2026· Journal of Machine Learning Innovations and Artificial Intelligence Horizons· 0 citations
TL;DR
It is concluded that while deep learning has achieved remarkable performance in many medical imaging benchmarks, substantial work remains to ensure generalization, interpretability, and ethical deployment in real-world clinicalsettings.
Abstract
Medical image analysis has witnessed substantial transformation through the application of deep learning models, yet the rapid proliferation of research in this domain poses significant challenges for synthesizing coherent insights. Our objective in this systematic literature review is to comprehensively map the landscape of deep learning approaches applied to medical imaging, with a focus on architectural innovations, task-specific solutions, learning paradigms, and emerging methodological trends. We conducted a structured review following established guidelines for systematic literature synthesis. The methodology involved a multi-stage screening process to identify relevant studies, followed by thematic categorization across eight dimensions, including model design, core tasks, data efficiency, explainability, clinical applications, pre-processing, emerging trends, and systemic challenges. Our analysis reveals that convolutional neural networks remain foundational, though transformer-based architectures and hybrid models are increasingly prevalent for tasks such as segmentation, classification, and detection. Data efficiency techniques, including self-supervised and few-shot learning, have become critical to address the scarcity of annotated medical datasets. We also observe a growing emphasis on explainability and uncertainty quantification to foster clinical trust, alongside rising concerns about privacy-preserving training and federated learning. The review further identifies persistent gaps, particularly in the validation of models across diverse populations and imaging modalities. We conclude that while deep learning has achieved remarkable performance in many medical imaging benchmarks, substantial work remains to ensure generalization, interpretability, and ethical deployment in real-world clinicalsettings. This systematic review provides a structured reference for researchers and practitioners navigating this interdisciplinary field.
Deep learning has revolutionized medical image analysis, playing a vital role in modern clinical applications. However, the deployment of large-scale models in real-world clinical settings remains challenging due to high computational costs, latency constraints, and patient data privacy concerns associated with cloud-based processing. To address these bottlenecks, this review provides a comprehensive synthesis of efficient and lightweight deep learning architectures specifically tailored for the medical domain. We categorize the landscape of modern efficient models into three primary streams: Convolutional Neural Networks (CNNs) for local inductive bias, Lightweight Transformers for global context under constrained attention, and emerging Linear Complexity Models for scalable global context. Furthermore, we examine key model compression strategies (including pruning, quantization, knowledge distillation, and low-rank factorization) and evaluate their efficacy in maintaining diagnostic performance while reducing hardware requirements. By identifying current limitations and discussing the transition toward on-device intelligence, this review serves as a roadmap for researchers and practitioners aiming to bridge the gap between high-performance AI and resource-constrained clinical environments.
C. M. Nguyen, Truong-Son Hy· Discover Artificial Intellig...· 0 citations
Accurate lung nodule segmentation is essential for early lung cancer diagnosis and treatment planning. Although deep learning (DL) has significantly advanced automated segmentation, existing review studies often lack an integrated technical–clinical perspective and provide limited guidance for real‐world deployment. In this study, we conduct a systematic literature review following PRISMA 2020 guidelines, analysing 311 publications and synthesizing 113 high‐quality studies published between 2019 and 2026. This review makes three key contributions. First, we provide a structured comparison of major DL architectures, including convolutional neural networks (CNNs), transformers, generative adversarial networks (GANs) and graph neural networks (GNNs), with respect to segmentation performance and computational efficiency. Second, we identify critical barriers to clinical adoption, such as data heterogeneity, annotation scarcity, high computational cost and limited model interpretability. Third, we analyse efficiency metrics (parameters, loss function and optimization) to categorize models suitable for point‐of‐care, resource‐constrained and mobile clinical environments. Based on these findings, we propose a roadmap for future research focusing on lightweight architectures, edge–cloud integration, federated learning and explainable AI (XAI). This study provides a clear and actionable framework to bridge the gap between DL‐based segmentation research and clinical deployment.
Muhammad Sufyan, Jun Qian, Jianqiang Li et al.· Expert systems· 0 citations
Empirical evidence from recent studies demonstrates that fine-tuning and network-based DTL strategies, including federated learning, consistently enhance diagnostic accuracy, robustness, and generalization across multiple medical imaging modalities, particularly in data-limited clinical scenarios.
M. A. S. Banu, A. Dhavapandiammal, K. Palanisamy· Current medical imaging· 0 citations
Transformer-based architectures have become central to medical image analysis, yet their practical value remains difficult to assess because studies vary widely in tasks, datasets, validation protocols, baselines, and reporting quality. This survey critically reviews recent transformer-based, hybrid, foundation, and transformer-alternative models across segmentation, classification, reconstruction, and image registration. A total of 128 studies published between 2021 and 2026 are organized using a task-, modality-, and architecture-aware taxonomy, with reported performance synthesized alongside baseline comparisons, reproducibility, computational cost, and clinical-readiness evidence. The findings indicate that the most convincing gains arise from task-adapted hybrid designs that combine local feature extraction with global context modeling, rather than from an unconditional superiority of transformers over convolutional networks. Persistent gaps include non-standardized benchmarks, limited external validation, incomplete code and weight availability, inconsistent efficiency reporting, weak uncertainty analysis, and insufficient clinical evaluation. Progress will require transparent reporting, multicenter validation, and clinically grounded assessment.
Sam Ansari, Nastaran Faraji, Luke K. Topham et al.· Frontiers in Artificial Inte...· 0 citations
Major deep learning architectures, including CNNs, residual networks, UNet, attention-based models, Vision Transformers, and hybrid approaches, along with their clinical applications are summarized and emerging directions such as self-supervised learning, Explainable AI, federated learning, and lightweight models are highlighted as promising approaches for more reliable and accessible medical image analysis.
Lakshmi Sai Anusha Dadi, Pravallika Devi Kommana· International Journal for Re...· 0 citations