孟誠(chéng),鄔芝雅,馮繼鋒
肺結(jié)節(jié)風(fēng)險(xiǎn)分層評(píng)估的研究進(jìn)展
孟誠(chéng)1,2,鄔芝雅1,2,馮繼鋒1,2
1.南京醫(yī)科大學(xué)附屬腫瘤醫(yī)院(江蘇省腫瘤醫(yī)院)腫瘤內(nèi)科,江蘇南京 210000;2.南京醫(yī)科大學(xué)第四臨床醫(yī)學(xué)院,江蘇南京 211166
近年來(lái),隨著肺癌早期篩查的逐步普及和低劑量螺旋CT的深入應(yīng)用,肺結(jié)節(jié)的檢出率越來(lái)越高,這會(huì)導(dǎo)致患者產(chǎn)生焦慮情緒。盡早對(duì)肺結(jié)節(jié)進(jìn)行評(píng)估并判斷其良惡性是緩解患者焦慮情緒、合理處置肺結(jié)節(jié)的關(guān)鍵所在?;颊邆€(gè)體臨床信息、影像學(xué)特征、生物標(biāo)志物和預(yù)測(cè)模型已成為肺結(jié)節(jié)風(fēng)險(xiǎn)分層評(píng)估的手段,本文對(duì)肺結(jié)節(jié)的臨床分類和風(fēng)險(xiǎn)分層評(píng)估的最新研究進(jìn)展進(jìn)行綜述。
肺結(jié)節(jié);肺癌;風(fēng)險(xiǎn)分層
肺癌是我國(guó)最常見(jiàn)的惡性腫瘤之一,也是癌癥中導(dǎo)致患者病死的首要原因,肺癌患者約占癌癥總病死人數(shù)的27.22%[1]。研究表明,對(duì)肺癌高風(fēng)險(xiǎn)人群進(jìn)行低劑量螺旋CT(low-dose computed tomography,LDCT)檢查,可早期發(fā)現(xiàn)肺癌,改善患者預(yù)后,降低患者病死率[2-4]。隨著LDCT技術(shù)的不斷普及,肺結(jié)節(jié)檢出率顯著升高,其精確管理也成為臨床治療面臨的嚴(yán)峻挑戰(zhàn)之一。本文對(duì)肺結(jié)節(jié)的臨床分類和風(fēng)險(xiǎn)分層評(píng)估的最新研究進(jìn)展進(jìn)行綜述。
影像學(xué)表現(xiàn)為直徑≤3cm的局灶性、類圓形、密度增高的實(shí)性或亞實(shí)性肺部陰影即為肺結(jié)節(jié)。肺結(jié)節(jié)可為孤立性或多發(fā)性,不伴肺不張、肺門(mén)淋巴結(jié)腫大和胸腔積液。根據(jù)病灶的數(shù)目,肺結(jié)節(jié)可分為孤立性肺結(jié)節(jié)、多發(fā)性肺結(jié)節(jié)及彌漫性肺結(jié)節(jié)。英國(guó)胸科協(xié)會(huì)(British Thoracic Society,BTS)依據(jù)結(jié)節(jié)的密度,將肺結(jié)節(jié)分為實(shí)質(zhì)性結(jié)節(jié)和亞實(shí)質(zhì)性結(jié)節(jié),并將亞實(shí)質(zhì)性結(jié)節(jié)又分為部分實(shí)質(zhì)性結(jié)節(jié)和純磨玻璃結(jié)節(jié)[5]。中華醫(yī)學(xué)會(huì)呼吸病學(xué)分會(huì)肺癌學(xué)組等[6]將直徑<5mm的肺結(jié)節(jié)定義為微小結(jié)節(jié),將直徑為5~10mm的肺結(jié)節(jié)定義為小結(jié)節(jié)。此外,在肺結(jié)節(jié)直徑因素的基礎(chǔ)上,BTS建議將初始體積和體積倍增時(shí)間(volume doubling time,VDT)納入其分類標(biāo)準(zhǔn)中[5],F(xiàn)leischner學(xué)會(huì)則將肺結(jié)節(jié)體積納入其分類標(biāo)準(zhǔn)中[7]。
美國(guó)胸科醫(yī)師協(xié)會(huì)(American College of Chest Physicians,ACCP)、Fleischner學(xué)會(huì)、BTS及我國(guó)的指南共識(shí)在很大程度上依賴于結(jié)節(jié)大小(直徑或體積)和密度來(lái)提供臨床處理意見(jiàn)[5-8]。然而,這些風(fēng)險(xiǎn)分層方法在區(qū)分良性與惡性肺結(jié)節(jié)方面仍有局限性。在肺結(jié)節(jié)評(píng)估的研究中發(fā)現(xiàn),研究納入的標(biāo)準(zhǔn)和來(lái)源人群并不一致,其準(zhǔn)確性和臨床效用取決于來(lái)源人群的病例組合及該人群中惡性腫瘤的患病率[5]。Fleischner學(xué)會(huì)則僅從肺結(jié)節(jié)的影像學(xué)特征方面提供管理策略。小結(jié)節(jié)和亞實(shí)質(zhì)性結(jié)節(jié)缺乏相應(yīng)的預(yù)測(cè)模型[7],ACCP呼吁應(yīng)進(jìn)一步研究、開(kāi)發(fā)并驗(yàn)證相應(yīng)的風(fēng)險(xiǎn)預(yù)測(cè)模型和新型非侵入性生物標(biāo)志物,以促進(jìn)肺結(jié)節(jié)的診斷并判斷其預(yù)后[8]。大量研究關(guān)注并改進(jìn)原有的分層方式,探索新的分層方法。
在對(duì)肺結(jié)節(jié)惡性概率進(jìn)行評(píng)估前,掌握患者的臨床信息至關(guān)重要。ACCP基于風(fēng)險(xiǎn)預(yù)測(cè)模型,將實(shí)質(zhì)性肺結(jié)節(jié)分為低惡性概率(<5%)、中惡性概率(5%~65%)和高惡性概率(>65%)[8]。低惡性概率實(shí)質(zhì)性肺結(jié)節(jié)的特征包括年輕、吸煙較少、既往腫瘤史、小結(jié)節(jié)、邊緣規(guī)則和不位于上葉;高惡性概率實(shí)質(zhì)性肺結(jié)節(jié)具有相反的特征;中惡性概率實(shí)質(zhì)性肺結(jié)節(jié)則具有上述二者的混合特征。中華醫(yī)學(xué)會(huì)呼吸病學(xué)分會(huì)肺癌學(xué)組等[6]研究指出,“肺結(jié)節(jié)位于上葉惡性概率大”的說(shuō)法不適用于我國(guó),因?yàn)樵谖覈?guó)上葉尖后段是肺結(jié)核的好發(fā)部位。評(píng)估肺結(jié)節(jié)惡性概率需結(jié)合我國(guó)人群的肺癌流行病學(xué)研究結(jié)果。統(tǒng)計(jì)數(shù)據(jù)顯示,2016年我國(guó)新發(fā)肺癌病例82.8萬(wàn)例,其中男性和女性發(fā)病率分別為49.78/10萬(wàn)和23.70/10萬(wàn),男性發(fā)病率顯著高于女性[1]。從2015年的數(shù)據(jù)來(lái)看,東、中、西三大經(jīng)濟(jì)地區(qū),肺癌的發(fā)病率也存在著較大差異,其中東部地區(qū)的發(fā)病率最高,中部次之,西部最低。年齡方面,我國(guó)肺癌患者的發(fā)病率在44歲前處于較低水平,45歲之后快速上升,80~84歲達(dá)到高峰[9-10]。綜上,性別、年齡和患者所在地區(qū)等也是評(píng)估肺結(jié)節(jié)良惡性概率的考量因素。
目前,LDCT已成為肺結(jié)節(jié)評(píng)估和肺癌篩查的首要手段。研究顯示,與胸部X線片相比,對(duì)肺癌高危人群進(jìn)行篩查時(shí),LDCT可使肺癌患者的病死率下降20%[2];荷蘭–比利時(shí)隨機(jī)肺癌篩查試驗(yàn)結(jié)果也證實(shí)了上述結(jié)論[11]。肺結(jié)節(jié)的直徑和密度是影像學(xué)評(píng)估的主要指標(biāo)。肺結(jié)節(jié)的直徑或體積越大,其發(fā)生惡性腫瘤的可能性越高[12]。含有磨玻璃成分的部分實(shí)質(zhì)性結(jié)節(jié)的惡性概率較實(shí)質(zhì)性結(jié)節(jié)更高,且以肺腺癌較為多見(jiàn)[13]。但有研究表明,胸部CT顯示的直徑并不能準(zhǔn)確估算肺結(jié)節(jié)的實(shí)際體積[14]。因此,VDT作為結(jié)節(jié)生長(zhǎng)快慢的指標(biāo)已成為惡性腫瘤的最佳預(yù)測(cè)因子之一。BTS將VDT在400~600d的結(jié)節(jié)視為中等風(fēng)險(xiǎn)組,VDT<400d的結(jié)節(jié)為高危組[5]。Horeweg等[12]對(duì)荷蘭–比利時(shí)隨機(jī)肺癌篩查試驗(yàn)的數(shù)據(jù)進(jìn)行分析,結(jié)果表明基于VDT的評(píng)估方案的敏感度為90.9%,特異性為94.9%;因其具有的高特異性,與其他方案相比,較少的患者需要接受隨訪CT檢查及額外的診斷程序,可顯著減少假陽(yáng)性概率和隨訪率。研究指出,正電子發(fā)射計(jì)算機(jī)體層掃描術(shù)(positron emission tomography,PET)/CT是目前區(qū)分惡性和良性孤立性肺結(jié)節(jié)的最敏感及最具特異性的成像技術(shù)[11]。PET/CT可減少呼吸運(yùn)動(dòng)偽影的影響,還能用于肺癌手術(shù)前的預(yù)分期[6,15]。近年來(lái),基于成纖維激活蛋白(fibroblast activation protein,F(xiàn)AP)的PET/CT(68Ga-FAPI PET/CT)也已進(jìn)入臨床研究。研究顯示,在惡性肺磨玻璃結(jié)節(jié)中,68Ga-FAPI PET/CT的示蹤劑攝取量高于18F-FDG PET/CT[16]。然而,研究顯示某些良性疾病也可能出現(xiàn)FAPI陽(yáng)性,包括椎體骨折、結(jié)核病、胰腺炎等[17-19]。相信隨著影像學(xué)技術(shù)的不斷更新,肺結(jié)節(jié)良惡性的判斷也會(huì)有更多的手段和更豐富的標(biāo)準(zhǔn)。
臨床上用來(lái)評(píng)估肺結(jié)節(jié)的預(yù)測(cè)模型主要包括Mayo模型、VA模型、Brock模型等。上述模型主要基于西方人群,其在中國(guó)人群的診斷價(jià)值尚不清楚。Cui等[20]比較不同模型和放射科醫(yī)生的評(píng)估能力,指出可能需要優(yōu)化預(yù)測(cè)模型才能應(yīng)用于亞洲人群,其評(píng)估肺結(jié)節(jié)的風(fēng)險(xiǎn)能力低于放射科醫(yī)生。早期的預(yù)測(cè)模型并未考慮腫瘤生物標(biāo)志物,但隨著其應(yīng)用越來(lái)越廣泛,國(guó)內(nèi)研究也開(kāi)始關(guān)注生物標(biāo)志物。Dong等[21]招募孤立性肺結(jié)節(jié)患者以開(kāi)發(fā)基于國(guó)內(nèi)人群的診斷模型,通過(guò)找尋國(guó)內(nèi)人群的預(yù)測(cè)因子和生物標(biāo)志物所建立的模型精度略高于Mayo模型,為國(guó)內(nèi)孤立性肺結(jié)節(jié)的風(fēng)險(xiǎn)評(píng)估提供一定的依據(jù)。近年來(lái),越來(lái)越多的研究將生物標(biāo)志物作為預(yù)測(cè)因子,結(jié)合患者的臨床信息和影像學(xué)特征,其得到的模型可能更適用于國(guó)內(nèi)人群[22-25]。Silvestri等[26]將血漿蛋白與預(yù)測(cè)模型結(jié)合,發(fā)現(xiàn)其可準(zhǔn)確識(shí)別預(yù)測(cè)癌癥概率低于50%的良性肺結(jié)節(jié)。Ostrin等[27]研究證實(shí),生物標(biāo)志物可改善肺結(jié)節(jié)風(fēng)險(xiǎn)預(yù)測(cè)模型,提高其識(shí)別良惡性肺結(jié)節(jié)的特異性和敏感度。另外,大量研究嘗試單獨(dú)使用生物標(biāo)志物識(shí)別良惡性結(jié)節(jié)。在Marquette等[28]開(kāi)展的一項(xiàng)多中心隊(duì)列中,在肺癌篩查的背景下對(duì)循環(huán)腫瘤細(xì)胞進(jìn)行前瞻性驗(yàn)證,結(jié)果表明其并不適合肺癌篩查。與前瞻性臨床試驗(yàn)已建立的風(fēng)險(xiǎn)模型相比,生物標(biāo)志物仍需臨床效用的驗(yàn)證。
深度學(xué)習(xí)主要是利用放射組學(xué)的定量特征來(lái)訓(xùn)練卷積神經(jīng)網(wǎng)絡(luò)。目前有證據(jù)顯示,卷積神經(jīng)網(wǎng)絡(luò)的風(fēng)險(xiǎn)評(píng)估能力與放射科醫(yī)生相當(dāng),甚至超過(guò)放射科醫(yī)生,對(duì)肺結(jié)節(jié)的評(píng)估有巨大潛力[29-31]。除卷積神經(jīng)網(wǎng)絡(luò),也有其他科學(xué)家基于胸部CT圖像,利用不同的數(shù)據(jù)庫(kù)來(lái)訓(xùn)練并驗(yàn)證自己的深度學(xué)習(xí)算法。Ali等[32]利用LIDC/IDRI肺結(jié)節(jié)數(shù)據(jù)庫(kù)數(shù)據(jù),結(jié)合放射科醫(yī)生的注釋作為深度學(xué)習(xí)模型的訓(xùn)練集,該模型在訓(xùn)練時(shí)的總體準(zhǔn)確率為99.1%,實(shí)際測(cè)試的準(zhǔn)確率為64.4%。Venkadesh等[33]的深度學(xué)習(xí)算法則是基于2002—2004年美國(guó)國(guó)家肺部篩查試驗(yàn)中收集的肺結(jié)節(jié)數(shù)據(jù)訓(xùn)練,并利用2004—2010年來(lái)自丹麥肺癌篩查試驗(yàn)中收集的數(shù)據(jù)予以驗(yàn)證,最終與加拿大肺癌早期檢測(cè)模型及臨床醫(yī)生組進(jìn)行比較,結(jié)果表明深度學(xué)習(xí)模型在整個(gè)隊(duì)列中的表現(xiàn)明顯優(yōu)于加拿大肺癌早期檢測(cè)模型。深度學(xué)習(xí)模型在肺癌富集隊(duì)列中與臨床醫(yī)生組表現(xiàn)相當(dāng)。另外也有學(xué)者嘗試改善卷積神經(jīng)網(wǎng)絡(luò)。Masquelin等[34]以多級(jí)小波變換取代早期卷積層,從而達(dá)到改善圖像分類的目的。目前國(guó)內(nèi)也有回顧性研究,利用既往分類好的肺結(jié)節(jié)臨床數(shù)據(jù)來(lái)開(kāi)發(fā)算法,為優(yōu)化臨床工作流程和提高診斷準(zhǔn)確性帶來(lái)希望[35-37]。綜合國(guó)內(nèi)外報(bào)道,深度學(xué)習(xí)算法能夠優(yōu)化肺結(jié)節(jié)篩查的管理,簡(jiǎn)化臨床工作流程,并有助于節(jié)省不必要的隨訪測(cè)試和支出,因此這個(gè)領(lǐng)域值得給予更多的關(guān)注。
未來(lái),肺結(jié)節(jié)的風(fēng)險(xiǎn)分層評(píng)估將以個(gè)體化的方式來(lái)評(píng)估其惡性概率。個(gè)體化方式依據(jù)的是患者的臨床信息、影像學(xué)特征和生物標(biāo)志物等,這意味著評(píng)估和管理肺結(jié)節(jié)的模式是可以重復(fù)的。但眼下的困境是缺乏不同的臨床環(huán)境對(duì)各類風(fēng)險(xiǎn)評(píng)估模型和算法進(jìn)行比較和校準(zhǔn),目前大多數(shù)風(fēng)險(xiǎn)評(píng)估都是基于隊(duì)列研究或回顧性研究,尚沒(méi)有正式的臨床試驗(yàn)比較不同評(píng)估方式的效果差異。隨著肺結(jié)節(jié)檢出率的不斷提高,將肺結(jié)節(jié)進(jìn)行風(fēng)險(xiǎn)分層既可減少侵入性手術(shù),避免過(guò)度醫(yī)療,又可早期識(shí)別惡性腫瘤,及早干預(yù)。深度學(xué)習(xí)算法的應(yīng)用能夠減少臨床醫(yī)生的工作量,緩解醫(yī)務(wù)人員供給需求。從長(zhǎng)遠(yuǎn)來(lái)看,生物標(biāo)志物、深度學(xué)習(xí)和原有的影像學(xué)方式的改進(jìn)都會(huì)帶來(lái)新的風(fēng)險(xiǎn)模型和分層策略,應(yīng)采取更積極的措施進(jìn)行驗(yàn)證、比較和應(yīng)用,進(jìn)而開(kāi)發(fā)出基于國(guó)內(nèi)人群的肺結(jié)節(jié)風(fēng)險(xiǎn)分層評(píng)估方式。
[1] ZHENG R S, ZHANG S W, ZENG H M, et al. Cancer incidence and mortality in China, 2016[J]. J Natl Cancer Cent, 2022, 2(1): 1–9.
[2] National Lung Screening Trial Research Team, ABERLE D R, ADAMS A M, et al. Reduced lung- cancer mortality with low-dose computed tomographic screening[J]. N Engl J Med, 2011, 365(5): 395–409.
[3] YANG W, QIAN F, TENG J, et al. Community-based lung cancer screening with low-dose CT in China: results of the baseline screening[J]. Lung Cancer, 2018, 117: 20–26.
[4] PACI E, PULITI D, LOPES PEGNA A, et al. Mortality, survival and incidence rates in the ITALUNG randomised lung cancer screening trial[J]. Thorax, 2017, 72(9): 825–831.
[5] CALLISTER M E, BALDWIN D R, AKRAM A R, et al. British Thoracic Society guidelines for the investigation and management of pulmonary nodules[J]. Thorax, 2015, 70(Suppl 2): ii1–ii54.
[6] 中華醫(yī)學(xué)會(huì)呼吸病學(xué)分會(huì)肺癌學(xué)組, 中國(guó)肺癌防治聯(lián)盟專家組. 肺結(jié)節(jié)診治中國(guó)專家共識(shí)(2018年版)[J]. 中華結(jié)核和呼吸雜志, 2018, 41(10): 763–771.
[7] MACMAHON H, NAIDICH D P, GOO J M, et al. Guidelines for management of incidental pulmonary nodules detected on CT images: from the Fleischner Society 2017[J]. Radiology, 2017, 284(1): 228–243.
[8] GOULD M K, DONINGTON J, LYNCH W R, et al. Evaluation of individuals with pulmonary nodules: when is it lung cancer? Diagnosis and management of lung cancer, 3rd ed: American College of Chest Physicians evidence-based clinical practice guidelines[J]. Chest, 2013, 143(5 Suppl): e93S–e120S.
[9] 孫可欣, 鄭榮壽, 張思維, 等. 2015年中國(guó)分地區(qū)惡性腫瘤發(fā)病和死亡分析[J]. 中國(guó)腫瘤, 2019, 28(1): 1–11.
[10] 孫可欣, 鄭榮壽, 曾紅梅, 等. 2014年中國(guó)肺癌發(fā)病和死亡分析[J]. 中華腫瘤雜志, 2018, 40(11): 805–811.
[11] DE KONING H J, VAN DER AALST C M, DE JONG P A, et al. Reduced lung-cancer mortality with volume CT screening in a randomized trial[J]. N Engl J Med, 2020, 382(6): 503–513.
[12] HOREWEG N, VAN ROSMALEN J, HEUVELMANS M A, et al. Lung cancer probability in patients with CT-detected pulmonary nodules: a prespecified analysis of data from the NELSON trial of low-dose CT screening[J]. Lancet Oncol, 2014, 15(12): 1332–1341.
[13] KURIYAMA K, YANAGAWA M. CT diagnosis of lung adenocarcinoma: radiologic-pathologic correlation and growth rate[J]. Radiology, 2020, 297(1): 199–200.
[14] HEUVELMANS M A, WALTER J E, VLIEGENTHARTR, et al. Disagreement of diameter and volume measurements for pulmonary nodule size estimation in CT lung cancer screening[J]. Thorax, 2018, 73(8): 779–781.
[15] LEE H J, SON H J, YUN M, et al. Prone position [18F]FDG PET/CT to reduce respiratory motion artefacts in the evaluation of lung nodules[J]. Eur Radiol, 2021, 31(7): 4606–4614.
[16] CHEN H, PANG Y, MENG T, et al.18F-FDG and68Ga-FAPI PET/CT in the evaluation of ground-glass opacity nodule[J]. Clin Nucl Med, 2021, 46(5): 424–426.
[17] WU J, LIU H, OU L, et al. FAPI uptake in a vertebral body fracture in a patient with lung cancer: a FAPI imaging pitfall[J]. Clin Nucl Med, 2021, 46(6): 520–522.
[18] GU B, LUO Z, HE X, et al.68Ga-FAPI and18F-FDG PET/CT images in a patient with extrapulmonary tuberculosis mimicking malignant tumor[J]. Clin Nucl Med, 2020, 45(11): 865–867.
[19] LUO Y, PAN Q, ZHANG W, et al. Intense FAPI uptake in inflammation may mask the tumor activity of pancreatic cancer in68Ga-FAPI PET/CT[J]. Clin Nucl Med, 2020, 45(4): 310–311.
[20] CUI X, HEUVELMANS M A, HAN D, et al. Comparison of Veterans Affairs, Mayo, Brock classification models and radiologist diagnosis for classifying the malignancy of pulmonary nodules in Chinese clinical population[J]. Transl Lung Cancer Res, 2019, 8(5): 605–613.
[21] DONG J, SUN N, LI J, et al. Development and validation of clinical diagnostic models for the probability of malignancy in solitary pulmonary nodules[J]. Thorac Cancer, 2014, 5(2): 162–168.
[22] XING W, SUN H, YAN C, et al. A prediction model based on DNA methylation biomarkers and radiological characteristics for identifying malignant from benign pulmonary nodules[J]. BMC Cancer, 2021, 21(1): 263.
[23] YANG D, ZHANG X, POWELL C A, et al. Probability of cancer in high-risk patients predicted by the protein-based lung cancer biomarker panel in China: LCBP study[J]. Cancer, 2018, 124(2): 262–270.
[24] LIANG W, LIU D, LI M, et al. Evaluating the diagnostic accuracy of a ctDNA methylation classifier for incidental lung nodules: protocol for a prospective, observational, and multicenter clinical trial of 10,560 cases[J]. Transl Lung Cancer Res, 2020, 9(5): 2016–2026.
[25] LING Z, CHEN J, WEN Z, et al. The value of a seven-autoantibody panel combined with the mayo model in the differential diagnosis of pulmonary nodules[J]. Dis Markers, 2021, 2021: 6677823.
[26] SILVESTRI G A, TANNER N T, KEARNEY P, et al. Assessment of plasma proteomics biomarker’s ability to distinguish benign from malignant lung nodules: results of the PANOPTIC (pulmonary nodule plasma proteomic classifier) trial[J]. Chest, 2018, 154(3): 491–500.
[27] OSTRIN E J, BANTIS L E, WILSON D O, et al. Contribution of a blood-based protein biomarker panel to the classification of indeterminate pulmonary nodules[J]. J Thorac Oncol, 2021, 16(2): 228–236.
[28] MARQUETTE C H, BOUTROS J, BENZAQUEN J, et al. Circulating tumour cells as a potential biomarker for lung cancer screening: a prospective cohort study[J]. Lancet Respir Med, 2020, 8(7): 709–716.
[29] ARDILA D, KIRALY A P, BHARADWAJ S, et al. Author correction: end-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography[J]. Nat Med, 2019, 25(8): 1319.
[30] BALDWIN D R, GUSTAFSON J, PICKUP L, et al. External validation of a convolutional neural network artificial intelligence tool to predict malignancy in pulmonary nodules[J]. Thorax, 2020, 75(4): 306–312.
[31] NAM J G, PARK S, HWANG E J, et al. Development and validation of deep learning-based automatic detection algorithm for malignant pulmonary nodules on chest radiographs[J]. Radiology, 2019, 290(1): 218–228.
[32] ALI I, HART G R, GUNABUSHANAM G, et al. Lung nodule detection via deep reinforcement learning[J]. Front Oncol, 2018, 8: 108.
[33] VENKADESH K V, SETIO A A A, SCHREUDER A, et al. Deep learning for malignancy risk estimation of pulmonary nodules detected at low-dose screening CT[J]. Radiology, 2021, 300(2): 438–447.
[34] MASQUELIN A H, CHENEY N, KINSEY C M, et al. Wavelet decomposition facilitates training on small datasets for medical image classification by deep learning[J]. Histochem Cell Biol, 2021, 155(2): 309–317.
[35] WANG Y W, WANG J W, YANG S X, et al. Proposing a deep learning-based method for improving the diagnostic certainty of pulmonary nodules in CT scan of chest[J]. Eur Radiol, 2021, 31(11): 8160–8167.
[36] LV W, WANG Y, ZHOU C, et al. Development and validation of a clinically applicable deep learning strategy (HONORS) for pulmonary nodule classification at CT: a retrospective multicentre study[J]. Lung Cancer, 2021, 155: 78–86.
[37] HU X, GONG J, ZHOU W, et al. Computer-aided diagnosis of ground glass pulmonary nodule by fusing deep learning and radiomics features[J]. Phys Med Biol, 2021, 66(6): 065015.
(2022–10–18)
(2023–08–24)
R521.6
A
10.3969/j.issn.1673-9701.2023.26.029
國(guó)家自然科學(xué)基金資助項(xiàng)目(81871873)
馮繼鋒,電子信箱:fjif@jszlyy.com.cn