15 Transparent and sustainable AI for brain tumor detection: from conventional to hybrid models in predictive healthcare
-
Kamini Lamba
Abstract
Detection of brain tumors is a serious problem in medical imaging, with the need for precise, interpretable, and cost-effective diagnostic techniques. This chapter provides an overview of the transition from traditional machine learning (ML) techniques to deep learning and hybrid artificial intelligence (AI) models for brain tumor categorization based on magnetic resonance imaging. Traditional ML methods – like support vector machines, random forests, and k-nearest neighbors – have interpretability but have limitations due to manual feature engineering and suboptimal performance on challenging imaging data. The advent of convolutional neural networks (CNNs) has greatly enhanced diagnostic performance via end-to-end learning from image inputs, although these models tend to lack transparency and require high computational power. To meet these challenges, hybrid AI models integrate CNN-based feature learning with traditional ML classifiers to achieve the best trade-off between accuracy, explainability, and computational cost. Furthermore, explainable AI methods – like Grad-CAM, LIME, and SHAP – allow model decisions to be visualized or quantified, promoting clinical trust and ethical deployment in healthcare. Along with technical precision, this chapter emphasizes the necessity for sustainability of AI-based healthcare. It addresses environmental, economic, and ethical aspects of sustainable AI with low-resource deployment, lifelong learning systems, and alignment with global AI ethics standards.
Abstract
Detection of brain tumors is a serious problem in medical imaging, with the need for precise, interpretable, and cost-effective diagnostic techniques. This chapter provides an overview of the transition from traditional machine learning (ML) techniques to deep learning and hybrid artificial intelligence (AI) models for brain tumor categorization based on magnetic resonance imaging. Traditional ML methods – like support vector machines, random forests, and k-nearest neighbors – have interpretability but have limitations due to manual feature engineering and suboptimal performance on challenging imaging data. The advent of convolutional neural networks (CNNs) has greatly enhanced diagnostic performance via end-to-end learning from image inputs, although these models tend to lack transparency and require high computational power. To meet these challenges, hybrid AI models integrate CNN-based feature learning with traditional ML classifiers to achieve the best trade-off between accuracy, explainability, and computational cost. Furthermore, explainable AI methods – like Grad-CAM, LIME, and SHAP – allow model decisions to be visualized or quantified, promoting clinical trust and ethical deployment in healthcare. Along with technical precision, this chapter emphasizes the necessity for sustainability of AI-based healthcare. It addresses environmental, economic, and ethical aspects of sustainable AI with low-resource deployment, lifelong learning systems, and alignment with global AI ethics standards.
Kapitel in diesem Buch
- Frontmatter I
- Preface V
- Contents VII
- List of contributors IX
- 1 Leveraging computational intelligence and mathematical modeling for sustainable future in agriculture: a unified paradigm for recognizing tomato leaf diseases 1
- 2 Digital dawn: how immersive technologies are shaping a sustainable future 37
- 3 Sustainable intelligence: ethical issues in the evolution of intelligent systems 57
- 4 Energy sustainability and computational intelligence based routing protocols in WSN: an analytical survey 75
- 5 Harnessing the metaverse for healthcare innovation: exploring predictive analytics and AI-driven personalization 91
- 6 Toward a sustainable future: a computational intelligence fusion framework of color and darknet features for the classification of crop leaf diseases 109
- 7 Sustainable computing approaches for complex medical image analysis: a neurodiagnostic perspective 139
- 8 Sustainability with artificial intelligence: obstacles, opportunities, and research agenda 161
- 9 Smart solutions to a sustainable future 177
- 10 Ethical issues in intelligent systems for sustainability 195
- 11 CPS: cyber-physical system security for the Industrial Internet of Things in smart grid 215
- 12 Integrating mathematical computing and deep learning for efficient monkeypox skin lesion detection: A pathway to sustainable health solutions 253
- 13 Leveraging heuristics based on CK Metrics Suite for quality enhancement in sustainable quantum software development 271
- 14 Intelligent systems for fire management and sustainability 297
- 15 Transparent and sustainable AI for brain tumor detection: from conventional to hybrid models in predictive healthcare 323
- Index 349
- Mathematical Methods in the Digital Age
Kapitel in diesem Buch
- Frontmatter I
- Preface V
- Contents VII
- List of contributors IX
- 1 Leveraging computational intelligence and mathematical modeling for sustainable future in agriculture: a unified paradigm for recognizing tomato leaf diseases 1
- 2 Digital dawn: how immersive technologies are shaping a sustainable future 37
- 3 Sustainable intelligence: ethical issues in the evolution of intelligent systems 57
- 4 Energy sustainability and computational intelligence based routing protocols in WSN: an analytical survey 75
- 5 Harnessing the metaverse for healthcare innovation: exploring predictive analytics and AI-driven personalization 91
- 6 Toward a sustainable future: a computational intelligence fusion framework of color and darknet features for the classification of crop leaf diseases 109
- 7 Sustainable computing approaches for complex medical image analysis: a neurodiagnostic perspective 139
- 8 Sustainability with artificial intelligence: obstacles, opportunities, and research agenda 161
- 9 Smart solutions to a sustainable future 177
- 10 Ethical issues in intelligent systems for sustainability 195
- 11 CPS: cyber-physical system security for the Industrial Internet of Things in smart grid 215
- 12 Integrating mathematical computing and deep learning for efficient monkeypox skin lesion detection: A pathway to sustainable health solutions 253
- 13 Leveraging heuristics based on CK Metrics Suite for quality enhancement in sustainable quantum software development 271
- 14 Intelligent systems for fire management and sustainability 297
- 15 Transparent and sustainable AI for brain tumor detection: from conventional to hybrid models in predictive healthcare 323
- Index 349
- Mathematical Methods in the Digital Age
