Auxiliary Diagnosis for COVID-19 with Deep Transfer Learning.

To assist physicians identify COVID-19 and its manifestations through the automatic COVID-19 recognition and classification in chest CT images with deep transfer learning. In this retrospective study, the used chest CT image dataset covered 422 subjects, including 72 confirmed COVID-19 subjects (260 studies, 30,171 images), 252 other pneumonia subjects (252 studies, 26,534 images) that contained 158 viral pneumonia subjects and 94 pulmonary tuberculosis subjects, and 98 normal subjects (98 studies, 29,838 images). In the experiment, subjects were split into training (70%), validation (15%) and testing (15%) sets. We utilized the convolutional blocks of ResNets pretrained on the public social image collections and modified the top fully connected layer to suit our task (the COVID-19 recognition). In addition, we tested the proposed method on a finegrained classification task; that is, the images of COVID-19 were further split into 3 main manifestations (ground-glass opacity with 12,924 images, consolidation with 7418 images and fibrotic streaks with 7338 images). Similarly, the data partitioning strategy of 70%-15%-15% was adopted. The best performance obtained by the pretrained ResNet50 model is 94.87% sensitivity, 88.46% specificity, 91.21% accuracy for COVID-19 versus all other groups, and an overall accuracy of 89.01% for the three-category classification in the testing set. Consistent performance was observed from the COVID-19 manifestation classification task on images basis, where the best overall accuracy of 94.08% and AUC of 0.993 were obtained by the pretrained ResNet18 (P < 0.05). All the proposed models have achieved much satisfying performance and were thus very promising in both the practical application and statistics. Transfer learning is worth for exploring to be applied in recognition and classification of COVID-19 on CT images with limited training data. It not only achieved higher sensitivity (COVID-19 vs the rest) but also took far less time than radiologists, which is expected to give the auxiliary diagnosis and reduce the workload for the radiologists.

CT synthesis from multi-sequence MRI using adaptive fusion network.

OBJECTIVE: To investigate a method using multi-sequence magnetic resonance imaging (MRI) to synthesize computed tomography (CT) for MRI-only radiation therapy. APPROACH: We proposed an adaptive multi-sequence fusion network (AMSF-Net) to exploit both voxel- and context-wise cross-sequence correlations from multiple MRI sequences to synthesize CT using element- and patch-wise fusions, respectively. The element- and patch-wise fusion feature spaces were combined, and the most representative features were selected for modeling. Finally, a densely connected convolutional decoder was applied to utilize the selected features to produce synthetic CT images. MAIN RESULTS: This study includes a total number of 90 patients' T1-weighted MRI, T2-weighted MRI and CT data. The AMSF-Net reduced the average mean absolute error (MAE) from 52.88-57.23 to 49.15 HU, increased the peak signal-to-noise ratio (PSNR) from 24.82-25.32 to 25.63 dB, increased the structural similarity index measure (SSIM) from 0.857-0.869 to 0.878, and increased the dice coefficient of bone from 0.886-0.896 to 0.903 compared to the other three existing multi-sequence learning models. The improvements were statistically significant according to two-tailed paired t-test. In addition, AMSF-Net reduced the intensity difference with real CT in five organs at risk, four types of normal tissue and tumor compared with the baseline models. The MAE decreases in parotid and spinal cord were over 8% and 16% with reference to the mean intensity value of the corresponding organ, respectively. Further, the qualitative evaluations confirmed that AMSF-Net exhibited superior structural image quality of synthesized bone and small organs such as the eye lens. SIGNIFICANCE: The proposed method can improve the intensity and structural image quality of synthetic CT and has potential for use in clinical applications.

The valuable role of dynamic 18F FDG PET/CT-derived kinetic parameter Ki in patients with nasopharyngeal carcinoma prior to radiotherapy: A prospective study.

BACKGROUND AND PURPOSE: Dynamic positron emission tomography/computed tomography (PET/CT) served the potential role of characterizing malignant foci. The main objective of this prospective study was to explore the advantage of dynamic PET/CT imaging in characterizing nasopharyngeal carcinoma (NPC). METHODS AND MATERIALS: Patients with probable head and neck disease underwent a local dynamic PET/CT scan followed by a whole-body static scan. Patlak analysis was used to generate parametric influx rate constant (Ki) images from 48 frames obtained from a dynamic PET/CT scan. By delineating the volumes-of-interest (VOIs) of: primary tumor (PT), lymph node (LN), and normal nasopharyngeal tissues (N), we acquired the corresponding Ki mean and SUVmean of each site respectively to perform the quantitative statistical analysis. RESULTS: Qualified images of 71 patients with newly diagnosed NPC and 8 without nasopharyngeal malignant lesions were finally included. We found the correlations between Ki mean-PT and critical clinical features, including clinical stage (r = 0.368), T category (r = 0.643) and EBV-DNA copy status (r = 0.351), and Ki mean-PT differed within the group. SUVmean-PT showed correlations with clinical stage (r = 0.280) and T category (r = 0.472), but could hardly differ systematically within group of clinical features except T category. Ki mean-LN offered the positive correlations with N category (r = 0.294), M category (r = 0.238) and EBV-DNA copy status (r = 0.446), and differed within the group. In addition, Ki mean represented a sensitivity of 94.4 % and a specificity of 100 %, in distinguishing NPC from the non-NPC, when the cut-off was defined as 0.0106. When the cut-off of SUV being defined as 2.03, the sensitivity and specificity were both 100 %. CONCLUSION: Our research confirmed Ki compared favorably to SUV in characterizing NPC and found that Ki can serve as an effective imaging marker of NPC.

The impact of new modes of electronic communication in the treatment of severe acute respiratory syndrome coronavirus 2 infection.

BACKGROUND: More than 26,760,000 cases of SARS-CoV-2 infection have been reported globally to date. This study aimed to analyze the impact of new electronic communication tools in the diagnosis and treatment of patients with SARS-CoV-2 infection. METHODS: From January 20 to February 26, 2020, adult patients with laboratory-confirmed severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection who were treated in The Fifth Affiliated Hospital, Sun Yat-sen University, in Zhuhai, China, were recruited. Forty-seven eligible patients were enrolled and randomly classified into either the test group or the control group. All of the patients received the standard therapeutic regimen and routine ward rounds. The test group was subdivided into three subgroups: the first subgroup (5-minute group) was given an extra 5-minute ward round by WeChat voice call once daily for basic disease communication; the second subgroup (10-minute group) received an extra 10-minute ward round by WeChat voice call once daily for further detail; and the third subgroup (20-minute group) was given an extra 10-minute ward round via WeChat voice call once daily, as well as an extra 10 minutes every 3 days. The primary outcome was the duration of positive-to-negative conversion of SARS-CoV-2 nucleic acid diagnosed by the NAT (nucleic acid testing). RESULTS: In the test groups, the median time from diagnosis to the endpoint was 7.0 days [interquartile range (IQR), 3.8-10.8], compared with 10.0 days (IQR, 6.5-14.5) in the control group. It showed significant reduced the duration time of virus from positive to negative by the NAT (nucleic acid testing), (P=0.032) especially between the 10-minute subgroup (3.0 days; IQR, 3.0-7.5) and the control group (P=0.0065). CONCLUSIONS: The use of new modes of electronic communication can benefit patients during the COVID-19 pandemic and could be extremely valuable in addressing the shortage of medical protective equipment and reducing occupational risk of exposure to infection.

Meta-analysis on inoperable pancreatic cancer: a comparison between gemcitabine-based combination therapy and gemcitabine alone.

AIM: To compare gemcitabine-based combination therapy and gemcitabine (GEM) alone in patients with advanced pancreatic cancer (APCa) through meta-analysis. METHODS: MEDLINE and EMBASE searches were supplemented by information from trial registers of randomized controlled trials (RCTs) for GEM-based combination therapy and GEM alone for APCa. A quantitative meta-analysis was carried out by two reviewers based on the inclusion criteria from all available RCTs. The meta-analysis involved overall survival (OS), objective remission rate (ORR), clinical benefit rate (CBR), time to progress/progress free survival (TTP/PFS) and toxicity. RESULTS: The meta-analysis included 22 RCTs. There was significant improvement in the GEM combination group with regard to the 6-mo survival rate (RD = 0.04, 95% CI 0.01-0.06, P = 0.008), 1-year survival rate (RD = 0.03, 95% CI 0.01-0.05, P = 0.01), ORR (RD = 0.04, 95% CI 0.01-0.07, P = 0.02), CBR (RD = 0.10, 95% CI 0.02-0.17, P = 0.01) and 6-mo TTP/PFS (RD = 0.07, 95% CI 0.04-0.10, P < 0.00001). However, the Grade 3-4 toxicity set by WHO was higher for the GEM combination group for neutropenia (RD = 0.05, 95% CI 0.01-0.10, P = 0.02), thrombocytopenia (RD = 0.05, 95% CI 0.02-0.08, P = 0.002) and vomiting/nausea (RD = 0.03, 95% CI 0.00-0.05, P = 0.02). CONCLUSION: GEM-based combination therapy may improve the overall survival and palliation in optimal patients with APCa as compared with GEM alone.