用混合变压器模型结合临床数据，精准预测肺纤维化进展。
Prognostic Model for Idiopathic Pulmonary Fibrosis Using Context-Aware Sequential-Parallel Hybrid Transformer and Enriched Clinical Information
- 设计上下文感知的串并行混合变压器捕捉影像与临床特征
- 在Kaggle数据集上实现-6.508的拉普拉斯对数似然得分
- 适合关注疾病进展预测与医疗AI融合的研究者
特发性肺纤维化（IPF）是一种进行性疾病，会将肺组织不可逆地转化为致密的纤维结构，导致呼吸困难和慢性疲劳等严重症状。该疾病的异质性和复杂性，尤其是病情严重程度和进展速度的差异，使预测其未来进程成为一项复杂挑战。传统基于临床评估和影像学的方法难以充分捕捉疾病复杂性。本文使用Kaggle肺纤维化进展数据集，包含计算机断层扫描图像和临床信息，预测用力肺活量（FVC）变化——这一关键进展指标。所提出的方法采用上下文感知的序列-并行混合变压器模型，并结合临床信息增强。该方法在测试中取得-6.508的拉普拉斯对数似然得分，优于已有方法，展现出更强的预测能力。结果表明，先进深度学习技术可提供更准确、及时的预测，为IPF的诊断与管理带来变革性可能，有望改善患者预后并推动治疗发展。
原文摘要 · Abstract (English)
Idiopathic pulmonary fibrosis (IPF) is a progressive disease that irreversibly transforms lung tissue into rigid fibrotic structures, leading to debilitating symptoms such as shortness of breath and chronic fatigue. The heterogeneity and complexity of this disease, particularly regarding its severity and progression rate, have made predicting its future course a complex and challenging task. Besides, traditional diagnostic methods based on clinical evaluations and imaging have limitations in capturing the disease's complexity. Using the Kaggle Pulmonary Fibrosis Progression dataset, which includes computed tomography images, and clinical information, the model predicts changes in forced vital capacity (FVC), a key progression indicator. Our method uses a proposed context-aware sequential-parallel hybrid transformer model and clinical information enrichment for its prediction. The proposed method achieved a Laplace Log-Likelihood score of -6.508, outperforming prior methods and demonstrating superior predictive capabilities. These results highlight the potential of advanced deep learning techniques to provide more accurate and timely predictions, offering a transformative approach to the diagnosis and management of IPF, with implications for improved patient outcomes and therapeutic advancements.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。