This study addressed the core challenge of intelligent pest and disease monitoring and early warning in smart horticultural production by proposing a multimodal deep learning framework based on multi-parameter environmental sensor arrays. The framework integrates visual information with electrical signals to overcome the inherent limitations of conventional single-modality approaches in terms of real-time capability, stability, and early detection performance. A long-term field experiment was conducted over 18 months in the Hetao Irrigation District of Bayannur, Inner Mongolia, using three representative horticultural crops—grape (Vitis vinifera), tomato (Solanum lycopersicum), and sweet pepper (Capsicum annuum)—to construct a multimodal dataset comprising illumination intensity, temperature, humidity, gas concentration, and high-resolution imagery, with a total of more than $2.6\times {10}^6$ recorded samples. The proposed framework consists of a lightweight convolution–Transformer hybrid encoder for electrical signal representation, a cross-modal feature alignment module, and an early-warning decision module, enabling dynamic spatiotemporal modeling and complementary feature fusion under complex field conditions. Experimental results demonstrated that the proposed model significantly outperformed both unimodal and traditional fusion methods, achieving an accuracy of $0.921$, a precision of $0.935$, a recall of $0.912$, an F1-score of $0.923$, and an area under curve (AUC) of $0.957$, confirming its superior recognition stability and early-warning capability. Ablation experiments further revealed that the electrical feature encoder, cross-modal alignment module, and early-warning module each played a critical role in enhancing performance. This research provides a low-cost, scalable, and energy-efficient solution for precise pest and disease management in intelligent horticulture, supporting efficient monitoring and predictive decision-making in greenhouses, orchards, and facility-based production systems. It offers a novel technological pathway and theoretical foundation for artificial-intelligence-driven sustainable horticultural production.

In precision oncology research, achieving joint modeling of tumor grading and treatment response, together with interpretable mechanism analysis, based on multimodal medical imaging and clinical data remains a challenging and critical problem. From a sensing perspective, these imaging and clinical data can be regarded as heterogeneous sensor-derived signals acquired by medical imaging sensors and clinical monitoring systems, providing continuous and structured observations of tumor characteristics and patient states. Existing approaches typically rely on invasive pathological grading, while grading prediction and treatment response modeling are often conducted independently. Moreover, multimodal fusion procedures generally lack explicit structural constraints, which limits their practical utility in clinical decision-making. To address these issues, a grade-guided multimodal collaborative modeling framework was proposed. Built upon mature deep learning models, including 3D ResNet-18, MLP, and CNN–Transformer, tumor grading was incorporated as a weakly supervised prior into the processes of multimodal feature fusion and treatment response modeling, thereby enabling an integrated solution for non-invasive grading prediction, treatment response subtype discovery, and intrinsic mechanism interpretation. Through a grade-guided feature fusion mechanism, discriminative information that is highly correlated with tumor malignancy and treatment sensitivity is emphasized in the multimodal joint representation, while irrelevant features are suppressed to prevent interference with model learning. Within a unified framework, grading prediction and grade-conditioned treatment response modeling are jointly realized. Experimental results on real-world clinical datasets demonstrate that the proposed method achieved an accuracy of $84.6\%$ and a kappa coefficient of $0.81$ in the tumor-grading prediction task, indicating a high level of consistency with pathological grading. In the treatment response prediction task, the proposed model attained an AUC of $0.85$, a precision of $0.81$, and a recall of $0.79$, significantly outperforming single-modality models, conventional early-fusion models, and multimodal CNN–Transformer models without grading constraints. In addition, treatment-sensitive and treatment-resistant subtypes identified under grading conditions exhibited stable and significant stratification differences in clustering consistency and survival analysis, validating the potential value of the proposed approach for clinical risk assessment and individualized treatment decision-making.