Quantization of Deep Neural Networks for Medical Image Analysis: A Systematic Review and Meta-Analysis.

Neural network quantization has become established as a key strategy for tran- 1 sitioning medical imaging models from research environments to clinical devices and 2 resource-constrained edge platforms; however, the available evidence remains fragmented 3 and focused on highly heterogeneous use cases. This study presents a systematic review 4 of 72 studies on quantization applied to medical images, following PRISMA guidelines, 5 with the aim of characterizing the relationship among quantization technique, network 6 architecture, imaging modality, and execution environment, as well as their impact on 7 latency, memory footprint, and clinical deployment. Based on a structured variable matrix, 8 we analyze—through tailored visualizations—usage patterns of post-training quantization 9 (PTQ), quantization-aware training (QAT), mixed precision, and binary/low-bit schemes 10 across frameworks such as PyTorch, TensorFlow, and TensorFlow Lite, executed on server- 11 class GPUs, edge/embedded devices, and specialized hardware. The results reveal a strong 12 concentration of evidence in PyTorch/TensorFlow pipelines using INT8 or mixed precision 13 on GPUs and edge platforms, contrasted with limited attention to PACS/RIS interoperabil- 14 ity, model lifecycle management, energy consumption, cost, and regulatory traceability. We 15 conclude that, although quantization can approximate real-time performance and reduce 16 memory footprint, its clinical adoption remains constrained by integration challenges, 17 model governance requirements, and the maturity of the hardware–software ecosystem.