We developed a computer-aided diagnosis (CADx) method for classification between benign nodule, primary lung cancer, and metastatic lung cancer and evaluated the following: (i) the usefulness of the deep convolutional neural network (DCNN) for CADx of the ternary classification, compared with a conventional method (hand-crafted imaging feature plus machine learning), (ii) the effectiveness of transfer learning, and (iii) the effect of image size as the DCNN input. Among 1240 patients of previously-built database, computed tomography images and clinical information of 1236 patients were included. For the conventional method, CADx was performed by using rotation-invariant uniform-pattern local binary pattern on three orthogonal planes with a support vector machine. For the DCNN method, CADx was evaluated using the VGG-16 convolutional neural network with and without transfer learning, and hyperparameter optimization of the DCNN method was performed by random search. The best averaged validation accuracies of CADx were 55.9%, 68.0%, and 62.4% for the conventional method, the DCNN method with transfer learning, and the DCNN method without transfer learning, respectively. For image size of 56, 112, and 224, the best averaged validation accuracy for the DCNN with transfer learning were 60.7%, 64.7%, and 68.0%, respectively. DCNN was better than the conventional method for CADx, and the accuracy of DCNN improved when using transfer learning. Also, we found that larger image sizes as inputs to DCNN improved the accuracy of lung nodule classification.

We aimed to evaluate a computer-aided diagnosis (CADx) system for lung nodule classification focussing on (i) usefulness of the conventional CADx system (hand-crafted imaging feature + machine learning algorithm), (ii) comparison between support vector machine (SVM) and gradient tree boosting (XGBoost) as machine learning algorithms, and (iii) effectiveness of parameter optimization using Bayesian optimization and random search. Data on 99 lung nodules (62 lung cancers and 37 benign lung nodules) were included from public databases of CT images. A variant of the local binary pattern was used for calculating a feature vector. SVM or XGBoost was trained using the feature vector and its corresponding label. Tree Parzen Estimator (TPE) was used as Bayesian optimization for parameters of SVM and XGBoost. Random search was done for comparison with TPE. Leave-one-out cross-validation was used for optimizing and evaluating the performance of our CADx system. Performance was evaluated using area under the curve (AUC) of receiver operating characteristic analysis. AUC was calculated 10 times, and its average was obtained. The best averaged AUC of SVM and XGBoost was 0.850 and 0.896, respectively; both were obtained using TPE. XGBoost was generally superior to SVM. Optimal parameters for achieving high AUC were obtained with fewer numbers of trials when using TPE, compared with random search. Bayesian optimization of SVM and XGBoost parameters was more efficient than random search. Based on observer study, AUC values of two board-certified radiologists were 0.898 and 0.822. The results show that diagnostic accuracy of our CADx system was comparable to that of radiologists with respect to classifying lung nodules.

Journal of Medical Internet Research - International Scientific Journal for Medical Research, Information and Communication on the Internet

Aims The aim of this study was to elucidate the epidemiological background of central serous chorioretinopathy (CSC), including its incidence and treatment pattern. Methods This was a population-based longitudinal cohort study using a nationwide health insurance claims database of the Japan Ministry of Health, Labour and Welfare (MHLW). As Japan employs universal health coverage, the database covers more than 95% of claims issued in Japan. We accessed all data stored in the database with permission from the MHLW. We traced all individuals aged 30 years or older and identified individuals with new onset of CSC between January 2011 and December 2018. CSC cases were categorised by age and sex for each year, and incidence rate was calculated. We also identified major treatments for CSC to elucidate the initial treatment pattern. Results During the 8-year period, 247 930 incidences of CSC were identified, among which 75.9% were men. The crude incidence rate (per 100 000 person-years) in the general population aged 30 years or older was 34.0 (95% CI 33.9 to 34.2), in men was 54.2 (95% CI 53.9 to 54.4) and in women was 15.7 (95% CI 15.5 to 15.8). The mean age of onset was lower in men than in women (50.5±12.5 years vs 54.7±13.5 years). Most of the patients with newly diagnosed CSC (86.8%) did not receive major treatment. Conclusions The current study provides the nationwide population-based evidence to clarify the detailed epidemiology of CSC. These results could help to understand the pathogenesis and mechanisms of CSC in the future.

Rationale: The effects of telemedicine on adherence in patients with obstructive sleep apnea with long-term continuous positive airway pressure (CPAP) use have never been investigated. Objectives: To examine effects of a telemedicine intervention on adherence in long-term CPAP users. Methods: In a prospective, randomized, multicenter noninferiority trial conducted in 17 sleep centers across Japan, patients who had used CPAP for >3 months and were receiving face-to-face follow-up by physicians every 1 or 2 months were randomized by a coordinating center in a blind manner to the following three groups: 1) follow-up every 3 months accompanied by a monthly telemedicine intervention (telemedicine group: TM-group), 2) follow-up every 3 months (3-month group: 3M-group), or 3) monthly follow-up (1-month group: 1M-group). Each group was followed up for 6 months. The change in percentage of days with ≥4 h/night of CPAP use from baseline to the end of the study period was evaluated. A decline of ≥5% from baseline was considered deterioration of adherence. Noninferiority of TM- and 3M-groups compared with the 1M-group according to the number of patients with deterioration of adherence was evaluated with the Farrington and Manning test (noninferiority margin 15%). Results: A total of 483 patients were analyzed (median duration of CPAP use, 29 [interquartile range, 12–71] mo), and deterioration of adherence was found in 41 of 161 (25.5%), 55 of 166 (33.1%), and 35 of 156 (22.4%) patients in the TM-, 3M-, and 1M-groups, respectively. The noninferiority of the TM-group compared with the 1M-group was verified (difference in percentage of patients with adherence deterioration, 3.0%; 95% confidence interval [CI], −4.8% to 10.9%; P < 0.01). Conversely, the 3M-group did not show noninferiority to the 1M-group (percentage difference, 10.7%; 95% CI, 2.6% to 18.8%; P = 0.19). In the stratified analysis, adherence in TM- and 1M-group patients with poor adherence at baseline improved (TM: 45.8% ± 18.2% to 57.3% ± 24.4%; P < 0.01; 1M: 43.1% ± 18.5% to 53.6% ± 24.3%; P < 0.01), whereas that of the 3M-group did not (39.3% ± 20.8% to 39.8% ± 24.8%; P = 0.84). Conclusions: Intensive telemedicine support could help to optimize CPAP adherence even after long-term CPAP use. Clinical trial registered with www.umin.ac.jp/ctr/index.htm (trial number: UMIN000023118).

Cardiovascular and cerebrovascular diseases are frequently interconnected due to underlying pathology involving atherosclerosis and thromboembolism. The aim of this study was to investigate the impact of clinical interactions among cardiovascular and cerebrovascular diseases on patient outcomes using a large-scale nationwide claims-based dataset. Cardiovascular diseases were defined as myocardial infarction, heart failure, atrial fibrillation, and aortic dissection. Cerebrovascular diseases were defined as cerebral infarction, intracerebral hemorrhage, and subarachnoid hemorrhage. This retrospective study included 2,736,986 inpatient records (1,800,255 patients) at 911 hospitals from 2015 to 2016 from Japanese registry of all cardiac and vascular disease-diagnostic procedure combination dataset. Interactions among comorbidities and complications, rehospitalization, and clinical outcomes including in-hospital mortality were investigated. Among hospitalization records that involved cardiovascular disease, 5.9% (32,686 records) had cerebrovascular disease as a comorbidity and 2.1% (11,362 records) included an incident cerebrovascular complication after hospitalization. Cerebrovascular disease as a comorbidity or complication was associated with higher in-hospital mortality than no cerebrovascular disease (adjusted odds ratio (OR) [95% confidence interval]: 1.10 [1.06–1.14], 2.02 [1.91–2.13], respectively). Among 367,904 hospitalization records that involved cerebrovascular disease, 17.7% (63,647 records) had cardiovascular disease listed as comorbidity and 3.3% (11,834 records) as a complication. Only cardiovascular disease as a complication was associated with higher in-hospital mortality (adjusted OR [95% confidence interval]: 1.29 [1.22–1.37]). In addition, in-hospital mortality during rehospitalization due to the other disease was significantly higher than mortality during the hospitalization due to the first disease. In conclusion, substantial associations were observed between cardiovascular and cerebrovascular disease in a large-scale nationwide claims-based dataset; these associations had a significant impact on clinical outcomes. More intensive prevention and management of cardiovascular and cerebrovascular disease might be crucial.