基于多区域影像组学的垂体瘤生物学行为预测研究进展

Advances in the Prediction of Biological Behavior of Pituitary Adenomas Based on Multi-Regional Radiomics

摘要: 垂体瘤(pituitary adenoma, PA)是鞍区最常见的原发性肿瘤，其生物学行为的异质性直接影响治疗策略的选择与预后判断。传统的肿瘤感兴趣区(intra-tumoral region of interest, ROI)影像组学虽已在垂体瘤的鉴别诊断、侵袭性预测、质地评估等方面取得一定进展，但单一瘤内区域难以全面反映肿瘤与周围微环境的相互作用及肿瘤内部的空间异质性。近年来，包括瘤周区域(peri-tumoral ROI)及基于体素聚类的肿瘤亚区域/生境(habitat)分析的多区域影像组学策略逐渐成为研究热点。该策略通过拓展影像组学的空间维度，从肿瘤内部不同功能亚区及瘤周微环境中提取更丰富的定量特征，为术前无创预测垂体瘤的生物学行为提供了新的技术路径。本文系统综述了多区域影像组学在垂体瘤分子分型与细胞谱系预测、肿瘤侵袭性评估、质地术前判断、血供特征分析及术后复发与预后预测等方面的研究进展。同时，分析了多区域策略相较于传统单一瘤内影像组学的增量价值，探讨了当前面临的瘤周区域边界划分缺乏统一标准、多中心数据异质性及模型可解释性不足等主要挑战。最后，展望了多中心外部验证、影像–分子–病理跨尺度多组学融合策略的深化及向临床转化等未来研究方向。

Abstract: Pituitary adenoma (PA) is the most common primary tumor in the sellar region, and the heterogeneity of its biological behavior directly guides the selection of treatment strategies and prognostic evaluation. Traditional radiomics based on a single intratumoral region of interest (ROI) has made considerable progress in the differential diagnosis, invasiveness prediction, and consistency evaluation of pituitary adenomas. However, a single intratumoral ROI is insufficient to fully reflect the interaction between tumors and the surrounding microenvironment, as well as the intratumoral spatial heterogeneity. In recent years, multi-regional radiomics strategies, including peritumoral ROI analysis and voxel clustering-based tumor subregion and habitat analysis, have emerged as a research focus. By expanding the spatial dimension of radiomic analysis, this method extracts abundant quantitative features from different intratumoral functional subregions and peritumoral microenvironments, providing a novel non-invasive technical approach for the preoperative prediction of biological behaviors of pituitary adenomas. This paper systematically reviews the research progress of multi-regional radiomics in the molecular subtyping and cell lineage prediction, invasiveness evaluation, preoperative tumor consistency judgment, vascular characteristic analysis, and postoperative recurrence and prognosis prediction of pituitary adenomas. Meanwhile, it analyzes the incremental value of multi-regional strategies compared with conventional single intratumoral radiomics. The current main challenges are also discussed, including inconsistent delineation criteria for peritumoral boundaries, multi-center data heterogeneity, and insufficient model interpretability. Finally, future research directions are envisioned, such as multi-center external validation, deepening of imaging-molecular-pathological cross-scale multi-omics integration, and translation into clinical practice.