Document Type : Systematic Review and Meta-analysis
Authors
1
Department of Pathology, School of Medicine, Iran University of Medical Sciences, Tehran, Iran.
2
Department of Radiology, School of Medicine, Iran University of Medical Sciences, Tehran, Iran.
3
Department of Molecular Biotechnology, Anhalt University of Applied Sciences, Köthen, Germany.
4
Department of Ophthalmology, Imam Hossein Hospital, Shahid Beheshti University of Medical Sciences, Tehran, Iran.
5
Department of Ophthalmology, Farabi Hospital, Tehran University of Medical Sciences, Tehran, Iran.
6
Student Research Committee, Ilam University of Medical Sciences, Ilam, Iran.
7
Department of Obstetrics and Gynecology, School of Medicine, Iran University of Medical Sciences, Tehran, Iran.
8
Department of Colorectal Surgery, AJA University of Medical Sciences, Tehran, Iran.
9
Department of Plastic Surgery, Iranshahr University of Medical Sciences, Iranshahr, Iran.
10
Breast Health and Cancer Research Center, Iran University of Medical Sciences, Tehran, Iran.
11
Department of Radiology, School of Medicine, Shahid Sadoughi University of Medical Sciences, Yazd, Iran.
12
Infectious Diseases Research Center, Shahid Sadoughi Hospital, Shahid Sadoughi University of Medical Sciences, Yazd, Iran.
13
Clinical Research Development Center, Shahid Sadoughi Hospital, Shahid Sadoughi University of Medical Sciences, Yazd, Iran.
Abstract
Graphics processing unit (GPU)-accelerated artificial intelligence has fundamentally transformed cancer diagnosis, imaging, and treatment planning by delivering unprecedented computational performance and clinical efficiency across multiple oncological domains. This comprehensive review demonstrates that GPU-optimized AI platforms, including NVIDIA Clara and MONAI frameworks, have achieved remarkable performance improvements ranging from 8x to 65x acceleration in cancer genomics and computational biology applications, while simultaneously reducing operational costs by up to 85%. In medical imaging applications, GPU-based systems have revolutionized cone-beam computed tomography reconstruction, achieving reconstruction times of 77-130 seconds compared to conventional approaches that require significantly longer processing periods, while enabling dramatic radiation dose reductions of 36-72 times without compromising diagnostic image quality. Digital pathology applications have benefited from GPU acceleration through enhanced histopathological image analysis capabilities, including automated gland segmentation for colorectal cancer grading and uncertainty quantification mechanisms that support clinical decision-making processes. The integration of GPU-accelerated AI tools into clinical workflows has enabled real-time processing of complex medical data, automated tumor contouring for radiation therapy planning, and sophisticated radiomics feature extraction that correlates imaging biomarkers with genetic and molecular tumor characteristics. These technological advances represent a paradigm shift toward precision oncology, where data-driven insights augment clinical expertise and reduce cognitive burden associated with complex oncological cases, ultimately enhancing diagnostic accuracy, treatment efficacy, and patient outcomes across diverse healthcare settings.
Keywords
Main Subjects
)