Enhancing brain tumor detection in MRI images through explainable AI using Grad-CAM with Resnet 50

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

Published: May 11, 2024

Abstract This study addresses the critical challenge of detecting brain tumors using MRI images, a pivotal task in medical diagnostics that demands high accuracy and interpretability. While deep learning has shown remarkable success image analysis, there remains substantial need for models are not only accurate but also interpretable to healthcare professionals. The existing methodologies, predominantly learning-based, often act as black boxes, providing little insight into their decision-making process. research introduces an integrated approach ResNet50, model, combined with Gradient-weighted Class Activation Mapping (Grad-CAM) offer transparent explainable framework tumor detection. We employed dataset enhanced through data augmentation, train validate our model. results demonstrate significant improvement model performance, testing 98.52% precision-recall metrics exceeding 98%, showcasing model’s effectiveness distinguishing presence. application Grad-CAM provides insightful visual explanations, illustrating focus areas making predictions. fusion explainability holds profound implications diagnostics, offering pathway towards more reliable detection tools.

Language: Английский

---

BrainNet: Precision Brain Tumor Classification with Optimized EfficientNet Architecture

International Journal of Intelligent Systems, Journal Year: 2024, Volume and Issue: 2024(1)

Published: Jan. 1, 2024

Brain tumors significantly impact human health due to their complexity and the challenges in early detection treatment. Accurate diagnosis is crucial for effective intervention, but existing methods often suffer from limitations accuracy efficiency. To address these challenges, this study presents a novel deep learning (DL) approach utilizing EfficientNet family enhanced brain tumor classification detection. Leveraging comprehensive dataset of 3064 T1‐weighted CE MRI images, our methodology incorporates advanced preprocessing augmentation techniques optimize model performance. The experiments demonstrate that EfficientNetB(07) achieved 99.14%, 98.76%, 99.07%, 99.69%, 99.07% accuracy, respectively. pinnacle research EfficientNetB3 model, which demonstrated exceptional performance with an rate 99.69%. This surpasses many state‐of‐the‐art (SOTA) techniques, underscoring efficacy approach. precision high‐accuracy DL promises improve diagnostic reliability speed clinical settings, facilitating earlier more treatment strategies. Our findings suggest significant potential improving patient outcomes diagnosis.

Language: Английский

---

A hybrid explainable model based on advanced machine learning and deep learning models for classifying brain tumors using MRI images

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Jan. 10, 2025

Brain tumors present a significant global health challenge, and their early detection accurate classification are crucial for effective treatment strategies. This study presents novel approach combining lightweight parallel depthwise separable convolutional neural network (PDSCNN) hybrid ridge regression extreme learning machine (RRELM) accurately classifying four types of brain (glioma, meningioma, no tumor, pituitary) based on MRI images. The proposed enhances the visibility clarity tumor features in images by employing contrast-limited adaptive histogram equalization (CLAHE). A PDSCNN is then employed to extract relevant tumor-specific patterns while minimizing computational complexity. RRELM model proposed, enhancing traditional ELM improved performance. framework compared with various state-of-the-art models terms accuracy, parameters, layer sizes. achieved remarkable average precision, recall, accuracy values 99.35%, 99.30%, 99.22%, respectively, through five-fold cross-validation. PDSCNN-RRELM outperformed pseudoinverse (PELM) exhibited superior introduction led enhancements performance parameters sizes those models. Additionally, interpretability was demonstrated using Shapley Additive Explanations (SHAP), providing insights into decision-making process increasing confidence real-world diagnosis.

Language: Английский

---

Detection of MRI brain tumor using residual skip block based modified MobileNet model

Cluster Computing, Journal Year: 2025, Volume and Issue: 28(4)

Published: Feb. 25, 2025

Language: Английский

---

Developments in Brain Tumor Segmentation Using MRI: Deep Learning Insights and Future Perspectives

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 26875 - 26896

Published: Jan. 1, 2024

The human brain is an incredible and wonderful organ that governs all body actions. Due to its great importance, any defect in the shape of regions should be reported quickly reduce death rate. abnormal region segmentation helps plan monitor treatment. most critical procedure isolating normal tissues from each other. So far, remarkable imaging modalities are being used diagnose abnormalities at their early stages, magnetic resonance (MRI) renowned noninvasive among those modalities. This paper investigates current landscape tumor (BTS) by exploring emerging deep learning (DL) methods for MRI analysis. findings offer a comprehensive comparison recent DL approaches, emphasizing effectiveness handling diverse types while addressing limitations associated with data scarcity robust validation. has shown vital improvement BTS, so our primary focus include significant models analyze MRI. However, outperforms traditional methods; still, there several limitations, especially related types, lack datasets, weak validations. future perspectives DL-based BTS present potential revolutionizing diagnosis treatment tumors.

Language: Английский

---

Enhancing Alzheimer’s disease classification through split federated learning and GANs for imbalanced datasets

PeerJ Computer Science, Journal Year: 2024, Volume and Issue: 10, P. e2459 - e2459

Published: Nov. 29, 2024

In the rapidly evolving healthcare sector, using advanced technologies to improve medical classification systems has become crucial for enhancing patient care, diagnosis, and treatment planning. There are two main challenges faced in this domain (i) imbalanced distribution of data, leading biased model performance (ii) need preserve privacy comply with data protection regulations. The primary goal project is develop a Alzheimer’s disease detection that can effectively learn from decentralized datasets without compromising on privacy. proposed system aims address these by employing an approach combines split federated learning (SFL) conditional generative adversarial networks (cGANs) enhance models. SFL enables efficient set distributed agents collaboratively train models sharing their thus improving integration GANs model’s ability generalize across classes generating realistic synthetic samples minority classes. provided accuracy approximately 83.54 percentage dataset.

Language: Английский

---

FKD-Med: Privacy-Aware, Communication-Optimized Medical Image Segmentation via Federated Learning and Model Lightweighting Through Knowledge Distillation

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 33687 - 33704

Published: Jan. 1, 2024

Advances in deep learning have revolutionized medical image segmentation, facilitating the precise delineation of complex anatomical structures. The scarcity annotated training samples remains a significant bottleneck. To tackle data limitation, federated (FL) offers promise pooling from multiple healthcare institutions. However, as models grow larger, increase communication costs restricts FL to fewer nodes, which constrains volume data. This situation necessitates simultaneous achievement model lightweighting. address this problem, study proposes FKD-Med, novel framework that integrates for privacy-sensitive amalgamation across institutions, and uses knowledge distillation (KD) enhance efficiency. "Med" FKD-Med refers application computational problems. Our principal contributions encompass design an open-source seamlessly blends KD, rendering it applicable broad spectrum informatics tasks. approach substantially augments volume, thereby boosting both efficiency throughput. Tested on two datasets segmentation using TransUNet ResUNet teacher models, achieves privacy, lowers costs, increases accuracy. parameters student were reduced 1/127 1/1027 those models. Additionally, subjected KD exhibited accuracy improvements 0.25%, 0.43%, 1.35%, 1.46% respectively, given same parameter volume. positions not only pivotal tool multi-institutional research but also versatile platform adaptable wide array real-world engineering applications. code is publicly available at https://github.com/SUN-1024/FKD-Med.

Language: Английский

---

ContourTL-Net: Contour-Based Transfer Learning Algorithm for Early-Stage Brain Tumor Detection

International Journal of Biomedical Imaging, Journal Year: 2024, Volume and Issue: 2024, P. 1 - 20

Published: April 29, 2024

Brain tumors are critical neurological ailments caused by uncontrolled cell growth in the brain or skull, often leading to death. An increasing patient longevity rate requires prompt detection; however, complexities of tissue make early diagnosis challenging. Hence, automated tools necessary aid healthcare professionals. This study is particularly aimed at improving efficacy computerized tumor detection a clinical setting through deep learning model. novel thresholding-based MRI image segmentation approach with transfer model based on contour (ContourTL-Net) suggested facilitate malignancies an initial phase. The utilizes contour-based analysis, which for object detection, precise segmentation, and capturing subtle variations morphology. employs VGG-16 architecture priorly trained “ImageNet” collection feature extraction categorization. designed utilize its ten nontrainable three trainable convolutional layers dropout layers. proposed ContourTL-Net evaluated two benchmark datasets four ways, among unseen case considered as aspect. Validating data crucial determine model’s generalization capability, domain adaptation, robustness, real-world applicability. Here, presented outcomes demonstrate highly accurate classification data, achieving perfect sensitivity negative predictive value (NPV) 100%, 98.60% specificity, 99.12% precision, 99.56% F1 -score, 99.46% accuracy. Additionally, compared state-of-the-art methodologies further enhance effectiveness. solution outperforms existing solutions both seen potential significantly improve efficiency accuracy, earlier diagnoses improved outcomes.

Language: Английский

---

Optimizing cervical cancer classification using transfer learning with deep gaussian processes and support vector machines

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

Published: Oct. 30, 2024

Abstract Background Cervical cancer is the fourth most frequent in women worldwide. Even though cervical deaths have decreased significantly Western countries, low and middle-income countries account for nearly 90% of deaths. While are leveraging powers artificial intelligence (AI) health sector, sub-Saharan Africa still lagging. In Uganda, cytologists manually analyze Pap smear images detection cancer, a process that highly subjective, slow, tedious. Machine learning (ML) algorithms been used automated classification cancer. However, MLs overfitting limitations which limits their deployment, especially sector where accurate predictions needed. Methods this study, we propose two kernel-based These (1) an optimized support vector machine (SVM), (2) deep Gaussian Process (DGP) model. The SVM model proposed uses radial basis kernel while DGP hybrid periodic local kernel. Results Experimental results revealed accuracy 100% 99.48% respectively. on precision, recall, F1 score were also reported. Conclusions models performed well classification, therefore suitable deployment. We plan to deploy our mobile application-based tool. limitation study was lack access high-performance computational resources.

Language: Английский

---

Modernizing Neuro-Oncology: The Impact of Imaging, Liquid Biopsies, and AI on Diagnosis and Treatment

International Journal of Molecular Sciences, Journal Year: 2025, Volume and Issue: 26(3), P. 917 - 917

Published: Jan. 22, 2025

Advances in neuro-oncology have transformed the diagnosis and management of brain tumors, which are among most challenging malignancies due to their high mortality rates complex neurological effects. Despite advancements surgery chemoradiotherapy, prognosis for glioblastoma multiforme (GBM) metastases remains poor, underscoring need innovative diagnostic strategies. This review highlights recent imaging techniques, liquid biopsies, artificial intelligence (AI) applications addressing current challenges. Advanced including diffusion tensor (DTI) magnetic resonance spectroscopy (MRS), improve differentiation tumor progression from treatment-related changes. Additionally, novel positron emission tomography (PET) radiotracers, such as 18F-fluoropivalate, 18F-fluoroethyltyrosine, 18F-fluluciclovine, facilitate metabolic profiling high-grade gliomas. Liquid biopsy, a minimally invasive technique, enables real-time monitoring biomarkers circulating DNA (ctDNA), extracellular vesicles (EVs), cells (CTCs), tumor-educated platelets (TEPs), enhancing precision. AI-driven algorithms, convolutional neural networks, integrate tools accuracy, reduce interobserver variability, accelerate clinical decision-making. These innovations advance personalized neuro-oncological care, offering new opportunities outcomes patients with central nervous system tumors. We advocate future research integrating these into workflows, accessibility challenges, standardizing methodologies ensure broad applicability neuro-oncology.

Language: Английский

---