国产日韩欧美一区二区三区三州_亚洲少妇熟女av_久久久久亚洲av国产精品_波多野结衣网站一区二区_亚洲欧美色片在线91_国产亚洲精品精品国产优播av_日本一区二区三区波多野结衣 _久久国产av不卡

?

卷積神經(jīng)網(wǎng)絡(luò)在T3/4期胃癌影像學(xué)診斷中應(yīng)用

2021-11-17 12:10張訓(xùn)營(yíng)張凱明張超馬金龍盧云王東升
關(guān)鍵詞:卷積神經(jīng)網(wǎng)絡(luò)診斷人工智能

張訓(xùn)營(yíng) 張凱明 張超 馬金龍 盧云 王東升

[摘要] 目的 利用卷積神經(jīng)網(wǎng)絡(luò)構(gòu)建T3/4期胃癌自動(dòng)識(shí)別平臺(tái)以達(dá)到輔助臨床診療目的。

方法 回顧性收集208例胃癌病人的增強(qiáng)CT圖像,并按照7∶1比例隨機(jī)分入訓(xùn)練集(182例)和驗(yàn)證集(26例),利用labelImg軟件標(biāo)識(shí)病變區(qū)域,用訓(xùn)練集對(duì)平臺(tái)進(jìn)行訓(xùn)練,用驗(yàn)證集進(jìn)行驗(yàn)證。通過(guò)對(duì)比平臺(tái)和影像學(xué)專家標(biāo)識(shí)圖像信息,采用受試者工作特征(ROC)曲線,對(duì)平臺(tái)性能進(jìn)行評(píng)估,評(píng)價(jià)指標(biāo)包括ROC曲線下面積(AUC)、準(zhǔn)確度、靈敏度、特異度、陽(yáng)性預(yù)測(cè)值及陰性預(yù)測(cè)值等。

結(jié)果 平臺(tái)的AUC為0.924,對(duì)T3/4期胃癌識(shí)別的準(zhǔn)確度、靈敏度、特異度分別為0.927、0.924、0.930,陽(yáng)性預(yù)測(cè)值為0.933,陰性預(yù)測(cè)值為0.921。

結(jié)論 平臺(tái)基于增強(qiáng)CT對(duì)T3/4期胃癌的識(shí)別準(zhǔn)確性與高年資影像學(xué)專家相當(dāng),并可以準(zhǔn)確識(shí)別出T3/4期胃癌病變區(qū)域,極大提高了胃癌術(shù)前診斷效率。

[關(guān)鍵詞] 卷積神經(jīng)網(wǎng)絡(luò);胃腫瘤;腫瘤分期;人工智能;診斷

[中圖分類號(hào)] R735.2;R445.3

[文獻(xiàn)標(biāo)志碼] A

[文章編號(hào)] 2096-5532(2021)05-0731-05

doi:10.11712/jms.2096-5532.2021.57.144

[開放科學(xué)(資源服務(wù))標(biāo)識(shí)碼(OSID)]

[網(wǎng)絡(luò)出版] https://kns.cnki.net/kcms/detail/37.1517.R.20210706.1547.010.html;2021-07-07 09:08:46

APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN RADIOLOGICAL DIAGNOSIS OF T3/4 GASTRIC CANCER

ZHANG Xunying, ZHANG Kaiming, ZHANG Chao, MA Jinlong, LU Yun, WANG Dongsheng

(Department of Gastrointestinal Surgery, The Affiliated Hospital of Qingdao University, Qingdao 266003, China)

[ABSTRACT] Objective To establish an automatic recognition platform for T3/4 gastric cancer based on convolutional neural network, and to achieve the purpose of assisting clinical diagnosis and treatment.

Methods A retrospective analysis was performed for the contrast-enhanced CT images of 208 patients with gastric cancer, and the patients were randomly divided into trai-

ning set with 182 patients and validation set with 26 patients at a ratio of 7∶1. The labelImg software was used to identify the lesion area, and the platform was trained by the training set and validated by the validation set. By comparing the image information identified by the platform and imaging experts, the receiver operating characteristic (ROC) curve was used to evaluate the perfor-

mance of this platform, and assessment indices included the area under the ROC curve (AUC), accuracy, sensitivity, specificity, positive predictive value, and negative predictive value.

Results The platform had an AUC of 0.924, with an accuracy of 0.927, a sensitivity of 0.924, and a specificity of 0.930 in identifying T3/4 gastric cancer, and the platform had a positive predictive value of 0.933 and a negative predictive value of 0.921.

Conclusion The platform based on contrast-enhanced CT has a comparable accuracy to senior imaging experts in identifying T3/4 gastric cancer and can accurately identify the area of the lesion of T3/4 gastric cancer, which greatly improves the efficiency of preoperative diagnosis of gastric cancer.

[KEY WORDS] convolutional neural network; stomach neoplasms; neoplasm staging; artificial intelligence; diagnosis

胃癌目前在全球癌癥發(fā)病率中位居第5,死亡率位居第3,嚴(yán)重威脅人類健康[1]。東亞是胃癌高發(fā)病及高死亡地區(qū),尤其是中國(guó)及日本[2]。中國(guó)胃癌病人占全球胃癌發(fā)病人數(shù)的42.6%,占相關(guān)死亡人數(shù)的45.0%[3-4]。盡管目前手術(shù)是治愈胃癌的唯一方法,但是新輔助化療的應(yīng)用可以明顯提高胃癌的治愈率及病人的生存率[5]。中國(guó)臨床腫瘤學(xué)會(huì)建議,新輔助化療適用于T3期及以上胃癌病人,T2期及更早的胃癌建議行手術(shù)治療[6]。因此,為了提高治愈率及降低新輔助化療的不利影響,需要對(duì)T3/4期胃癌病人進(jìn)行有效篩選[7]。CT由于具有非侵入性、實(shí)用性、便利性及穩(wěn)定性等優(yōu)點(diǎn),是術(shù)前評(píng)估胃癌分期的常規(guī)檢查方法[8]。但是,CT預(yù)測(cè)胃癌T分期的總體準(zhǔn)確率為43%~82%,容易對(duì)胃癌T分期產(chǎn)生誤判,造成不必要的姑息性手術(shù)及過(guò)度的放化療治療[9-12]。在這種情況下,需要一種替代技術(shù)對(duì)T3/4期胃癌病人進(jìn)行有效的篩選。

人工智能處理數(shù)據(jù)具有運(yùn)算速度快、精度高等優(yōu)點(diǎn)[13-14]。近年來(lái),在臨床實(shí)踐中卷積神經(jīng)網(wǎng)絡(luò)(CNN)越來(lái)越多地被用來(lái)識(shí)別和區(qū)分醫(yī)學(xué)圖像。該技術(shù)在影像圖片診斷中已經(jīng)顯示出具有較高的診斷性能,例如在檢測(cè)冠狀動(dòng)脈粥樣硬化、乳癌、轉(zhuǎn)移淋巴結(jié)[15-17]、皮膚病變的分類[18]及糖尿病視網(wǎng)膜病變篩查[19]等時(shí),在各種深度學(xué)習(xí)模型中CNN是最成熟的算法。本研究主要基于CNN在圖像處理及識(shí)別方面的強(qiáng)大能力,探索利用上腹部增強(qiáng)CT圖像建立CNN對(duì)T3/4期胃癌的自動(dòng)識(shí)別平臺(tái),并驗(yàn)證、評(píng)估其準(zhǔn)確性?,F(xiàn)將結(jié)果報(bào)告如下。

1 資料和方法

1.1 病人選擇

回顧性收集2018年6月—2019年12月在青島大學(xué)附屬醫(yī)院行根治性胃癌手術(shù)的564例病人的上腹部增強(qiáng)CT圖片。病人的納入標(biāo)準(zhǔn):術(shù)前行胃鏡檢查經(jīng)病理診斷為胃癌;術(shù)前于我院行上腹部增強(qiáng)CT檢查;于我院行根治性切除手術(shù),術(shù)后病理確診為T3/4期胃癌。排除標(biāo)準(zhǔn):腫瘤直徑較小無(wú)法勾畫感興趣區(qū)域(ROI);術(shù)前接受新輔助放化療;胃腔充盈狀態(tài)不理想或胃部蠕動(dòng)導(dǎo)致成像不理想病人;手術(shù)后復(fù)發(fā)的病人。最終共208例病人被納入研究,其中T3期病人90例,T4期病人118例。收集病人性別、年齡及腫瘤病理分期、部位等基本信息。本研究經(jīng)青島大學(xué)附屬醫(yī)院倫理委員會(huì)批準(zhǔn)。

1.2 病人分組及CT檢查方法

以腫瘤分期及腫瘤部位為分類標(biāo)準(zhǔn)將病人按7∶1比例隨機(jī)分入訓(xùn)練集(182例)和驗(yàn)證集(26例)。研究小組在訓(xùn)練集中共挑選出1 200張優(yōu)質(zhì)圖像確定為陽(yáng)性圖像,同理在測(cè)試集中挑選出210張陽(yáng)性圖像。本研究所有病人均采用飛利浦Brilliance iCT掃描儀行上腹部增強(qiáng)CT掃描,掃描層厚為1 mm,層間隔為1 mm,間距為0.985。檢查前所有病人均簽署碘對(duì)比劑知情同意書,禁食4~6 h,檢查前20 min給予病人500~1 000 mL飲用水。通過(guò)高壓注射器以3 mL/s的流量將90 mL非離子造影劑碘海醇注入前肘靜脈進(jìn)行增強(qiáng)掃描。在動(dòng)脈期延遲掃描33 s,在靜脈期延遲掃描65 s,在平衡期延遲掃描120 s。掃描范圍為橫膈膜到臍部平面。

1.3 圖像標(biāo)識(shí)及數(shù)據(jù)增強(qiáng)處理

利用labelImg軟件對(duì)圖像進(jìn)行標(biāo)識(shí),由兩名高年資放射科醫(yī)師分別獨(dú)立閱讀CT圖像并標(biāo)記腫瘤病變,標(biāo)識(shí)方法采用腫瘤分割方法。根據(jù)相關(guān)文獻(xiàn)的研究結(jié)果,與鄰近胃壁相比,局灶性胃壁增厚≥6 mm確定為異常增厚和癌變[20]。兩名影像科醫(yī)師結(jié)合病人胃鏡報(bào)告及術(shù)后最終病理結(jié)果,僅標(biāo)識(shí)影像圖像中腫瘤浸潤(rùn)胃壁最深的位置。根據(jù)術(shù)后病理結(jié)果,由第三位影像科醫(yī)師檢查上腹部增強(qiáng)CT圖像上腫瘤標(biāo)識(shí)部位,以保證增強(qiáng)CT圖像中病變部位的準(zhǔn)確性及一致性。

利用CNN提取上腹部增強(qiáng)CT圖像上不同大小的ROI,然后對(duì)1 200張陽(yáng)性圖片中的ROI使用裁剪、翻轉(zhuǎn)等數(shù)據(jù)增強(qiáng)方法進(jìn)行數(shù)據(jù)擴(kuò)增,最后篩選出2 500張陽(yáng)性圖像作為訓(xùn)練數(shù)據(jù)集,以增強(qiáng)研究數(shù)據(jù)集,同時(shí)減輕模型處理數(shù)據(jù)集時(shí)產(chǎn)生的過(guò)度擬合問(wèn)題[21]。

1.4 識(shí)別平臺(tái)構(gòu)建及驗(yàn)證

1.4.1 構(gòu)建識(shí)別平臺(tái) 構(gòu)建識(shí)別平臺(tái)前對(duì)圖像進(jìn)行預(yù)處理,包括采用圖像強(qiáng)度范圍歸一化和直方圖均衡化方法來(lái)處理圖像[22]。統(tǒng)一將訓(xùn)練圖像縮放為512×557像素大小,然后對(duì)識(shí)別平臺(tái)進(jìn)行訓(xùn)練,識(shí)別平臺(tái)在學(xué)習(xí)陽(yáng)性圖像同時(shí),將訓(xùn)練集中正常胃部解剖圖像默認(rèn)為陰性圖像一并學(xué)習(xí)。本研究采用的CNN是一個(gè)具有101層深度的CNN,可以對(duì)圖像特征進(jìn)行提取。每個(gè)層面的模型分別經(jīng)過(guò)800個(gè)epoch的訓(xùn)練。優(yōu)化器采用SGD優(yōu)化器,初始學(xué)習(xí)率為0.000 2。對(duì)CNN學(xué)習(xí)成果分析采用Python編程語(yǔ)言,提取Metric模塊里面的結(jié)果生成結(jié)論。

1.4.2 平臺(tái)驗(yàn)證 利用驗(yàn)證集中210張陽(yáng)性圖像和200張陰性圖像對(duì)該平臺(tái)識(shí)別性能進(jìn)行驗(yàn)證。對(duì)比影像科醫(yī)師對(duì)胃癌腫瘤區(qū)域標(biāo)注結(jié)果,判定平臺(tái)對(duì)驗(yàn)證集識(shí)別結(jié)果準(zhǔn)確性。通過(guò)繪制受試者工作特征(ROC)曲線,計(jì)算ROC曲線下面積(AUC),評(píng)估診斷平臺(tái)識(shí)別T3/4期胃癌圖像的準(zhǔn)確性,并統(tǒng)計(jì)識(shí)別的準(zhǔn)確度、靈敏度、特異度、陽(yáng)性預(yù)測(cè)值及陰性預(yù)測(cè)值等指標(biāo)。

1.5 統(tǒng)計(jì)學(xué)分析

應(yīng)用SPSS 20.0軟件對(duì)數(shù)據(jù)進(jìn)行統(tǒng)計(jì)學(xué)處理。統(tǒng)計(jì)所有結(jié)點(diǎn)處的真陽(yáng)性和假陽(yáng)性的數(shù)目,計(jì)算得到不同概率閾值下真陽(yáng)性率和假陽(yáng)性率,從而繪制出ROC曲線,通過(guò)計(jì)算AUC,得出平臺(tái)識(shí)別T3/4期胃癌的準(zhǔn)確率。

2 結(jié)? 果

2.1 平臺(tái)的學(xué)習(xí)效果

為評(píng)估平臺(tái)的學(xué)習(xí)效果,研究小組將驗(yàn)證集輸入經(jīng)過(guò)訓(xùn)練的識(shí)別平臺(tái)中進(jìn)行驗(yàn)證。由診斷平臺(tái)學(xué)習(xí)結(jié)果的損失函數(shù)(loss)學(xué)習(xí)曲線可知,診斷平臺(tái)在進(jìn)行800個(gè)epoch學(xué)習(xí)后達(dá)到最佳優(yōu)化參數(shù)。用于識(shí)別T3/4期胃癌診斷平臺(tái)的AUC為0.924,準(zhǔn)確度、靈敏度、特異度分別為0.927、0.924、0.930,陽(yáng)性預(yù)測(cè)值為0.933,陰性預(yù)測(cè)值為0.921。見(jiàn)圖1。

2.2 平臺(tái)驗(yàn)證

如圖2所示,A、C圖片是影像科醫(yī)師基于病理結(jié)果手動(dòng)標(biāo)識(shí)的腫瘤位置,B、D圖片是識(shí)別平臺(tái)對(duì)圖片中腫瘤的分割及識(shí)別??梢缘贸鼋Y(jié)論,基于上腹部增強(qiáng)CT圖像,該識(shí)別平臺(tái)識(shí)別T3/4期胃癌具有很高的準(zhǔn)確性。

3 討? 論

準(zhǔn)確的術(shù)前T分期對(duì)胃癌病人圍手術(shù)期選擇治療方案以及評(píng)估預(yù)后均至關(guān)重要[23]。有研究證實(shí),胃癌病理T3/4期是術(shù)后切緣陽(yáng)性的獨(dú)立危險(xiǎn)因素,術(shù)后切緣陽(yáng)性病人總體預(yù)后較差[24]。中國(guó)臨床腫瘤學(xué)會(huì)建議治療胃癌之前需要準(zhǔn)確地區(qū)分胃癌T分期以制定精準(zhǔn)的治療計(jì)劃。第8版美國(guó)國(guó)立綜合癌癥網(wǎng)絡(luò)指南(NCCN指南)提出上腹部增強(qiáng)CT是診斷胃癌T分期的主要影像學(xué)方法。影像科醫(yī)師主要通過(guò)術(shù)前腹部CT等影像資料評(píng)估胃癌病人T分期,繼而指導(dǎo)臨床醫(yī)師選擇治療方案。然而,目前實(shí)際臨床工作中術(shù)前通過(guò)腹部CT判斷T分期還存在一些問(wèn)題:①不同影像科醫(yī)師通過(guò)上腹部增強(qiáng)CT評(píng)估胃癌病人T分期存在主觀差異;②在目前國(guó)內(nèi)各三級(jí)甲等醫(yī)院病人數(shù)量龐大的背景下,面對(duì)通過(guò)CT等影像資料進(jìn)行胃癌臨床分期的復(fù)雜性,影像科醫(yī)生承受著巨大的工作量。因此,迫切需要一種新的方法來(lái)提高臨床診斷效率。而深度學(xué)習(xí)網(wǎng)絡(luò)技術(shù)的發(fā)展,為解決這一問(wèn)題創(chuàng)造了可能。有研究報(bào)道,CNN-CAD系統(tǒng)已經(jīng)應(yīng)用于乳房組織病理學(xué)圖像的分類檢測(cè)[25]和結(jié)腸直腸癌的檢測(cè)[26]。本研究基于CNN建立了識(shí)別T3/4期胃癌的診斷平臺(tái),實(shí)現(xiàn)了術(shù)前利用上腹部增強(qiáng)CT對(duì)胃癌進(jìn)行快速精確篩選。

本研究小組在前期已經(jīng)開發(fā)出基于深度學(xué)習(xí)網(wǎng)絡(luò)的直腸癌轉(zhuǎn)移淋巴結(jié)的MRI圖像自動(dòng)識(shí)別系統(tǒng)[27]。在前期經(jīng)驗(yàn)的基礎(chǔ)上,本研究建立了基于CNN的T3/4期胃癌自動(dòng)識(shí)別平臺(tái),并評(píng)估了其臨床價(jià)值。上腹部增強(qiáng)CT為胃癌病人的常規(guī)輔助檢查手段,有研究證實(shí),上腹部增強(qiáng)CT靜脈期圖像對(duì)胃癌腫瘤浸潤(rùn)的診斷性能優(yōu)于動(dòng)脈期圖像,所以本研究選用上腹部增強(qiáng)CT靜脈期圖像[28]。最新版胃癌NCCN指南指出,上腹部CT對(duì)胃癌T分期的識(shí)別準(zhǔn)確率為43%~82%[23]。蘭州大學(xué)第二醫(yī)院的一項(xiàng)回顧性研究將胃癌術(shù)后病理結(jié)果與影像科高年資醫(yī)師讀片報(bào)告對(duì)比,結(jié)果顯示,增強(qiáng)CT評(píng)估T3、T4期胃癌的準(zhǔn)確度分別為76.7%和92.7%[29]。本研究中通過(guò)放射科高年資醫(yī)生結(jié)合病理結(jié)果對(duì)增強(qiáng)CT圖片的標(biāo)識(shí),對(duì)T3/4期胃癌識(shí)別平臺(tái)進(jìn)行深度訓(xùn)練,經(jīng)驗(yàn)證識(shí)別平臺(tái)對(duì)于T3/4期胃癌病人增強(qiáng)CT靜脈期圖像具有較高的識(shí)別準(zhǔn)確度,其AUC為0.924。表明識(shí)別平臺(tái)的準(zhǔn)確性接近于影像科高年資醫(yī)師的診斷水平??紤]其原因可能為:①T4期胃癌腫瘤較大、浸透漿膜層,在CT圖片中易于辨認(rèn);②T3期胃癌侵犯至胃壁的固有肌層,而固有肌層在增強(qiáng)CT中構(gòu)成了低密度條紋層的外層[30],在病理切片中占據(jù)了胃壁的大部分,這降低了識(shí)別T分期的難度;③T3/4期胃癌圖片較多,診斷平臺(tái)參數(shù)優(yōu)化較完善?;谠\斷平臺(tái)對(duì)病人全靜脈期連續(xù)圖像的T分期最終判定結(jié)果與術(shù)后病理T分期診斷結(jié)果完全相符。上述結(jié)果表明,該診斷平臺(tái)具有較高的可行性、準(zhǔn)確性、客觀性和高效性,可以輔助放射科醫(yī)生完成胃癌的篩選工作,減少放射科醫(yī)生的工作量;可以輔助臨床醫(yī)生制定診療方案,從而有利于胃癌病人接受更加精準(zhǔn)和高效的治療。

本研究的局限性:①本研究為單中心試驗(yàn)研究,數(shù)據(jù)量有限;②本研究是基于CNN的監(jiān)督學(xué)習(xí),平臺(tái)的訓(xùn)練準(zhǔn)確性依賴于放射科醫(yī)師對(duì)增強(qiáng)CT圖片腫瘤區(qū)域的精確標(biāo)識(shí)。因此,為了進(jìn)一步提升人工智能輔助平臺(tái)的可靠性,今后研究將聯(lián)合多中心增加數(shù)據(jù)量,并優(yōu)化算法和提升標(biāo)識(shí)效率,最終實(shí)現(xiàn)臨床驗(yàn)證,以達(dá)到輔助醫(yī)師診斷與治療的目的。

綜上所述,本研究建立的T3/4期胃癌自動(dòng)識(shí)別平臺(tái)能夠利用上腹部增強(qiáng)CT圖像自動(dòng)分割識(shí)別T3/4期胃癌,顯示出了與經(jīng)驗(yàn)豐富的影像科醫(yī)師相當(dāng)?shù)臏?zhǔn)確性,有望協(xié)助影像科醫(yī)師做出更為精準(zhǔn)、直觀及高效的診斷,極大減輕影像科醫(yī)師的工作負(fù)擔(dān);有望指導(dǎo)臨床醫(yī)師制定診療方案,利于病人接受更加精準(zhǔn)及個(gè)性化的治療。

[參考文獻(xiàn)]

[1]Global Burden Of Disease Cancer Collaboration, FITZMAURICE C, ALLEN C, et al. Global, regional, and national cancer incidence, mortality, years of life lost, years lived with disability, and disability-adjusted life-years for 32 cancer groups, 1990 to 2015: a systematic analysis for the global burden of disease study[J].? JAMA Oncology, 2017,3(4):524-548.

[2]RAHMAN R, ASOMBANG A W, IBDAH J A. Characteristics of gastric cancer in Asia[J].? World Journal of Gastroente-

rology, 2014,20(16):4483-4490.

[3]FERLAY J, SOERJOMATARAM I, DIKSHIT R, et al. Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012[J].? Int J Cancer,? 2015,136(5):E359-E386.

[4]WANG Z, XING J, CAI J, et al. Short-term surgical outcomes of laparoscopy-assisted versus open D2 distal gastrectomy for locally advanced gastric cancer in north China: a multi- ?center randomized controlled trial[J].? Surg Endosc, 2019,33(1):33-45.

[5]BIONDI A, LIROSI M C, DUGO D, et al. Neo-adjuvant chemo (radio) therapy in gastric cancer: Current status and future perspectives[J].? World J Gastrointest Oncol, 2015,7(12):389-400.

[6]WANG F H, SHEN L, LI J, et al. The Chinese Society of Clinical Oncology (CSCO): clinical guidelines for the diagnosis and treatment of gastric cancer[J].? Cancer Communications (London, England), 2019,39(1):10.

[7]FUKAGAWA T, KATAI H, MIZUSAWA J, et al. A prospective multi-institutional validity study to evaluate the accuracy of clinical diagnosis of pathological stage Ⅲ gastric cancer (JCOG1302A)[J].? Gastric Cancer, 2018,21(1):68-73.

[8]LIU S, SHI H, JI C, et al. Preoperative CT texture analysis of gastric cancer: correlations with postoperative TNM staging[J].? Clinical Radiology, 2018,73(8):756.e1-756.e9.

[9]TAN C Q, PENG L B, ZENG X H, et al. Cost-utility analysis of the newly recommended adjuvant chemotherapy for resec-

table gastric cancer patients in the 2011 Chinese National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines in Oncology: Gastric Cancer[J].? PharmacoEconomics, 2014,32(3):235-243.

[10]YAN C, ZHU Z G, YAN M, et al. Value of multidetector-row computed tomography in the preoperative T and N staging of gastric carcinoma: a large-scale Chinese study[J].? Journal of Surgical Oncology, 2009,100(3):205-214.

[11]SAITO T, KUROKAWA Y, TAKIGUCHI S, et al. Accuracy of multidetector-row CT in diagnosing lymph node metastasis in patients with gastric cancer[J].? European Radiology, 2015,25(2):368-374.

[12]JOO I, LEE J M, KIM J H, et al. Prospective comparison of 3 T MRI with diffusion-weighted imaging and MDCT for the preoperative TNM staging of gastric cancer[J].? Journal of Magnetic Resonance Imaging: JMRI, 2015,41(3):814-821.

[13]KUMAR V, GU Y, BASU S, et al. Radiomics: the process and the challenges[J].? Magn Reson Imaging, 2012,30(9):1234-1248.

[14]LAMBIN P, RIOS-VELAZQUEZ E, LEIJENAAR R, et al. Radiomics: extracting more information from medical images using advanced feature analysis[J].? European Journal of Can-

cer, 2012,48(4):441-446.

[15]TATSUGAMI F, HIGAKI T, NAKAMURA Y, et al. Deep learning-based image restoration algorithm for coronary CT angiography[J].? Eur Radiol, 2019,29(10):5322-5329.

[16]GERAS K J, MANN R M, MOY L. Artificial intelligence for mammography and digital breast tomosynthesis: current concepts and future perspectives[J].? Radiology, 2019,293(2):246-259.

[17]EHTESHAMI B B, MITKO V, JOHANNES V D P, et al. Diagnostic assessment of deep learning algorithms for detection of lymph node metastases in women with breast cancer[J].? JAMA, 2017,318(22):2199-2210.

[18]ESTEVA A, KUPREL B, NOVOA R A, et al. Corrigendum: dermatologist-level classification of skin cancer with deep neural networks[J].? Nature, 2017,546(7660):686.

[19]VARUN G, LILY P, MARC C, et al. Development and validation of a deep learning algorithm for detection of diabetic re-

tinopathy in retinal fundus photographs[J].? JAMA, 2016,316(22):2402-2410.

[20]KIM H J, KIM A Y, OH S T, et al.

Gastric cancer staging at multi-detector row CT gastrography: comparison of transverse and volumetric CT scanning[J].? Radiology, 2005,236(3):879-885.

[21]ZHONG Z, ZHENG L, KANG G L, et al. Random erasing data augmentation[J].? Proceedings of the AAAI Conference on Artificial Intelligence, 2020,34(7):13001-13008.

[22]DROZDZAL M, CHARTRAND G, VORONTSOV E, et al. Learning normalized inputs for iterative estimation in medical image segmentation[J].? Comput Vis Pattern Recognit, 2017,44:1-10.

[23]AJANI J A, D′AMICO T A, ALMHANNA K, et al. Gastric cancer, version 3.2016, NCCN clinical practice guidelines in oncology[J].? Journal of the National Comprehensive Cancer Network: JNCCN, 2016,14(10):1286-1312.

[24]SCHUHMACHER C, GRETSCHEL S, LORDICK F, et al. Neoadjuvant chemotherapy compared with surgery alone for locally advanced cancer of the stomach and cardia: European Organisation for Research and Treatment of Cancer rando-

mized trial 40954[J].? J Clin Oncol, 2010,28(35):5210-5218.

[25]GANDOMKAR Z, BRENNAN P C, MELLO-THOMS C. MuDeRN: multi-category classification of breast histopathological image using deep residual networks[J].? Artificial Intelligence in Medicine, 2018,88:14-24.

[26]HORNBROOK M C, GOSHEN R, CHOMAN E, et al. Early colorectal cancer detected by machine learning model using gender, age, and complete blood count data[J].? Dig Dis Sci, 2017,62(10):2719-2727.

[27]LU Y, YU Q, GAO Y, et al. Identification of metastatic lymph nodes in MR imaging with faster region-based convolutional neural networks[J]. Cancer Res, 2018,78(17):5135-5143.

[28]CHEN X H, REN K, LIANG P, et al. Spectral computed tomography in advanced gastric cancer: can iodine concentration non-invasively assess angiogenesis[J]? World Journal of Gastroenterology, 2017,23(9):1666-1675.

[29]武贊凱,王宇,張耕源,等. 增強(qiáng)CT對(duì)進(jìn)展期胃癌術(shù)前的評(píng)估價(jià)值[J].? 中華普通外科學(xué)文獻(xiàn)(電子版), 2016,10(2):133-137.

[30]KIM J W, SHIN S S, HEO S H, et al. Diagnostic perfor-

mance of 64-section CT using CT gastrography in preoperative T staging of gastric cancer according to 7th edition of AJCC cancer staging manual[J].? Eur Radiol, 2012,22(3):654-662.

(本文編輯 馬偉平)

猜你喜歡
卷積神經(jīng)網(wǎng)絡(luò)診斷人工智能
2019:人工智能
人工智能與就業(yè)
數(shù)讀人工智能
基于深度卷積神經(jīng)網(wǎng)絡(luò)的物體識(shí)別算法
深度學(xué)習(xí)算法應(yīng)用于巖石圖像處理的可行性研究
基于深度卷積網(wǎng)絡(luò)的人臉年齡分析算法與實(shí)現(xiàn)
紅外線測(cè)溫儀在汽車診斷中的應(yīng)用
窄帶成像聯(lián)合放大內(nèi)鏡在胃黏膜早期病變?cè)\斷中的應(yīng)用
下一幕,人工智能!
基于卷積神經(jīng)網(wǎng)絡(luò)的樹葉識(shí)別的算法的研究