Research and implementation of high-precision medical image segmentation technology based on deep residual u-net++ and adaptive dynamic planning matching algorithm

Medical image segmentation plays a vital role in disease diagnosis, treatment planning, and follow-up procedures. However, achieving high-precision segmentation in complex anatomical structures remains a challenge due to image noise, inter-subject variability, and inconsistent imaging modalities. This study proposes a novel hybrid framework combining a Deep Residual U-Net++ architecture with an Adaptive Dynamic Planning Matching Algorithm (ADPMA) to enhance segmentation accuracy and robustness. The Deep Residual U-Net++ improves feature learning via nested and residual connections, while the ADPMA dynamically adjusts segmentation parameters based on image-specific characteristics. The model was evaluated on public MRI and CT datasets, including BraTS and ISIC, using Dice Similarity Coefficient (DSC), Intersection over Union (IoU), and pixel accuracy metrics. Compared to standard U-Net and U-Net++ architectures, the proposed method achieved a DSC of 0.89, IoU of 0.80, and pixel accuracy of 96.3%, demonstrating significant improvements in segmenting complex and noisy medical images.