Year: 2024, Volume and Issue: unknown, P. 70 - 85

Published: April 15, 2024

Brain tumors develop when cells in the brain multiply rapidly and unchecked. It can be fatal if not addressed its early stages. Getting segmentation classification right is still a challenge, despite lot of work good results this field. Radiologists may now more easily locate tumor regions with use experimental medical imaging techniques like magnetic resonance (MRI). Image processing such as pre-processing, segmentation, contour detection, feature extraction using SIFT (scale invariant transformation), VGG16, CNN, Fed-VGG16, Fed-CNN classifiers, evaluation confusion matrices are presented study. The models reach up to 97%, 98.51%, 99.28%, 100% accuracy for used correspondingly, according data. In order facilitate detection subsequent research activity, it seeks mitigate some problems that have already been addressed.

Language: Английский

---

Preoperative prediction of histopathological grading in patients with chondrosarcoma using MRI-based radiomics with semantic features

BMC Medical Imaging, Journal Year: 2024, Volume and Issue: 24(1)

Published: July 11, 2024

Abstract Background Distinguishing high-grade from low-grade chondrosarcoma is extremely vital not only for guiding the development of personalized surgical treatment but also predicting prognosis patients. We aimed to establish and validate a magnetic resonance imaging (MRI)-based nomogram preoperative grading in patients with chondrosarcoma. Methods Approximately 114 (60 54 cases chondrosarcoma, respectively) were recruited this retrospective study. All treated via surgery histopathologically proven, they randomly divided into training ( n = 80) validation 34) sets at ratio 7:3. Next, radiomics features extracted two sequences using least absolute shrinkage selection operator (LASSO) algorithms. The rad-scores calculated then subjected logistic regression develop model. A combining independent predictive semantic radiomic by multivariate was established. performance each model assessed receiver operating characteristic (ROC) curve analysis area under curve, while clinical efficacy evaluated decision (DCA). Results Ultimately, six optimal signatures T1-weighted (T1WI) T2-weighted fat suppression (T2WI-FS) Tumour cartilage abundance, which emerged as an predictor, significantly related p < 0.05). AUC values 0.85 (95% CI, 0.76 0.95) sets, corresponding 0.82 0.65 0.98), far superior 0.68 0.58 0.79) 0.72 0.57 0.87) sets. demonstrated good distinction DCA revealed that had markedly higher usefulness preoperatively than either rad-score or alone. Conclusion based on MRI combined factors better differentiation between has potential noninvasive tool personalizing plans.

Language: Английский

---