A fine-tuned vision transformer based enhanced multi-class brain tumor classification using MRI scan imagery

Frontiers in Oncology, Journal Year: 2024, Volume and Issue: 14

Published: July 18, 2024

Brain tumors occur due to the expansion of abnormal cell tissues and can be malignant (cancerous) or benign (not cancerous). Numerous factors such as position, size, progression rate are considered while detecting diagnosing brain tumors. Detecting in their initial phases is vital for diagnosis where MRI (magnetic resonance imaging) scans play an important role. Over years, deep learning models have been extensively used medical image processing. The current study primarily investigates novel Fine-Tuned Vision Transformer (FTVTs)—FTVT-b16, FTVT-b32, FTVT-l16, FTVT-l32—for tumor classification, also comparing them with other established ResNet50, MobileNet-V2, EfficientNet - B0. A dataset 7,023 images (MRI scans) categorized into four different classes, namely, glioma, meningioma, pituitary, no classification. Further, presents a comparative analysis these including accuracies evaluation metrics recall, precision, F1-score across each class. ResNet-50, EfficientNet-B0, MobileNet-V2 obtained accuracy 96.5%, 95.1%, 94.9%, respectively. Among all FTVT models, FTVT-l16 model achieved remarkable 98.70% whereas FTVT-b16, FTVT-132 98.09%, 96.87%, 98.62%, respectively, hence proving efficacy robustness FTVT’s

Language: Английский

---

Lexicon and Deep Learning-Based Approaches in Sentiment Analysis on Short Texts

Journal of Computer and Communications, Journal Year: 2024, Volume and Issue: 12(01), P. 11 - 34

Published: Jan. 1, 2024

Social media is an essential component of our personal and professional lives. We use it extensively to share various things, including opinions on daily topics feelings about different subjects. This sharing posts provides insights into someone's current emotions. In artificial intelligence (AI) deep learning (DL), researchers emphasize opinion mining analysis sentiment, particularly social platforms such as Twitter (currently known X), which has a global user base. research work revolves explicitly around comparison between two popular approaches: Lexicon-based Deep learning-based Approaches. To conduct this study, study used dataset called sentiment140, contains over 1.5 million data points. The primary focus was the Long Short-Term Memory (LSTM) sequence model. beginning, we particular techniques preprocess data. divided training test evaluated performance model using Simultaneously, have applied lexicon-based approach same recorded outputs. Finally, compared approaches by creating confusion matrices based their respective allows us assess precision, recall, F1-Score, enabling determine yields better accuracy. achieved 98% accuracy for algorithms 95% approach.

Language: Английский

---

Predictive modeling for breast cancer classification in the context of Bangladeshi patients by use of machine learning approach with explainable AI

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: April 11, 2024

Abstract Breast cancer has rapidly increased in prevalence recent years, making it one of the leading causes mortality worldwide. Among all cancers, is by far most common. Diagnosing this illness manually requires significant time and expertise. Since detecting breast a time-consuming process, preventing its further spread can be aided creating machine-based forecasts. Machine learning Explainable AI are crucial classification as they not only provide accurate predictions but also offer insights into how model arrives at decisions, aiding understanding trustworthiness results. In study, we evaluate compare accuracy, precision, recall, F1 scores five different machine methods using primary dataset (500 patients from Dhaka Medical College Hospital). Five supervised techniques, including decision tree, random forest, logistic regression, naive bayes, XGBoost, have been used to achieve optimal results on our dataset. Additionally, study applied SHAP analysis XGBoost interpret model’s understand impact each feature output. We compared accuracy with which several algorithms classified data, well contrasted other literature field. After final evaluation, found that achieved best 97%.

Language: Английский

---

Optimizing medical image analysis through MViTX on multiple datasets with explainable AI

Neural Computing and Applications, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 28, 2025

Language: Английский

---

Radiomics-driven neuro-fuzzy framework for rule generation to enhance explainability in MRI-based brain tumor segmentation

Frontiers in Neuroinformatics, Journal Year: 2025, Volume and Issue: 19

Published: April 17, 2025

Brain tumors are a leading cause of mortality worldwide, with early and accurate diagnosis being essential for effective treatment. Although Deep Learning (DL) models offer strong performance in tumor detection segmentation using MRI, their black-box nature hinders clinical adoption due to lack interpretability. We present hybrid AI framework that integrates 3D U-Net Convolutional Neural Network MRI-based radiomic feature extraction. Dimensionality reduction is performed machine learning, an Adaptive Neuro-Fuzzy Inference System (ANFIS) employed produce interpretable decision rules. Each experiment constrained small set high-impact features enhance clarity reduce complexity. The was validated on the BraTS2020 dataset, achieving average DICE Score 82.94% core 76.06% edema segmentation. Classification tasks yielded accuracies 95.43% binary (healthy vs. tumor) 92.14% multi-class edema) problems. A concise 18 fuzzy rules generated provide clinically outputs. Our approach balances high diagnostic accuracy enhanced interpretability, addressing critical barrier applying DL settings. Integrating ANFIS radiomics supports transparent decision-making, facilitating greater trust applicability real-world medical diagnostics assistance.

Language: Английский

---

A Hybrid Transfer Learning Framework for Brain Tumor Diagnosis

Advanced Intelligent Systems, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 5, 2025

Brain tumors are among the most severe health challenges, necessitating early and precise diagnosis for effective treatment planning. This study introduces an optimized hybrid transfer learning (TL) framework brain tumor classification using magnetic resonance imaging images. The proposed system integrates advanced preprocessing techniques, ensemble of pretrained deep models, explainable artificial intelligence (XAI) methods to achieve high accuracy reliability. methodology enhances image quality through noise reduction contrast enhancement, facilitating robust feature extraction. model combines VGG16 ResNet152V2 architectures, achieving a 99.47% on challenging four‐class dataset. Additionally, gradient‐weighted class activation mapping SHapley Additive exPlanations (SHAP)‐based XAI techniques provide visual quantitative insights into predictions, improving interpretability clinical trust. comprehensive demonstrates potential TL in advancing diagnostic supporting decision‐making detection. results underscore its applicability settings, particularly resource‐constrained environments.

Language: Английский

---

Machine Learning Model Development for Malignant Prostate Lesion Prediction Using Texture Analysis Features from Ultrasound Shear-Wave Elastography

Cancers, Journal Year: 2025, Volume and Issue: 17(8), P. 1358 - 1358

Published: April 18, 2025

Introduction: Artificial intelligence (AI) is increasingly utilized for texture analysis and the development of machine learning (ML) techniques to enhance diagnostic accuracy. ML algorithms are trained differentiate between normal malignant conditions based on provided data. Texture feature analysis, including first-order second-order features, a critical step in development. This study aimed evaluate quantitative features prostate cancer tissues identified through ultrasound B-mode shear-wave elastography (SWE) imaging develop assess models predicting classifying versus tissues. Methodology: First-order were extracted from SWE imaging, four reconstructed regions interest (ROIs) images A total 94 derived, intensity, Gray-Level Co-Occurrence Matrix (GLCM), Dependence Length (GLDLM), Run (GLRLM), Size Zone (GLSZM). Five developed evaluated using 5-fold cross-validation predict Results: Data 62 patients analyzed. All ROIs, except those derived exhibited statistically significant differences Among models, Support Vector Machines (SVM), Random Forest (RF), Naive Bayes (NB) demonstrated highest performance across all ROIs. These consistently achieved strong predictive accuracy Gray Pure Reconstructed Provided sensitivity specificity PCa prediction by 82%, 90%, 98%, 96%, respectively. Conclusions: with SWE-US effectively differentiates benign lesions, like contrast, entropy, correlation playing key role. Forest, SVM, Naïve showed classification performance, while grayscale reconstructions (GPSWE GRRI) enhanced detection

Language: Английский

---

Advancements in Brain Tumour Analysis: A Review of Machine Learning, Deep Learning, Image Processing, and Explainable AI Techniques

Operations Research Forum, Journal Year: 2025, Volume and Issue: 6(2)

Published: May 5, 2025

Language: Английский

---

Malignancy pattern analysis of breast ultrasound images using clinical features and a graph convolutional network

Digital Health, Journal Year: 2024, Volume and Issue: 10

Published: Jan. 1, 2024

Early diagnosis of breast cancer can lead to effective treatment, possibly increase long-term survival rates, and improve quality life. The objective this study is present an automated analysis classification system for using clinical markers such as tumor shape, orientation, margin, surrounding tissue. novelty uniqueness the lie in approach considering medical features based on radiologists.

Language: Английский

---

Enhancing Bangla Fake News Detection Using Bidirectional Gated Recurrent Units and Deep Learning Techniques

Published: April 18, 2024

Language: Английский

---