Traditional disease detection methods typically depend on visual assessments conducted by human experts, which are time-consuming and subjective. Thus, there is an urgent demand for automated and efficient approaches to accurately detect and classify tea diseases. This study presents an enhanced Real-Time Detection Transformer (RT-DETR), tailored for the accurate and efficient identification of tea diseases in natural environments. The proposed method integrates three novel components: Faster-LTNet, CG Attention Module, and RMT Spatial Prior Block, to significantly improve computational efficiency, feature representation, and detection capabilities. Faster-LTNet employs partial convolution and hierarchical design to optimize computational resources, while the CG Attention Module enhances multi-head self-attention by introducing grouped feature inputs and cascading operations to reduce redundancy and increase attention diversity. The RMT Spatial Prior Block integrates a Manhattan distance-based spatial decay matrix and linear decomposition strategy to improve global and local context modeling, reducing attention complexity. The enhanced RT-DETR model achieves a detection precision of 89.20% and a processing speed of 346.40 FPS. While the precision improves, the FPS value also increases by 109, which is superior to the traditional model in terms of precision and real-time processing. Additionally, compared to the baseline model, the FLOPs are reduced by 50%, and the overall model size and parameter size are decreased by approximately 50%. These findings indicate that the proposed algorithm is well-suited for efficient, real-time, and lightweight agricultural disease detection.

Accurate medical image segmentation is a core task in biomedical engineering and computer-assisted image analysis. Prompt-based segmentation models, such as Med-SAM, allow anatomical structures or lesions to be delineated from sparse user inputs, but two-dimensional clinical images often contain small targets, weak boundaries, heterogeneous lesion textures, and modality-specific noise. These factors can reduce the reliability of prompt-conditioned masks when early structural cues are weakened during deep encoder propagation. We propose HBF-BCER MedSAM, a prompt-preserving enhancement framework that primarily strengthens MedSAM image-encoder representations while retaining the original point- and box-prompt interface. Historical Branch Fusion (HBF) reinjects cross-layer structural summaries from previous enhanced blocks, whereas Balanced Convolutional Expert Routing (BCER) performs soft pixel-wise weighting of heterogeneous convolutional experts in bottleneck space under a load-balancing constraint. Experiments on REFUGE2, ISIC2016, and TNMIX under matched simulated-prompt protocols show consistent improvements over the matched MedSAM baseline. Dice/IoU increased from 82.1%/73.8% to 91.0%/83.7% for REFUGE2 optic cup segmentation, from 87.5%/78.6% to 95.9%/92.4% on ISIC2016, and from 81.6%/75.1% to 91.4%/85.1% on TNMIX. Qualitative comparisons, convergence analysis, and ablation studies further support the effectiveness of preserving historical structural evidence and applying adaptive local expert weighting for prompt-conditioned medical image segmentation.

Promptable medical segmentation models can be sensitive to small targets, weak boundaries, heterogeneous texture, and inaccurate user prompts. We propose HBF-BCER, a prompt-preserving adaptation of the SAM ViT-B architecture for medical image segmentation. Historical Branch Fusion (HBF) reinjects aligned bottleneck summaries from previous enhanced encoder blocks, while Balanced Convolutional Expert Routing (BCER) applies load-balanced, pixel-wise soft weighting to heterogeneous convolutional experts. The evaluation uses fixed held-out evaluation sets, three training seeds, matched original SAM ViT-B, image-encoder and mask-decoder full fine-tuning, decoder-adapter, LoRA, and DoRA baselines, repeated prompt perturbations, paired statistics, and internal module ablations. HBF-BCER achieved Dice scores of $96.69\pm 0.12\%$ and $92.24\pm 0.18\%$ for REFUGE2 disc and cup, $96.18\pm 0.31\%$ for ISIC2016, and $92.08\pm 0.19\%$ for TNMIX. Image-encoder and mask-decoder full fine-tuning produced slightly higher Dice for the REFUGE2 disc and TNMIX, whereas HBF-BCER produced higher Dice for the REFUGE2 cup and ISIC2016 and the lowest HD95 on all four targets. Among the four paired comparisons with available per-image outputs, HBF-BCER showed significant gains for REFUGE2 cup versus SAM + Adapter and LoRA and for TNMIX versus SAM + Adapter; the ISIC2016 difference versus full fine-tuning was not significant. The method reduced trainable parameters by 81.5% relative to image-encoder and mask-decoder full fine-tuning but increased total parameters, FLOPs, and inference latency. These results support HBF-BCER as a parameter-efficient training strategy rather than a generally lightweight or uniformly superior model.