目的·探讨基于影像后处理和3D打印技术设计的改良病理切片盒运用于骨肿瘤并辅助人工智能（artificial intelligence，AI）拼接病理图像的可行性。方法·收集于2022年2月至2022年8月间以骨肿瘤收治并行切除重建术的患者，对患者术前及其术后肿瘤标本行CT和MRI检查，以骨肿瘤的骨性结构为标志点，实现术前患者CT、MRI图像及术后肿瘤标本CT图像间的配准。设计经过改良的个性化病理切片盒并3D打印，指导病理切片方向，网格状分割病理大切片后行苏木精-伊红染色。借助AI拼接病理切片图像，并与术前肿瘤影像边界进行共定位比较。结果·收集4例使用改良病理切片盒的骨肿瘤患者（骨转移瘤2例，骨肉瘤1例，软骨肉瘤1例），男性3例，女性1例，平均年龄（40.25±25.70）岁。骨肿瘤累及股骨3例，髂骨1例。骨肿瘤最大直径为（12.10±4.02）cm。改良后的3D打印病理切片盒能较好地拟合骨肿瘤表面，解决了骨肿瘤标本在初代病理切片盒内移动度过大而导致的切割困难或切割定位不准确的问题，其能完整获得肿瘤三维病理边界的信息，并能与术前的影像肿瘤边界进行共定位比较。切片盒逐层各异的轮廓特征信息能辅助AI筛选并还原切片盒内病理碎片的相关关系，图像拼接时间由70余小时缩短为1 h，边缘符合率增加至90%。结论·基于3D打印技术的改良病理切片盒能准确辅助AI拼接病理图像，大幅提升病理图像拼接效率，为实现骨肿瘤的术前影像和病理边界精准对应提供数据支持。

Objective ·To explore the feasibility of applying improved pathological section boxes based on image post-processing and 3D printing technology to assisting artificial intelligence (AI) in stitching pathological images of bone tumors. Methods ·Bone tumor patients who underwent tumor resection surgery between February 2022 and August 2022 were enrolled. Patients and their postoperative tumor specimens were examined by CT and MRI. The skeletal anatomical landmarks of bone tumors were used to register the preoperative CT, and MRI and the postoperative CT images. The personalized pathological slice box was improved and 3D printed to guide the cutting direction of bone tumor specimens. The large pathological slices were grid-like segmented and performed with hematoxylin-eosin staining. After AI stitching the pathological images, the pathological tumor boundaries were compared with the preoperative radiological boundaries. Results ·Four patients with bone tumors (2 bone metastatic neoplasm, 1 osteosarcoma, and 1 chondrosarcoma; 3 males and 1 female) were collected to design the pathological section box with an average age of (40.25±25.70) years. The lesions included 3 cases of femur and 1 case of ilium. The mean maximum tumor diameter was (12.10±4.02) cm. The modified 3D printed pathological section box could fit the surface of bone tumor individually,and overcame the problems of difficult cutting or inaccurate cutting location caused by excessive movement of bone tumor specimens in the primary pathological section box. The pathological boundaries could be completely obtained and compared with the preoperative MRI boundaries for colocalization. The outline features of the section boxes could help AI to restore the permutation of pathological fragments, and the image stitching time decreased from the previous seventy hours to one hour while the boundary coincidence rates increased to 90%. Conclusion ·The improved pathological section box of 3D printing can accurately assist AI in stitching pathological images, greatly improve the efficiency of pathological image stitching and achieve the colocalization between preoperative MRI and pathological images of bone tumor specimens.