Previous studies have suggested that factors such as the treatment interval and aetiology may influence the initial response rate to first-line treatment for infantile epileptic spasms syndrome (IESS). However, few children with IECSS have undergone clinically accessible tests to determine the aetiology. Using a dataset from our previously published research, we constructed and tested a predictive model for the initial response to first-line treatment in children with IESS. Random sampling and 5-fold cross-validation were performed, with synthetic minority oversampling technique to correct data imbalance. Machine learning algorithms and evaluation metrics optimised model accuracy and efficacy. This study included 532 children with IESS who had completed monotherapy first-line treatment, of whom 160 achieved an initial response. The model’s accuracy, F1 score, and area under the curve (AUC) in the validation set were 0.7836 ± 0.0229 (ranging from 0.75167 to 0.80536), 0.7833 ± 0.0229 (ranging from 0.75145 to 0.80531), and 0.8516 ± 0.0165 (ranging from 0.82468 to 0.86936), respectively. Factors such as the age of seizure onset, age of spasm onset, lead time, MRI subtype, treatment choice, and age at treatment consistently ranked in the top six for importance in contributing to the model. The study findings suggest that this model may help effectively predict the initial response to first-line treatment, supporting clinical decision-making for children with IESS. Key predictors such as the age of seizure onset and MRI subtype enable early, data-driven intervention strategies in clinical practice.

Extracting relation quintuple and feature words from unstructured text is a prelude to the construction of the scientific knowledge base. At present, the prior works use explicit clues between entities to study this task but ignore the use and the association of the feature words. In this work, we propose a new method to generate self-adaptive feature words from the original text for every single sample. These words can add additional correlation information to the knowledge graph. We allow the model to generate a new word representation and apply it to the original sentence to judge the relation type and locate the head and tail of the relation quintuple. Compared with the previous works, the feature words increase the flexibility of relying on information and improve the explanatory ability. Extensive experiments on scientific field datasets illustrate that the self-adaptive feature words method (SAFW) is good at ferreting out the unique feature words and obtaining the core part for the quintuple. It achieves good performance on four public datasets and obtains a markable performance improvement compared with other baselines.

Postoperative pulmonary complications (PPCs) are a major cause of morbidity and mortality following thoracic surgery, making accurate perioperative risk prediction crucial for clinical decision-making. However, existing predictive models mainly rely on structured clinical data and fail to effectively capture the complex correlations across heterogeneous medical data modalities. To address this issue, this paper proposes a novel deep learning framework based on a hierarchical cross-modal attention mechanism that emulates the clinical reasoning process. Anchoring the fusion process on structured tabular features, the model gradually integrates the feature representations of unstructured clinical notes and chest CT images, thus realizing the organic integration of three complementary medical data modalities. Experimental results on a retrospective multimodal dataset of thoracic surgery patients show that the proposed method achieves an Area Under the Curve (AUC) of 0.832 and an accuracy of 80.3%, outperforming several traditional machine learning baseline models. The study confirms that the hierarchical fusion strategy can effectively capture complementary information across multiple modalities and significantly improve the performance of PPC prediction. The method proposed in this paper constructs a robust and interpretable multimodal framework for clinical risk assessment, which has practical application potential for optimizing perioperative decision support in thoracic surgery.

Large language models (LLMs) are advancing intelligent writing systems from local text continuation and language polishing toward long-form structured text generation. However, directly generating full-length academic paper drafts remains challenging due to unclear research objectives, unstable discourse structures, insufficient long-text coherence, and the lack of explicit quality control mechanisms. To address this long-form structured generation task, we propose MetricDraft, a metric-driven framework for academic paper draft generation. The framework organizes the drafting process as a closed-loop pipeline comprising research ideation clarification, structural anchoring, section-by-section generation, quality assessment, and feedback-driven revision. Its key components include adversarial research ideation clarification, staged structural anchoring, the PRISM structured metric system, progressive context injection with section-type-aware guided generation (PCI+STAGG), and a metric-feedback-driven generation–evaluation co-optimization mechanism. Experimental results demonstrate that MetricDraft achieves higher composite quality scores than one-shot generation, summary-based context passing, and context-accumulation-only baselines, improving MQS over Base1, Base2, and Base3 by +5.5, +7.9, and +7.0 points, respectively, with paired tests reaching statistical significance. To examine whether this advantage is tied to a single LLM backend, we further conduct a cross-model validation on all 15 tasks using Qwen3.7-Max in addition to the original DeepSeek-V4-Pro setting. MetricDraft remains the best-performing strategy under both models. To address citation reliability, an additional citation verification-and-retrieval-based replacement (CVRR) experiment reduces the fabricated citation rate of DeepSeek MetricDraft drafts from 56.0% to 15.0%. Furthermore, PRISM exhibits moderate-to-high positive correlations with expert ratings, providing preliminary evidence that it can serve as an auxiliary evaluation reference for draft quality diagnosis and iterative revision. This work reformulates academic writing as an adjustable, assessable, and iteratively optimizable long-form structured text generation problem, offering methodological insights for human–AI collaborative writing and intelligent text generation system design.