何付權(quán) 李亞飛 楊嘉 劉曉強(qiáng) 朱磊
【摘要】 目的:探究人工智能(AI)輔助冠脈CT血管成像(CTA)“雙低”成像診斷冠狀動(dòng)脈病變的可行性。方法:選取2022年7月—2023年6月于上海交通大學(xué)附屬第一人民醫(yī)院就診的116例疑似冠狀動(dòng)脈病變患者,所有患者均行冠脈CTA“雙低”成像診斷,圖像均經(jīng)醫(yī)生診斷和AI輔助診斷,比較兩種方式的診斷時(shí)間及對(duì)圖像質(zhì)量評(píng)估結(jié)果,以選擇性冠狀動(dòng)脈造影術(shù)(CAG)結(jié)果為金標(biāo)準(zhǔn),比較兩種方式對(duì)冠狀動(dòng)脈病變的診斷價(jià)值。以醫(yī)生診斷結(jié)果為標(biāo)準(zhǔn),采用Kappa檢驗(yàn)分析AI輔助診斷與醫(yī)生診斷對(duì)冠狀動(dòng)脈狹窄程度、冠狀動(dòng)脈斑塊診斷的一致性。結(jié)果:AI組診斷時(shí)間短于醫(yī)生組,圖像評(píng)優(yōu)率高于醫(yī)生組(P<0.05);醫(yī)生組和AI組診斷對(duì)冠狀動(dòng)脈病變的準(zhǔn)確率分為93.97%、93.10%,敏感度分別為95.33%、94.39%,特異度分別為77.78%、77.78%,差異均無(wú)統(tǒng)計(jì)學(xué)意義(P>0.05);Kappa檢驗(yàn)結(jié)果顯示,醫(yī)生組與AI組對(duì)冠狀動(dòng)脈狹窄程度、冠狀動(dòng)脈斑塊診斷的Kappa值分別為0.954、0.922,具有高度一致性。結(jié)論:AI輔助冠脈CTA“雙低”成像對(duì)冠狀動(dòng)脈病變有較高的診斷價(jià)值,能顯著縮短冠狀動(dòng)脈病變的診斷時(shí)間,可作為醫(yī)生診斷冠狀動(dòng)脈病變的有效輔助工具。
【關(guān)鍵詞】 冠狀動(dòng)脈病變 人工智能 冠脈CT血管成像 輔助診斷 可行性
Feasibility Study of Artificial Intelligence-assisted Coronary CTA Imaging under "Double Low Scheme" in the Diagnosis of Coronary Artery Lesion/HE Fuquan, LI Yafei, YANG Jia, LIU Xiaoqiang, ZHU Lei. //Medical Innovation of China, 2024, 21(13): -134
[Abstract] Objective: To explore the feasibility of artificial intelligence (AI)-assisted coronary CT angiography (CTA) imaging under "double low scheme" in the diagnosis of coronary artery lesion. Method: A total of 116 patients with suspected coronary artery lesion in the First People's Hospital Affiliated to Shanghai Jiao Tong University from July 2022 to June 2023 were selected. All patients underwent coronary CTA imaging diagnosis under "double low scheme". The images were diagnosed by doctors and AI-assisted diagnosis. The diagnosis time and image quality assessment results of the two methods were compared. Selective coronary angiography (CAG) result was used as the gold standard to compare the diagnostic value of the two methods on coronary artery disease. Based on the doctor's diagnosis result, Kappa test was used to analyze the consistency of AI-assisted diagnosis and doctor's diagnosis in the diagnosis of coronary artery stenosis degree and coronary plaque. Result: The diagnosis time in AI group was shorter than that in doctor group, and the rate of excellent image assessment was higher than that in doctor group (P<0.05). The accuracy rates of diagnosis of coronary artery disease in doctor group and AI group were 93.97% and 93.10%, the sensitivities were 95.33% and 94.39%, and the specificities were 77.78% and 77.78% respectively, the differences were no significant differences (P>0.05). Kappa test results showed that the Kappa values of doctor group and AI group for the diagnosis of coronary artery stenosis degree and coronary plaque were 0.954 and 0.922 respectively, with high consistency. Conclusion: AI-assisted coronary CTA imaging under "double low scheme" has a high diagnostic value on coronary artery lesion, and it can significantly shorten the diagnosis time of coronary artery lesion and can be used as an effective auxiliary tool for doctors to diagnose coronary artery lesion.
[Key words] Coronary artery lesion Artificial intelligence Coronary CT angiography Assisted diagnosis Feasibility
First-author's address: Department of Radiology, the First People's Hospital Affiliated to Shanghai Jiao Tong University, Shanghai 201620, China
doi:10.3969/j.issn.1674-4985.2024.13.030
冠狀動(dòng)脈病變是指冠狀動(dòng)脈供血不足或阻塞的疾病,其病因主要是動(dòng)脈粥樣硬化,即血管壁內(nèi)脂質(zhì)沉積形成斑塊,逐漸增大堵塞血管腔[1]。冠狀動(dòng)脈病變臨床表現(xiàn)為胸痛、氣短、胸悶、心悸、乏力等,發(fā)病通常與高血脂、高血壓、糖尿病、吸煙、肥胖、炎癥反應(yīng)和家族遺傳等因素有關(guān),近年來(lái),其發(fā)病率和病死率上升,且呈年輕化趨勢(shì)發(fā)展,因此,早期診斷和臨床干預(yù)極其重要[2-3]。冠狀動(dòng)脈病變?cè)\斷通常以冠狀動(dòng)脈造影(CAG)為金標(biāo)準(zhǔn),但該法為有創(chuàng)檢查,且檢查結(jié)果較為復(fù)雜,使其在臨床診斷中的應(yīng)用受到限制[4]。冠脈CT血管(CTA)“雙低”成像是一種非侵入性檢查方法,掃描中使用低劑量成像和低噪聲成像,以提高對(duì)冠狀動(dòng)脈的分辨率和顯示能力,有助于準(zhǔn)確診斷冠狀動(dòng)脈病變,但圖像處理依賴(lài)于醫(yī)生,仍需人力資源支撐[5]。隨著醫(yī)學(xué)人工智能(AI)發(fā)展,AI輔助冠脈CTA診斷疾病已成為研究熱點(diǎn)。本文旨在探究人工智能輔助冠脈CTA“雙低”成像診斷冠狀動(dòng)脈病變的可行性,以期為冠狀動(dòng)脈病變臨床診斷提供支持。
1 資料與方法
1.1 一般資料
選取2022年7月—2023年6月于上海交通大學(xué)附屬第一人民醫(yī)院就診的116例疑似冠狀動(dòng)脈病變患者。納入標(biāo)準(zhǔn):(1)臨床表現(xiàn)為胸悶、氣短、心悸等癥狀;(2)血流動(dòng)力學(xué)異常,疑似冠狀動(dòng)脈狹窄;(3)均行CAG及冠脈CTA“雙低”成像檢查,兩種檢查間隔<2周;(4)年齡≥18歲。排除標(biāo)準(zhǔn):(1)對(duì)檢查所用試劑過(guò)敏;(2)安裝心臟起搏器、搭橋術(shù);(3)冠脈CTA“雙低”成像所得圖像存在偽影,影響診斷;(4)嚴(yán)重心律不齊、心率過(guò)快;(5)AI軟件無(wú)法自動(dòng)計(jì)算已完成的病例;(6)無(wú)法配合完成檢查。116例患者中男75例,女41例;平均年齡(62.43±6.17)歲;平均體重指數(shù)(23.77±2.24)kg/m2。研究已通過(guò)上海交通大學(xué)附屬第一人民醫(yī)院醫(yī)學(xué)倫理委員會(huì)審核?;颊呋蚣覍僦橥?。
1.2 方法
CAG檢查:采用FD20數(shù)字血管造影機(jī)(飛利浦),對(duì)比劑選用400 mg/mL的碘美普爾注射液(生產(chǎn)廠家:上海博萊科信誼藥業(yè)有限責(zé)任公司,批準(zhǔn)文號(hào):國(guó)藥準(zhǔn)字J20150053,規(guī)格:100 mL︰40 g)。術(shù)前檢查后,局部麻醉,經(jīng)皮穿刺右股動(dòng)脈或橈動(dòng)脈置入血管鞘,將冠狀動(dòng)脈導(dǎo)管分別置入左、右冠狀動(dòng)脈,選擇不同投照角度,檢查冠狀動(dòng)脈病情情況,每個(gè)體位對(duì)比劑注射量為4~6 mL。
冠脈CTA“雙低”成像檢查:采用第三代192層雙源CT掃描機(jī)(Somatom Force,Simens Healthineers,F(xiàn)orchheim,Germany)檢查,對(duì)比劑選用碘美普爾注射液(400 mg/mL)?;颊呷⊙雠P位,連接電極后對(duì)患者進(jìn)行呼吸和屏氣訓(xùn)練,常規(guī)舌下噴0.5 mg硝酸甘油氣霧劑,患者深吸氣未屏氣狀態(tài)下掃描,掃描范圍為氣管分叉下1 cm至心底位置。掃描參數(shù)設(shè)置:管電壓70~120 kV,根據(jù)患者定位窗由機(jī)器自動(dòng)調(diào)節(jié);管電流根據(jù)噪聲指數(shù)自動(dòng)調(diào)節(jié);旋轉(zhuǎn)時(shí)間:0.25 s/r;螺距:0.2;準(zhǔn)直器寬度:128×0.652 mm;重建矩陣:512×512;重建厚度:0.6 mm;重建間隔:0.3 mm。采用高壓注射器以3.5~5.5 mL/s速度經(jīng)肘靜脈注射對(duì)比劑39~60 mL,隨后以同速度注射生理鹽水40 mL,運(yùn)用對(duì)比劑追蹤觸發(fā)技術(shù)選擇主動(dòng)脈根部降主動(dòng)脈管腔層面為感興趣區(qū)(ROI),ROI的CT值>100 HU時(shí),延遲5 s自動(dòng)觸發(fā)掃描,采用心電門(mén)控技術(shù),選擇最佳舒張期重建圖像。
圖像處理:掃描完成后將原始圖像傳送值工作站,由資深心內(nèi)科醫(yī)生對(duì)前降支、右冠脈、回旋支冠脈三大主支進(jìn)行評(píng)估。參照美國(guó)心臟血管CT學(xué)會(huì)評(píng)估冠狀動(dòng)脈狹窄、狹窄程度、斑塊情況等,其中<25%為無(wú)狹窄或輕微狹窄;25%~50%為輕度狹窄;51%~70%為中度狹窄;71%~100%為重度狹窄或閉塞[6]。AI輔助診斷均在醫(yī)生診斷之后,采用AI輔助診斷軟件(數(shù)坤)一鍵生成冠脈CTA“雙低”成像結(jié)果和分析報(bào)告。
1.3 觀察指標(biāo)及評(píng)價(jià)標(biāo)準(zhǔn)
(1)診斷時(shí)間和圖像質(zhì)量評(píng)估比較:記錄醫(yī)生診斷和AI輔助診斷處理和分析圖像所需時(shí)間,統(tǒng)計(jì)醫(yī)生診斷和AI輔助診斷對(duì)圖像質(zhì)量評(píng)價(jià)結(jié)果,其中醫(yī)生診斷時(shí)間包括整個(gè)冠脈后處理時(shí)間加報(bào)告書(shū)寫(xiě)時(shí)間,AI處理時(shí)間為醫(yī)生點(diǎn)擊患者選項(xiàng)開(kāi)始至報(bào)告生成時(shí)間。圖像質(zhì)量評(píng)價(jià):圖像無(wú)偽影,血管連續(xù)完整,邊緣光滑、清晰為優(yōu);圖像有少量偽影,血管邊緣欠光滑為良好;圖像偽影較多,可勉強(qiáng)診斷為差。(2)診斷效能:以CAG檢查結(jié)果為金標(biāo)準(zhǔn),分析醫(yī)生診斷和AI輔助診斷對(duì)冠狀動(dòng)脈狹窄的診斷效能,其中檢出動(dòng)脈狹窄為陽(yáng)性,未檢出動(dòng)脈狹窄為陰性。(3)冠狀動(dòng)脈狹窄程度診斷:以醫(yī)生診斷結(jié)果為標(biāo)準(zhǔn),采用Kappa檢驗(yàn)分析AI輔助冠脈CTA“雙低”成像對(duì)冠狀動(dòng)脈狹窄程度診斷的一致性。(4)冠狀動(dòng)脈斑塊診斷:以醫(yī)生診斷結(jié)果為標(biāo)準(zhǔn),采用Kappa檢驗(yàn)分析AI輔助冠脈CTA“雙低”成像對(duì)冠狀動(dòng)脈斑塊診斷的一致性。
1.4 統(tǒng)計(jì)學(xué)處理
采用SPSS 22.0軟件分析數(shù)據(jù)。計(jì)數(shù)資料以率(%)表示,采用字2檢驗(yàn),計(jì)量資料以(x±s)表示,行t檢驗(yàn)。以醫(yī)生診斷為標(biāo)準(zhǔn),采用Kappa檢驗(yàn)分析AI輔助冠脈CTA“雙低”成像對(duì)冠狀動(dòng)脈狹窄程度、冠狀動(dòng)脈斑塊診斷的一致性,Kappa值≥0.75為一致性較好,0.4 2 結(jié)果 2.1 醫(yī)生組和AI組診斷時(shí)間及圖像質(zhì)量比較 AI組診斷時(shí)間顯著短于醫(yī)生組,且AI組診斷評(píng)估圖像質(zhì)量評(píng)優(yōu)率高于醫(yī)生組(P<0.05),見(jiàn)表1。 2.2 醫(yī)生組和AI組診斷冠狀動(dòng)脈狹窄的檢出情況 以CAG檢查結(jié)果為金標(biāo)準(zhǔn),醫(yī)生組對(duì)冠狀動(dòng)脈狹窄總體檢出的準(zhǔn)確度、敏感度、特異度、陽(yáng)性預(yù)測(cè)值、陰性預(yù)測(cè)值分別為93.97%、95.33%、77.78%、98.08%、58.33%;AI組對(duì)冠狀動(dòng)脈狹窄總體檢出的準(zhǔn)確度、敏感度、特異度、陽(yáng)性預(yù)測(cè)值、陰性預(yù)測(cè)值分別為93.10%、94.39%、77.78%、98.06%、53.85%,差異均無(wú)統(tǒng)計(jì)學(xué)意義(P>0.05)。見(jiàn)表2。 2.3 醫(yī)生組和AI組診斷冠狀動(dòng)脈狹窄程度的一致性 以醫(yī)生組結(jié)果為標(biāo)準(zhǔn),Kappa分析結(jié)果顯示,AI組對(duì)冠狀動(dòng)脈狹窄程度診斷的Kappa值為0.922,敏感度為91.46%,特異度為94.12%,見(jiàn)表3。 2.4 醫(yī)生組和AI組診斷對(duì)冠狀動(dòng)脈斑塊檢出的一致性 以醫(yī)生組結(jié)果為標(biāo)準(zhǔn),Kappa分析結(jié)果顯示,AI組對(duì)冠狀動(dòng)脈斑塊檢出的Kappa值為0.954,敏感度為95.37%,特異度為87.50%,見(jiàn)表4。 3 討論 冠狀動(dòng)脈病變?cè)\斷主要采用影像學(xué)檢查,其檢查金標(biāo)準(zhǔn)為CAG,通過(guò)在患者的手腕或大腿動(dòng)脈插入導(dǎo)管,經(jīng)過(guò)血管引導(dǎo)器將導(dǎo)管導(dǎo)入冠狀動(dòng)脈,并注入造影劑,采用X線攝影觀察造影劑在冠狀動(dòng)脈中的分布情況,以此判斷冠狀動(dòng)脈是否病變、病變程度、范圍及位置,但CAG為有創(chuàng)檢查,會(huì)對(duì)機(jī)體產(chǎn)生損傷,且部分患者排斥該法[7-9]。冠脈CTA“雙低”成像利用冠脈CTA中的“低劑量”和“低碘質(zhì)量”成像技術(shù)評(píng)估冠狀動(dòng)脈病變,還能從多角度、全方位評(píng)估動(dòng)脈管壁斑塊情況,該技術(shù)在保證圖像質(zhì)量的同時(shí),最大限度地減少患者接受輻射劑量和碘對(duì)比劑的風(fēng)險(xiǎn)[10-11]。近年來(lái),隨著AI技術(shù)的發(fā)展,AI被廣泛應(yīng)用于醫(yī)學(xué)影像,可減少醫(yī)生簡(jiǎn)單重復(fù)性工作,提高醫(yī)生診斷效率,同時(shí)減少人為錯(cuò)誤,提高診斷準(zhǔn)確性[12-13]。本研究納入116例疑似冠狀動(dòng)脈病變患者,分析人工智能輔助冠脈CTA“雙低”成像診斷冠狀動(dòng)脈病變的可行性。 本研究發(fā)現(xiàn),AI組診斷時(shí)間顯著短于醫(yī)生組,圖像質(zhì)量評(píng)優(yōu)率高于醫(yī)生組,且兩組對(duì)冠狀動(dòng)脈狹窄總體檢準(zhǔn)確度無(wú)統(tǒng)計(jì)學(xué)差異,提示AI診斷可有效縮短冠狀動(dòng)脈狹窄診斷和報(bào)告處理時(shí)間,且兩種診斷方式均具有較高的準(zhǔn)確度。究其原因可能是醫(yī)生診斷過(guò)程中為獲得較高質(zhì)量的圖像,需對(duì)原始數(shù)據(jù)進(jìn)行判斷,選擇最佳心動(dòng)周期圖像,并對(duì)圖像進(jìn)行后處理,該過(guò)程醫(yī)生需對(duì)多支冠狀動(dòng)脈進(jìn)行判讀,過(guò)程較為煩瑣,需要大量重復(fù)勞動(dòng)力,且受醫(yī)生熟練程度、個(gè)人經(jīng)驗(yàn)等影響,因此診斷所需時(shí)間較長(zhǎng)[14-15]。AI已具有圖像自動(dòng)后處理、診斷等多種深度學(xué)習(xí)的功能,對(duì)不同起點(diǎn)、路徑和終點(diǎn)的多支橋血管具有良好的識(shí)別和評(píng)估能力,通過(guò)計(jì)算機(jī)算法分析和解讀影像圖像來(lái)輔助診斷,且AI可以對(duì)大量的影像數(shù)據(jù)進(jìn)行自動(dòng)處理和分析,快速處理影像圖片及診斷報(bào)告,縮短診斷時(shí)間,且診斷結(jié)果與醫(yī)生診斷一致性較高。此外,AI主要通過(guò)像素面積識(shí)別進(jìn)行分級(jí),避免醫(yī)生診斷中人眼對(duì)中間質(zhì)量層級(jí)的感知誤差,提高閱片的準(zhǔn)確性,因此AI輔助診斷對(duì)影片的評(píng)優(yōu)率較高[16-18]。 本文顯示,以醫(yī)生診斷為標(biāo)準(zhǔn),AI組對(duì)冠狀動(dòng)脈狹窄程度診斷、斑塊檢出的Kappa值分別為0.922、0.954,說(shuō)明AI組對(duì)冠狀動(dòng)脈狹窄程度診斷、斑塊檢出與醫(yī)生診斷具有極高一致性。這可能是因?yàn)槿搜蹖?duì)冠狀動(dòng)脈狹窄程度及斑塊判斷主要依靠臨床經(jīng)驗(yàn)和醫(yī)學(xué)知識(shí),通過(guò)升主動(dòng)脈壁上的凸起盲端等,結(jié)合患者的癥狀、體征及相關(guān)的實(shí)驗(yàn)室檢查等方面綜合判斷,準(zhǔn)確率較高,但醫(yī)生對(duì)冠狀動(dòng)脈狹窄程度的判斷具有主觀性。AI軟件在學(xué)習(xí)數(shù)據(jù)時(shí)模型沒(méi)有代表性,會(huì)出現(xiàn)血管分割遺漏,出現(xiàn)誤診風(fēng)險(xiǎn),且偽影也會(huì)增加AI判斷誤差[19-20]。因此,AI輔助診斷在冠狀動(dòng)脈病變的診斷中可以起到輔助作用,但醫(yī)生的經(jīng)驗(yàn)和臨床判斷仍然是不可替代的,只有醫(yī)生和AI技術(shù)相結(jié)合,才能提高冠狀動(dòng)脈病變的診斷準(zhǔn)確性。 綜上所述,AI輔助冠脈CTA“雙低”成像對(duì)冠狀動(dòng)脈病變有較高的診斷價(jià)值,能顯著縮短冠狀動(dòng)脈病變的診斷時(shí)間,可作為醫(yī)生診斷冠狀動(dòng)脈病變的有效輔助工具。 參考文獻(xiàn) [1] MEDINA-LEYTE D J,ZEPEDA-GARC?A O,DOM?NGUEZ-P?REZ M,et al.Endothelial dysfunction, inflammation and coronary artery disease: potential biomarkers and promising therapeutical approaches[J].Int J Mol Sci,2021,22(8):3850. [2]劉震,李慧珠,高明康,等.血管內(nèi)光學(xué)相干斷層成像在頸動(dòng)脈粥樣硬化病變中的研究進(jìn)展[J].中國(guó)臨床實(shí)用醫(yī)學(xué),2022,13(4):70-73. [3] SUCATO V,COPPOLA G,MANNO G,et al.Coronary artery disease in south asian patients:cardiovascular risk factors,pathogenesis and treatments[J].Curr Probl Cardiol,2023,48(8):101228. [4] CURY R C,LEIPSIC J,ABBARA S,et al.CAD-RADS? 2.0-2022 coronary artery disease-reporting and data system: an expert consensus document of the society of cardiovascular computed tomography (SCCT), the American College of Cardiology (ACC), the American College of Radiology (ACR), and the North America Society of Cardiovascular Imaging (NASCI)[J].J Cardiovasc Comput Tomogr,2022,16(6):536-557. [5] CADEMARTIRI F,CASOLO G,CLEMENTE A,et al.Coronary CT angiography:a guide to examination, interpretation, and clinical indications[J].Expert Rev Cardiovasc Ther,2021,19(5):413-425. [6] CURY R C,ABBARA S,ACHENBACH S,et al.Coronary artery disease-reporting and data system (CAD-RADS): an expert consensus document of SCCT,ACR and NASCI:endorsed by the ACC[J].JACC Cardiovasc Imaging,2016,9(9):1099-1113. [7] FISCHER A M,EID M,DE CECCO C N,et al.Accuracy of an artificial intelligence deep learning algorithm implementing a recurrent neural network with long short-term memory for the automated detection of calcified plaques from coronary computed tomography angiography[J].J Thorac Imaging,2020,1(1):S49-S57. [8]馬方偉,張蕾,陳首名,等.64層CCTA對(duì)冠心病患者冠狀動(dòng)脈狹窄的診斷價(jià)值[J].西部醫(yī)學(xué),2023,35(7):1079-1082,1088. [9]朱軼,陳旭鋒,李偉,等.早期全身CT聯(lián)合冠狀動(dòng)脈造影術(shù)在體外心肺復(fù)蘇中的應(yīng)用[J].中華急診醫(yī)學(xué)雜志,2022,31(12):1603-1607. [10] QIAO H Y,TANG C X,SCHOEPF U J,et al.Impact of machine learning-based coronary computed tomography angiography fractional flow reserve on treatment decisions and clinical outcomes in patients with suspected coronary artery disease[J].Eur Radiol,2020,30(11):5841-5851. [11]文雨婷,李萬(wàn)江,李真林,等.超低管電壓聯(lián)合DLIR-H算法在冠脈CT血管成像“雙低”掃描中的應(yīng)用價(jià)值[J].中國(guó)醫(yī)療設(shè)備,2022,37(2):78-81,88. [12]左晨,劉暢,付麗媛.人工智能冠狀動(dòng)脈CT血管造影檢測(cè)冠狀動(dòng)脈狹窄的應(yīng)用價(jià)值研究[J].中國(guó)醫(yī)學(xué)裝備,2022,19(12):11-14. [13]宋丹丹,齊星亮,劉佳林,等.人工智能技術(shù)在胸部CT惡性及非鈣化結(jié)節(jié)檢出中的價(jià)值[J].醫(yī)學(xué)影像學(xué)雜志,2023,33(1):154-157. [14] LUO N,HE Y,F(xiàn)AN J,et al.Performance evaluation of deep learning-based post-processing and diagnostic reporting system for coronary CT angiography: a clinical comparative study[J].Chin Med J (Engl),2022,135(19):2366-2368. [15] PAUL J F,ROHNEAN A,GIROUSSENS H,et al.Evaluation of a deep learning model on coronary CT angiography for automatic stenosis detection[J].Diagn Interv Imaging,2022,103(6):316-323. [16] LIN A,MANRAL N,MCELHINNEY P,et al.Deep learning-enabled coronary CT angiography for plaque and stenosis quantification and cardiac risk prediction: an international multicentre study[J/OL].Lancet Digit Health,2022,4(4):e256-e265.https://pubmed.ncbi.nlm.nih.gov/35337643/. [17] LIU C Y,TANG C X,ZHANG X L,et al.Deep learning powered coronary CT angiography for detecting obstructive coronary artery disease:the effect of reader experience,calcification and image quality[J].Eur J Radiol,2021,142(1):109835. [18]張曉浩,劉軍波,范麗娟.人工智能技術(shù)應(yīng)用于冠狀動(dòng)脈CTA圖像后處理的可行性[J].放射學(xué)實(shí)踐,2021,36(8):994-999. [19]羅南,張婷婷,原媛,等.冠狀動(dòng)脈CT血管成像人工智能系統(tǒng)對(duì)冠狀動(dòng)脈旁路移植術(shù)后橋血管的重建及評(píng)估能力分析[J].實(shí)用放射學(xué)雜志,2023,39(8):1244-1248. [20]王憲凱,賈學(xué)燕,程祥科,等.人工智能在冠狀動(dòng)脈CTA圖像后處理及狹窄評(píng)估中的應(yīng)用價(jià)值[J].醫(yī)學(xué)影像學(xué)雜志,2022,32(4):588-590,720. (收稿日期:2024-03-14) (本文編輯:何玉勤)