Establishment and preliminary validation of an artificial intelligence-assisted model for identification of perforators of anterolateral thigh perforator flaps based on preoperative indocyanine green angiography (ICGA) videos
-
摘要:
目的 建立基于吲哚菁绿血管造影（ICGA）视频的人工智能辅助股前外侧穿支皮瓣穿支识别模型并进行初步验证。 方法 该研究为横断面研究。2024年1—7月,苏州大学苏州瑞华骨科医院收治70例符合入选标准且接受股前外侧穿支皮瓣修复手术的患者,术前行彩色多普勒超声（CDU）和ICGA检查。采用Python3.13中的random函数选择50例患者作为训练集用于模型参数训练；剩余20例患者作为验证集,用于独立验证。以术中探查结果为金标准,预设每例患者股前外侧穿支皮瓣区域内有4个穿支定位点。统计训练集和验证集患者CDU、ICGA-人工判读、ICGA-人工智能辅助识别（ICGA-AI）穿支定位点与术中实际穿支入皮点距离≤1.0 cm的情况（真阳性）并计算阳性率。在训练集和验证集患者中,统计3种方法定位穿支的敏感度、特异度、阳性预测值、阴性预测值、准确率、平衡准确率,3种方法穿支定位结果与术中探查结果的一致性,测量3种方法确定的穿支定位点与对应的术中实际穿支入皮点之间的距离（误差距离）,记录ICGA-AI定位的优势穿支与术中选择的优势穿支是否一致并计算优势穿支定位准确率。 结果 3种方法定位训练集和验证集患者穿支的阳性率比较,差异均无统计学意义（P>0.05）。在训练集患者中,CDU、ICGA-人工判读、ICGA-AI定位穿支的敏感度、特异度、阳性预测值、阴性预测值、准确率、平衡准确率分别为88.70%、85.88%、89.47%、84.88%、87.50%、87.29%,93.91%、89.41%、92.31%、91.57%、92.00%、91.66%,82.61%、83.53%、87.16%、78.02%、83.00%、83.07%；CDU、ICGA-人工判读和ICGA-AI穿支定位结果与术中探查结果的一致性较强或几乎完全一致（κ分别为0.74、0.84、0.66,P<0.05）。在验证集患者中,CDU、ICGA-人工判读、ICGA-AI定位穿支的敏感度、特异度、阳性预测值、阴性预测值、准确率、平衡准确率分别为95.24%、78.95%、83.33%、93.75%、87.50%、87.09%,95.24%、92.11%、93.02%、94.59%、93.75%、93.67%,95.24%、94.74%、95.24%、94.74%、95.00%、94.99%；CDU、ICGA-人工判读和ICGA-AI穿支定位结果与术中探查结果的一致性较强或几乎完全一致（κ分别为0.75、0.88、0.90,P<0.05）。在训练集患者中,ICGA-AI穿支定位误差距离为0.84（0.70,0.98）cm,明显短于CDU的1.41（0.80,2.00）cm（Z=-2.56,P<0.05）和ICGA-人工判读的1.50（1.03,1.80）cm（Z=-5.88,P<0.05）。在验证集患者中,ICGA-AI穿支定位误差距离为0.52（0.35,0.71）cm,明显短于CDU的1.50（1.00,2.25）cm（Z=-3.23,P<0.05）和ICGA-人工判读的1.56（1.41,2.00）cm（Z=-3.52,P<0.05）。在训练集和验证集患者中,ICGA-AI定位优势穿支的准确率分别为82.00%（41/50）和90.00%（18/20）。 结论 本研究初步建立了基于ICGA视频的人工智能辅助股前外侧穿支皮瓣穿支识别方法,在验证集患者中显示出较小的定位误差和较高的优势穿支定位准确率,可辅助术前穿支定位。 Abstract:Objective To establish and preliminarily validate an artificial intelligence-assisted model for identification of perforators of anterolateral thigh perforator flaps based on preoperative indocyanine green angiography (ICGA) videos. Methods This study was a cross-sectional study. From January to July 2024, 70 patients who met the inclusion criteria and underwent anterolateral thigh perforator flap repair surgery were admitted to Suzhou Ruihua Orthopedic Hospital of Soochow University. All patients received preoperative color Doppler ultrasound (CDU) and ICGA examinations. Fifty patients were selected as the training set for model parameter training using the random function in Python 3.13, and the remaining 20 patients were assigned as the validation set for independent validation. With the intraoperative exploration result as the gold standard, 4 perforator sites were preset in the anterolateral thigh perforator flap area of each patient. In the training set and the validation set of patients, cases where the distance between the perforator localization points of CDU, ICGA-manual interpretation, and ICGA-based artificial intelligence (ICGA-AI) and the actual intraoperative cutaneous perforator entry point was ≤1.0 cm (defined as true positive) were calculated, and the positive rate was calculate. In the training set and validation set of patients, the sensitivity, specificity, positive predictive value, negative predictive value, accuracy, and balanced accuracy of the three methods for perforator localization were calculated, the consistency of the perforator localization results between the three methods and the intraoperative exploration; the distance between the perforator localization points determined by the three methods and the corresponding actual intraoperative cutaneous perforator entry point (error distance) was measured; the consistency between the dominant perforator localization by ICGA-AI and that determined intraoperatively was recorded, and the localization accuracy of the dominant perforator was calculated. Results There were no statistically significant differences in the positive rates of perforator localization among the three methods in both the training set and the validation set of patients (P>0.05). In the training set of patients, the sensitivity, specificity, positive predictive value, negative predictive value, accuracy, and balanced accuracy of CDU, ICGA-manual interpretation, and ICGA-AI were 88.70%, 85.88%, 89.47%, 84.88%, 87.50%, and 87.29%, 93.91%, 89.41%, 92.31%, 91.57%, 92.00%, and 91.66%, and 82.61%, 83.53%, 87.16%, 78.02%, 83.00%, and 83.07%, respectively; the consistency of the perforator localization results between the three methods and the intraoperative exploration was strong or almost perfect (with κ of 0.74, 0.84, and 0.66, respectively, P<0.05). In the validation set of patients, the sensitivity, specificity, positive predictive value, negative predictive value, accuracy, and balanced accuracy of CDU, ICGA-manual interpretation, and ICGA-AI were 95.24%, 78.95%, 83.33%, 93.75%, 87.50%, and 87.09%, 95.24%, 92.11%, 93.02%, 94.59%, 93.75%, and 93.67%, 95.24%, 94.74%, 95.24%, 94.74%, 95.00%, and 94.99%, respectively; the consistency of the perforator localization results between the three methods and the intraoperative exploration was strong or almost perfect (with κ of 0.75, 0.88, and 0.90, respectively, P<0.05). In the training set of patients, the error distance of perforator localization of ICGA-AI was 0.84 (0.70, 0.98) cm, which was significantly shorter than 1.41 (0.80, 2.00) cm of CDU (Z=-2.56, P<0.05) and 1.50 (1.03, 1.80) cm of ICGA-manual interpretation (Z=-5.88, P<0.05). In the validation set of patients, the error distance of perforator localization of ICGA-AI was 0.52 (0.35, 0.71) cm, which was significantly shorter than 1.50 (1.00, 2.25) cm of CDU (Z=-3.23, P<0.05) and 1.56 (1.41, 2.00) cm of ICGA-manual interpretation (Z=-3.52, P<0.05). In the training set and the validation set of patients, the accuracy of ICGA-AI for dominant perforator localization was 82.00% (41/50) and 90.00% (18/20), respectively. Conclusions This study preliminarily establishes an artificial intelligence-assisted method for identification of perforators of anterolateral thigh perforator flaps based on ICGA videos. In the validation set of patients, this method showed smaller localization errors and higher accuracy of dominant perforator localization, which can assist in preoperative perforator localization. -
Key words:
- Perforator flap /
- Artificial intelligence /
- Angiography /
- Indocyanine green /
- Anterolateral thigh flap /
- Wound repair
-
参考文献
(38) [1] 徐永清,唐举玉,刘元波,等.股前外侧皮瓣穿支解剖学特征与定位方法的专家共识(2024版)[J].中国临床解剖学杂志,2024,42(5):489-499.DOI: 10.13418/j.issn.1001-165x.2024.5.01. [2] SongYG,ChenGZ,SongYL.The free thigh flap: a new free flap concept based on the septocutaneous artery[J].Br J Plast Surg,1984,37(2):149-159.DOI: 10.1016/0007-1226(84)90002-x. [3] WeiFC,JainV,CelikN,et al.Have we found an ideal soft-tissue flap? An experience with 672 anterolateral thigh flaps[J].Plast Reconstr Surg,2002,109(7):2219-2226; discussion 2227-2230.DOI: 10.1097/00006534-200206000-00007. [4] KimataY,UchiyamaK,EbiharaS,et al.Anatomic variations and technical problems of the anterolateral thigh flap: a report of 74 cases[J].Plast Reconstr Surg,1998,102(5):1517-1523.DOI: 10.1097/00006534-199810000-00026. [5] WongCH,WeiFC.Anterolateral thigh flap[J].Head Neck,2010,32(4):529-540.DOI: 10.1002/hed.21204. [6] LakhianiC,LeeMR,Saint-CyrM.Vascular anatomy of the anterolateral thigh flap: a systematic review[J].Plast Reconstr Surg,2012,130(6):1254-1268.DOI: 10.1097/PRS.0b013e31826d1662. [7] LeeYC,ChenWC,ChouTM,et al.Anatomical variability of the anterolateral thigh flap perforators: vascular anatomy and its clinical implications[J].Plast Reconstr Surg,2015,135(4):1097-1107.DOI: 10.1097/PRS.0000000000001103. [8] SmithRK,WykesJ,MartinDT,et al.Perforator variability in the anterolateral thigh free flap: a systematic review[J].Surg Radiol Anat,2017,39(7):779-789.DOI: 10.1007/s00276-016-1802-y. [9] 董帅,王石,滕志成,等.吲哚菁绿血管造影联合彩色多普勒超声在股前外侧穿支皮瓣穿支定位中的临床应用效果[J].中华烧伤与创面修复杂志,2025,41(11):10.DOI: 10.3760/cma.j.cn501225-20240629-00254. [10] 王石,董帅,曹阳,等.高选择性动脉吲哚菁绿造影在游离股前外侧皮瓣设计中的应用[J].中华烧伤与创面修复杂志,2024,40(10):948-954.DOI: 10.3760/cma.j.cn501225-20240513-00174. [11] ThomasB,FalknerF,DidzunO,et al.Optimizing ALT flap harvest: the role of combined preoperative duplex ultrasound and intraoperative ICG angiography for perforator selection[J].Life (Basel),2025,15(4):620.DOI: 10.3390/life15040620. [12] ChungMJ,ChenWH,LuYC,et al.Artificial intelligence-assisted indocyanine green angiography for perforators identification in the anterolateral thigh flap[J].Ann Plast Surg,2026,96(2 Suppl 2):S25-S30.DOI: 10.1097/SAP.0000000000004545. [13] JiangZ,LeiY,ZhangL,et al.Automated quantitative analysis of blood flow in extracranial-intracranial arterial bypass based on indocyanine green angiography[J].Front Surg,2021,8:649719.DOI: 10.3389/fsurg.2021.649719. [14] ShadidO,SethI,CuomoR,et al.Artificial intelligence in microsurgical planning: a five-year leap in clinical translation[J].J Clin Med,2025,14(13):4574.DOI: 10.3390/jcm14134574. [15] LinTC,YangHA,HuangRW,et al.Artificial intelligence and machine learning in reconstructive microsurgery[J].Semin Plast Surg,2025,39(3):190-198.DOI: 10.1055/s-0045-1810062. [16] LandisJR,KochGG.The measurement of observer agreement for categorical data[J].Biometrics,1977,33(1):159-174. [17] MiyamotoS,FujisawaK,KuritaD,et al.New anatomic landmark for perforator of anterolateral thigh flap in preoperative ultrasonography[J].Laryngoscope,2025,135(10):3623-3629.DOI: 10.1002/lary.32223. [18] YuL,JiangY,ZhangJ,et al.Three dimensional-CTA versus Doppler ultrasound for anterolateral thigh flap reconstruction in limb defects: a retrospective study[J].Plast Reconstr Surg Glob Open,2025,13(11):e7308.DOI: 10.1097/GOX.0000000000007308. [19] WeiB,LuG,LiQ,et al.Finder-I for locating anterolateral thigh perforators in head and neck reconstruction: a diagnostic study[J].Head Neck,2025,47(11):3019-3027.DOI: 10.1002/hed.28227. [20] 张丕红.浅谈薄型穿支皮瓣切取与穿支血管探测和皮瓣血运评估[J].中华烧伤与创面修复杂志,2023,39(10):911-918.DOI: 10.3760/cma.j.cn501225-20230812-00047. [21] 邓呈亮,简喜超.浅谈显微外科技术和游离穿支皮瓣修复创面[J].中华烧伤与创面修复杂志,2026,42(5):421-427.DOI: 10.3760/cma.j.cn501225-20260224-00085. [22] YuP,YoussefA.Efficacy of the handheld Doppler in preoperative identification of the cutaneous perforators in the anterolateral thigh flap[J].Plast Reconstr Surg,2006,118(4):928-933.DOI: 10.1097/01.prs.0000232216.34854.63. [23] TsukinoA,KurachiK,InamiyaT,et al.Preoperative color Doppler assessment in planning of anterolateral thigh flaps[J].Plast Reconstr Surg,2004,113(1):241-246.DOI: 10.1097/01.PRS.0000095949.41413.88. [24] ChengHT,LinFY,ChangSC.Diagnostic efficacy of color Doppler ultrasonography in preoperative assessment of anterolateral thigh flap cutaneous perforators: an evidence-based review[J].Plast Reconstr Surg,2013,131(3):471e-473e.DOI: 10.1097/PRS.0b013e31827c733e. [25] 王晓冰,纪旭,马春燕.彩色多普勒超声在股前外侧穿支皮瓣移植术前穿支动脉评估中的作用[J].中国医科大学学报,2022,51(6):513-517.DOI: 10.12007/j.issn.0258-4646.2022.06.007. [26] MooreR,MullnerD,NicholsG,et al.Color Doppler ultrasound versus computed tomography angiography for preoperative anterolateral thigh flap perforator imaging: a systematic review and meta-analysis[J].J Reconstr Microsurg,2022,38(7):563-570.DOI: 10.1055/s-0041-1740958. [27] KimTH, ChaoJS, KwonJG, et al. Refining surgical precision: the impact of color doppler ultrasound-guided perforator mapping on anterolateral thigh flap for oncologic patients[J/OL]. J Reconstr Microsurg,2025[2026-05-31]. https://pubmed.ncbi.nlm.nih.gov/41360097/. DOI:10.1055/a-2751-8691. [published online ahead of print December 8, 2025]. [28] TangY,GuoQ,LiX,et al.Augmented reality-assisted systematic mapping of anterolateral thigh perforators[J].BMC Musculoskelet Disord,2022,23(1):1047.DOI: 10.1186/s12891-022-06013-1. [29] YuXX,YangSF,JiCS,et al.A novel computed tomography angiography technique: guided preoperative localization and design of anterolateral thigh perforator flap[J].Insights Imaging,2022,13(1):190.DOI: 10.1186/s13244-022-01318-0. [30] ZhangA,DongY,YuanS,et al.Preoperative perforator mapping of anterolateral thigh flaps using computed tomographic angiography and visual coordinate system for the reconstruction of head and neck defects[J].Ear Nose Throat J,2025,104(4):NP195-NP202.DOI: 10.1177/01455613221108367. [31] KimH,ChaIH,KimHJ,et al.Perforators detected in computed tomography angiography for anterolateral thigh free flap: am I the only one who feels inaccurate?[J].J Clin Med,2023,12(12):4139.DOI: 10.3390/jcm12124139. [32] 张月恒,崔文举,宋坤修,等.改良CT血管造影术辅助基于浅筋膜穿支的股前外侧皮瓣穿支评估与偏心设计的前瞻性研究[J].中华烧伤与创面修复杂志,2023,39(2):141-149.DOI: 10.3760/cma.j.cn501225-20220902-00376. [33] 尹其翔,糜菁熠,蔡华忠,等.基于多层CT血管成像的穿支皮瓣设计和切取的三维导航流程的建立及其临床应用效果[J].中华烧伤与创面修复杂志,2025,41(2):171-179.DOI: 10.3760/cma.j.cn501225-20240407-00123. [34] SingaraveluA,DalliJ,PotterS,et al.Artificial intelligence for optimum tissue excision with indocyanine green fluorescence angiography for flap reconstructions: proof of concept[J].JPRAS Open,2024,41:389-393.DOI: 10.1016/j.jpra.2024.07.014. [35] 赖子森,吕嘉晖,刘广文,等.超声引导下经皮经肝门静脉穿刺注射吲哚菁绿在腹腔镜解剖性肝切除术中的应用价值[J].中华消化外科杂志,2024,23(12):1544-1549.DOI: 10.3760/cma.j.cn115610-20241120-00500. [36] DidzunO,ThomasB,PanayiAC,et al.Indocyanine green fluorescence angiography as a predictor of distal flap necrosis in latissimus dorsi free flaps[J].Microsurgery,2025,45(2):e70019.DOI: 10.1002/micr.70019. [37] 刘一秀,汤喜,吴剑,等.基于混合现实与人工智能算法在股前外穿支血管定位中的前瞻性研究[J].华西口腔医学杂志,2024,(6):9.DOI: 10.7518/hxkq.2024.2024129. [38] LiuY,FanW,YuM,et al.Comparison between mixed reality with artificial algorithms and ultrasound in localization of anterior thigh flap perforators: a prospective randomized controlled study[J].BMC Med,2025,23(1):374.DOI: 10.1186/s12916-025-04181-0. -
图 1 术前CDU和ICGA-人工判读定位穿支指导股前外侧穿支皮瓣移植修复右小腿创面及ICGA-AI穿支识别效果。1A.清创后可见软组织缺损；1B.术前CDU（蓝色箭头）及ICGA-人工判读（白色箭头）定位穿支及设计股前外侧穿支皮瓣；1C.术中穿支（箭头所示）选择；1D.股前外侧穿支皮瓣切取；1E. ICGA-AI对ICGA视频进行离线分析并识别、排序候选穿支（红色表示优势穿支定位点,黄色表示候选穿支定位点）；1F.ICGA-AI识别的候选穿支定位点转换至股前外侧体表后的结果（红色表示优势穿支定位点,黄色表示候选穿支定位点）；1G.股前外侧穿支皮瓣移植修复右小腿创面术后即刻；1H.术后12 d,皮瓣顺利成活
注：CDU为彩色多普勒超声,ICGA为吲哚菁绿血管造影,ICGA-AI为ICGA及人工智能辅助识别
Table 1. 训练集50例接受股前外侧穿支皮瓣手术的患者采用3种方法定位穿支的效能
方法 穿支定位点（个） 敏感度（%） 特异度（%） 阳性预测值（%） 阴性预测值（%） 准确率（%） 平衡准确率（%） 真阳性 假阳性 真阴性 假阴性 CDU 102 12 73 13 88.70 85.88 89.47 84.88 87.50 87.29 ICGA-人工判读 108 9 76 7 93.91 89.41 92.31 91.57 92.00 91.66 ICGA-AI 95 14 71 20 82.61 83.53 87.16 78.02 83.00 83.07 注：CDU为彩色多普勒超声,ICGA为吲哚菁绿血管造影,ICGA-AI为ICGA-人工智能辅助识别；将术前定位点与术中实际穿支入皮点之间的距离≤1.0 cm判定为真阳性；若术中在该区域内未探查到穿支,而术前探查存在穿支定位点或术前定位点与术中实际穿支入皮点之间的距离>1.0 cm,则判定为假阳性；若术中在该区域内探查到穿支,而术前探查不存在穿支定位点,则判定为假阴性；若术中在该区域内未探查到穿支且术前探查不存在穿支定位点,则判定为真阴性 
下载: 导出CSV
Table 2. 验证集20例接受股前外侧穿支皮瓣手术的患者采用3种方法定位穿支的效能
方法 穿支定位点（个） 敏感度（%） 特异度（%） 阳性预测值（%） 阴性预测值（%） 准确率（%） 平衡准确率（%） 真阳性 假阳性 真阴性 假阴性 CDU 40 8 30 2 95.24 78.95 83.33 93.75 87.50 87.09 ICGA-人工判读 40 3 35 2 95.24 92.11 93.02 94.59 93.75 93.67 ICGA-AI 40 2 36 2 95.24 94.74 95.24 94.74 95.00 94.99 注：CDU为彩色多普勒超声,ICGA为吲哚菁绿血管造影,ICGA-AI为ICGA-人工智能辅助识别；将术前定位点与术中实际穿支入皮点之间的距离≤1.0 cm判定为真阳性；若术中在该区域内未探查到穿支,而术前探查存在穿支定位点或术前定位点与术中实际穿支入皮点之间的距离>1.0 cm,则判定为假阳性；若术中在该区域内探查到穿支,而术前探查不存在穿支定位点,则判定为假阴性；若术中在该区域内未探查到穿支且术前探查不存在穿支定位点,则判定为真阴性
下载: 导出CSV
-
下载:
图(2) / 表(2)
计量
- 文章访问数: 0
- HTML全文浏览量: 0
- PDF下载量: 0
- 被引次数: 0
