RESUMO
Background Evaluation of interstitial lung disease (ILD) at CT is a challenging task that requires experience and is subject to substantial interreader variability. Purpose To investigate whether a proposed content-based image retrieval (CBIR) of similar chest CT images by using deep learning can aid in the diagnosis of ILD by readers with different levels of experience. Materials and Methods This retrospective study included patients with confirmed ILD after multidisciplinary discussion and available CT images identified between January 2000 and December 2015. Database was composed of four disease classes: usual interstitial pneumonia (UIP), nonspecific interstitial pneumonia (NSIP), cryptogenic organizing pneumonia, and chronic hypersensitivity pneumonitis. Eighty patients were selected as queries from the database. The proposed CBIR retrieved the top three similar CT images with diagnosis from the database by comparing the extent and distribution of different regional disease patterns quantified by a deep learning algorithm. Eight readers with varying experience interpreted the query CT images and provided their most probable diagnosis in two reading sessions 2 weeks apart, before and after applying CBIR. Diagnostic accuracy was analyzed by using McNemar test and generalized estimating equation, and interreader agreement was analyzed by using Fleiss κ. Results A total of 288 patients were included (mean age, 58 years ± 11 [standard deviation]; 145 women). After applying CBIR, the overall diagnostic accuracy improved in all readers (before CBIR, 46.1% [95% CI: 37.1, 55.3]; after CBIR, 60.9% [95% CI: 51.8, 69.3]; P < .001). In terms of disease category, the diagnostic accuracy improved after applying CBIR in UIP (before vs after CBIR, 52.4% vs 72.8%, respectively; P < .001) and NSIP cases (before vs after CBIR, 42.9% vs 61.6%, respectively; P < .001). Interreader agreement improved after CBIR (before vs after CBIR Fleiss κ, 0.32 vs 0.47, respectively; P = .005). Conclusion The proposed content-based image retrieval system for chest CT images with deep learning improved the diagnostic accuracy of interstitial lung disease and interreader agreement in readers with different levels of experience. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Wielpütz in this issue.
Assuntos
Aprendizado Profundo , Doenças Pulmonares Intersticiais/diagnóstico por imagem , Interpretação de Imagem Radiográfica Assistida por Computador/métodos , Tomografia Computadorizada por Raios X/métodos , Diagnóstico Diferencial , Feminino , Humanos , Pulmão/diagnóstico por imagem , Masculino , Pessoa de Meia-Idade , Reprodutibilidade dos Testes , Estudos RetrospectivosRESUMO
BACKGROUND. CT-guided percutaneous transthoracic needle biopsy (PTNB) is widely used for evaluation of indeterminate pulmonary lesions, although guidelines are lacking regarding the experience needed to gain sufficient skill. OBJECTIVE. The purpose of our study was to investigate the learning curve among a large number of operators in a tertiary referral hospital and to determine the number of procedures required to obtain acceptable performance. METHODS. This retrospective study included CT-guided PTNBs with coaxial technique performed by 17 thoracic imaging fellows from March 2, 2011, to August 8, 2017, who were novices in the procedure. A maximum number of 200 consecutive procedures per operator were included. The cumulative summation method was used to assess learning curves for diagnostic accuracy, false-negative rate, pneumothorax rate, and hemoptysis rate. Operators were assessed individually and in a pooled analysis. Pneumothorax risk was also assessed in a model adjusting for risk factors. Acceptable failure rates were defined as 0.1 for diagnostic accuracy and false-negative rate, 0.45 for pneumothorax rate, and 0.05 for hemoptysis rate. RESULTS. The study included 3261 procedures in 3134 patients (1876 men, 1258 women; mean age, 67.7 ± 12.1 [SD] years). Overall diagnostic accuracy was 94.2% (2960/3141). All 17 operators achieved acceptable diagnostic accuracy (37 procedures required in the pooled analysis; median, 33 procedures required [range, 19-67 procedures required]). Overall false-negative rate was 7.6% (179/2370). All 17 operators achieved acceptable false-negative rate (52 procedures required in the pooled analysis; median, 33 procedures required [range, 19-95 procedures required]). Pneumothorax occurred in 32.6% of the procedures (1063/3261 procedures), and hemoptysis occurred in 2.7% of the procedures (89/3261 procedures). All 17 operators achieved acceptable pneumothorax rate (20 procedures required in the pooled analysis; median, 19 procedures required [range, 7-63 procedures required]). In the risk-adjusted model, 15 operators achieved acceptable pneumothorax rate (54 procedures required in the pooled analysis; median, 36 procedures required [range, 10-192 procedures required]). Sixteen operators achieved acceptable hemoptysis rate (67 procedures required in the pooled analysis; median, 55 procedures required [range, 41-152 procedures required]). CONCLUSION. For CT-guided PTNB, at least 37 and 52 procedures are required to achieve acceptable diagnostic accuracy and false-negative rate, respectively. Not all operators achieved acceptable complication rates. CLINICAL IMPACT. The findings may help set standards for training, supervision, and ongoing assessment of operator proficiency for this procedure.
Assuntos
Competência Clínica/estatística & dados numéricos , Curva de Aprendizado , Neoplasias Pulmonares/diagnóstico por imagem , Neoplasias Pulmonares/patologia , Radiografia Intervencionista/métodos , Tomografia Computadorizada por Raios X/métodos , Adulto , Idoso , Idoso de 80 Anos ou mais , Biópsia por Agulha , Bolsas de Estudo , Feminino , Humanos , Biópsia Guiada por Imagem , Pulmão/diagnóstico por imagem , Pulmão/patologia , Masculino , Pessoa de Meia-Idade , Reprodutibilidade dos Testes , Estudos Retrospectivos , Sensibilidade e Especificidade , Centros de Atenção Terciária , Adulto JovemRESUMO
PURPOSE: To evaluate the added value of threshold growth (TG) for imaging criteria for diagnosing hepatocellular carcinoma (HCC) on gadoxetic acid-enhanced MRI. METHODS: Patients who underwent preoperative gadoxetic acid-enhanced MRI because of absence of 'definite HCC' (Liver Imaging Reporting and Data System category 5) on prior CT or MRI between January 2016 and December 2020 were retrospectively analyzed. The sensitivity and specificity for 'definite HCC' according to the criteria of the European Association for the Study of the Liver [EASL], Asian Pacific Association for the Study of the Liver [APASL], and Korean Liver Cancer Association-National Cancer Center [KLCA-NCC] were separately calculated with and without TG as a major imaging feature. The results were compared using generalized estimating equations. RESULTS: Of 202 nodules in 154 patients, 19 % showed TG. When TG was used as a major imaging feature, the sensitivity of EASL were significantly higher than when it was not used (59.2 % vs. 51.4 %, p = 0.001), whereas the sensitivities of APASL and KLCA-NCC did not significantly differ. No significant difference was found in the specificities of the three imaging criteria when TG was used or not (p ≥ 0.16). Of 11 HCCs additionally detected when TG was added to EASL criteria, 9 showed transitional-phase or hepatobiliary-phase hypointensity without portal venous-phase washout. CONCLUSION: TG had added value for improving the sensitivity of EASL criteria for gadoxetic acid-enhanced MRI without extending washout to transitional-phase or hepatobiliary-phase images.
Assuntos
Carcinoma Hepatocelular , Neoplasias Hepáticas , Humanos , Carcinoma Hepatocelular/diagnóstico por imagem , Carcinoma Hepatocelular/patologia , Neoplasias Hepáticas/diagnóstico por imagem , Neoplasias Hepáticas/patologia , Estudos Retrospectivos , Meios de Contraste , Gadolínio DTPA , Imageamento por Ressonância Magnética/métodos , Sensibilidade e EspecificidadeRESUMO
CT volumetry (CTV) has been widely used for pre-operative graft weight (GW) estimation in living-donor liver transplantation (LDLT), and the use of a deep-learning algorithm (DLA) may further improve its efficiency. However, its accuracy has not been well determined. To evaluate the efficiency and accuracy of DLA-assisted CTV in GW estimation, we performed a retrospective study including 581 consecutive LDLT donors who donated a right-lobe graft. Right-lobe graft volume (GV) was measured on CT using the software implemented with the DLA for automated liver segmentation. In the development group (n = 207), a volume-to-weight conversion formula was constructed by linear regression analysis between the CTV-measured GV and the intraoperative GW. In the validation group (n = 374), the agreement between the estimated and measured GWs was assessed using the Bland-Altman 95% limit-of-agreement (LOA). The mean process time for GV measurement was 1.8 ± 0.6 min (range, 1.3-8.0 min). In the validation group, the GW was estimated using the volume-to-weight conversion formula (estimated GW [g] = 206.3 + 0.653 × CTV-measured GV [mL]), and the Bland-Altman 95% LOA between the estimated and measured GWs was -1.7% ± 17.1%. The DLA-assisted CT volumetry allows for time-efficient and accurate estimation of GW in LDLT.