Deep Convolutional Neural Network System Research Articles

Artificial intelligence-assisted system for the assessment of Forrest classification of peptic ulcer bleeding: a multicenter diagnostic study.

Inaccurate Forrest classification may significantly affect clinical outcomes, especially in high risk patients. Therefore, this study aimed to develop a real-time deep convolutional neural network (DCNN) system to assess the Forrest classification of peptic ulcer bleeding (PUB). A training dataset (3868 endoscopic images) and an internal validation dataset (834 images) were retrospectively collected from the 900th Hospital, Fuzhou, China. In addition, 521 images collected from four other hospitals were used for external validation. Finally, 46 endoscopic videos were prospectively collected to assess the real-time diagnostic performance of the DCNN system, whose diagnostic performance was also prospectively compared with that of three senior and three junior endoscopists. The DCNN system had a satisfactory diagnostic performance in the assessment of Forrest classification, with an accuracy of 91.2% (95%CI 89.5%-92.6%) and a macro-average area under the receiver operating characteristic curve of 0.80 in the validation dataset. Moreover, the DCNN system could judge suspicious regions automatically using Forrest classification in real-time videos, with an accuracy of 92.0% (95%CI 80.8%-97.8%). The DCNN system showed more accurate and stable diagnostic performance than endoscopists in the prospective clinical comparison test. This system helped to slightly improve the diagnostic performance of senior endoscopists and considerably enhance that of junior endoscopists. The DCNN system for the assessment of the Forrest classification of PUB showed satisfactory diagnostic performance, which was slightly superior to that of senior endoscopists. It could therefore effectively assist junior endoscopists in making such diagnoses during gastroscopy.

Diagnosis of Breast Cancer with Strongly Supervised Deep Learning Neural Network

The strongly supervised deep convolutional neural network (DCNN) has better performance in assessing breast cancer (BC) because of the more accurate features from the slice-level precise labeling compared with the image-level labeling weakly supervised DCNN. However, manual slice-level precise labeling is time consuming and expensive. In addition, the slice-level diagnosis adopted in the DCNN system is incomplete and defective because of the lack of other slices’ information. In this paper, we studied the impact of the region of interest (ROI) and lesion-level multi-slice diagnosis in the DCNN auxiliary diagnosis system. Firstly, we proposed an improved region-growing algorithm to generate slice-level precise ROI. Secondly, we adopted the average weighting method as the lesion-level diagnosis criteria after exploring four different weighting methods. Finally, we proposed our complete system, which combined the densely connected convolutional network (DenseNet) with the slice-level ROI and the average weighting lesion-level diagnosis after evaluating the performance of five DCNNs. The proposed system achieved an AUC of 0.958, an accuracy of 92.5%, a sensitivity of 95.0%, and a specificity of 90.0%. The experimental results showed that our proposed system had a better performance in BC diagnosis because of the more precise ROI and more complete information of multi-slices.

Fully automated measurement on coronal alignment of lower limbs using deep convolutional neural networks on radiographic images

BackgroundA deep convolutional neural network (DCNN) system is proposed to measure the lower limb parameters of the mechanical lateral distal femur angle (mLDFA), medial proximal tibial angle (MPTA), lateral distal tibial angle (LDTA), joint line convergence angle (JLCA), and mechanical axis of the lower limbs.MethodsStanding X-rays of 1000 patients’ lower limbs were examined for the DCNN and assigned to training, validation, and test sets. A coarse-to-fine network was employed to locate 20 key landmarks on both limbs that first recognised the regions of hip, knee, and ankle, and subsequently outputted the key points in each sub-region from a full-length X-ray. Finally, information from these key landmark locations was used to calculate the above five parameters.ResultsThe DCNN system showed high consistency (intraclass correlation coefficient > 0.91) for all five lower limb parameters. Additionally, the mean absolute error (MAE) and root mean squared error (RMSE) of all angle predictions were lower than 3° for both the left and right limbs. The MAE of the mechanical axis of the lower limbs was 1.124 mm and 1.416 mm and the RMSE was 1.032 mm and 1.321 mm, for the right and left limbs, respectively. The measurement time of the DCNN system was 1.8 ± 1.3 s, which was significantly shorter than that of experienced radiologists (616.8 ± 48.2 s, t = -180.4, P < 0.001).ConclusionsThe proposed DCNN system can automatically measure mLDFA, MPTA, LDTA, JLCA, and the mechanical axis of the lower limbs, thus helping physicians manage lower limb alignment accurately and efficiently.

A deep learning-based segmentation system for rapid onsite cytologic pathology evaluation of pancreatic masses: A retrospective, multicenter, diagnostic study.

SummaryBackgroundWe aimed to develop a deep learning-based segmentation system for rapid on-site cytopathology evaluation (ROSE) to improve the diagnostic efficiency of endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA) biopsy.MethodsA retrospective, multicenter, diagnostic study was conducted using 5345 cytopathological slide images from 194 patients who underwent EUS-FNA. These patients were from Nanjing Drum Tower Hospital (109 patients), Wuxi People's Hospital (30 patients), Wuxi Second People's Hospital (25 patients), and The Second Affiliated Hospital of Soochow University (30 patients). A deep convolutional neural network (DCNN) system was developed to segment cell clusters and identify cancer cell clusters with cytopathological slide images. Internal testing, external testing, subgroup analysis, and human–machine competition were used to evaluate the performance of the system.FindingsThe DCNN system segmented stained cells from the background in cytopathological slides with an F1-score of 0·929 and 0·899–0·938 in internal and external testing, respectively. For cancer identification, the DCNN system identified images containing cancer clusters with AUCs of 0·958 and 0·948–0·976 in internal and external testing, respectively. The generalizable and robust performance of the DCNN system was validated in sensitivity analysis (AUC > 0·900) and was superior to that of trained endoscopists and comparable to cytopathologists on our testing datasets.InterpretationThe DCNN system is feasible and robust for identifying sample adequacy and pancreatic cancer cell clusters. Prospective studies are warranted to evaluate the clinical significance of the system.FundingJiangsu Natural Science Foundation; Nanjing Medical Science and Technology Development Funding; National Natural Science Foundation of China.

A Novel Deep Learning System for Diagnosing Early Esophageal Squamous Cell Carcinoma: A Multicenter Diagnostic Study.

This study aims to construct a real-time deep convolutional neural networks (DCNNs) system to diagnose early esophageal squamous cell carcinoma (ESCC) with white light imaging endoscopy. A total of 4,002 images from 1,078 patients were used to train and cross-validate the DCNN model for diagnosing early ESCC. The performance of the model was further tested with independent internal and external validation data sets containing 1,033 images from 243 patients. The performance of the model was then compared with endoscopists. The accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and Cohen kappa coefficient were measured to assess performance. The DCNN model had excellent performance in diagnosing early ESCC with a sensitivity of 0.979, a specificity of 0.886, a positive predictive value of 0.777, a negative predictive value of 0.991, and an area under curve of 0.954 in the internal validation data set. The model also depicted a tremendously generalized performance in 2 external data sets and exhibited superior performance compared with endoscopists. The performance of the endoscopists was markedly elevated after referring to the predictions of the DCNN model. An open-accessed website of the DCNN system was established to facilitate associated research. A real-time DCNN system, which was constructed to diagnose early ESCC, showed good performance in validation data sets. However, more prospective validation is needed to understand its true clinical significance in the real world.

A DCNN system based on an iterative method for automatic landmark detection in cephalometric X-ray images

Cephalometric analysis is a commonly used method in modern orthodontic clinical diagnosis and treatment. The traditional way of manually marking cephalometric landmarks takes a long time, and it is difficult to obtain stable detection accuracy because of the uneven professionalism of doctors. In order to efficiently and accurately locate the landmarks of X-ray cephalograms, we propose a deep convolutional neural network (DCNN) system for automatic landmark detection in cephalometric X-ray images. And the system incorporates an iterative method while introducing the ideas of transfer learning and data enhancement. We first preprocess and expand the dataset of the cephalometric X-ray images. Then we use the preprocessed image to train a deep convolutional neural network and obtain preliminary prediction areas for various landmarks. Next, for each type of landmark, a detection process based on the iterative method is used to gradually reduce the prediction range in five iterations, which achieves more precise landmark positioning. Finally, the system restores the final predicted position in the original image by reversing the predicted coordinates. Our training set and test set proposed in the 2015 IEEE International Symposium on Biomedical Imaging each contain 150 cephalometric X-ray images. The accuracy of the system predicting 19 pairs of landmarks in the image within the error range of two millimeters reaches 87.51% on average. The experimental results and comparisons show that the iterative method does improve the accuracy and stability of location prediction and our cephalometric image landmark detector is effective in locating landmarks.

Development and validation of a real-time artificial intelligence-assisted system for detecting early gastric cancer: A multicentre retrospective diagnostic study.

BackgroundWe aimed to develop and validate a real-time deep convolutional neural networks (DCNNs) system for detecting early gastric cancer (EGC).MethodsAll 45,240 endoscopic images from 1364 patients were divided into a training dataset (35823 images from 1085 patients) and a validation dataset (9417 images from 279 patients). Another 1514 images from three other hospitals were used as external validation. We compared the diagnostic performance of the DCNN system with endoscopists, and then evaluated the performance of endoscopists with or without referring to the system. Thereafter, we evaluated the diagnostic ability of the DCNN system in video streams. The accuracy, sensitivity, specificity, positive predictive value, negative predictive value and Cohen's kappa coefficient were measured to assess the detection performance.FindingThe DCNN system showed good performance in EGC detection in validation datasets, with accuracy (85.1%–91.2%), sensitivity (85.9%–95.5%), specificity (81.7%–90.3%), and AUC (0.887–0.940). The DCNN system showed better diagnostic performance than endoscopists and improved the performance of endoscopists. The DCNN system was able to process oesophagogastroduodenoscopy (OGD) video streams to detect EGC lesions in real time.InterpretationWe developed a real-time DCNN system for EGC detection with high accuracy and stability. Multicentre prospective validation is needed to acquire high-level evidence for its clinical application.FundingThis work was supported by the National Natural Science Foundation of China (grant nos. 81672935 and 81871947), Jiangsu Clinical Medical Center of Digestive System Diseases and Gastrointestinal Cancer (grant no. YXZXB2016002), and Nanjing Science and Technology Development Foundation (grant no. 2017sb332019).

A Fast Deep Perception Network for Remote Sensing Scene Classification

Current scene classification for high-resolution remote sensing images usually uses deep convolutional neural networks (DCNN) to extract extensive features and adopts support vector machine (SVM) as classifier. DCNN can well exploit deep features but ignore valuable shallow features like texture and directional information; and SVM can hardly train a large amount of samples in an efficient way. This paper proposes a fast deep perception network (FDPResnet) that integrates DCNN and Broad Learning System (BLS), a novel effective learning system, to extract both deep and shallow features and encapsulates a designed DPModel to fuse the two kinds of features. FDPResnet first extracts the shallow and the deep scene features of a remote sensing image through a pre-trained model on residual neural network-101 (Resnet101). Then, it inputs the two kinds of features into a designed deep perception module (DPModel) to obtain a new set of feature vectors that can describe both higher-level semantic and lower-level space information of the image. The DPModel is the key module responsible for dimension reduction and feature fusion. Finally, the obtained new feature vector is input into BLS for training and classification, and we can obtain a satisfactory classification result. A series of experiments are conducted on the challenging NWPU-RESISC45 remote sensing image dataset, and the results demonstrate that our approach outperforms some popular state-of-the-art deep learning methods, and present high-accurate scene classification within a shorter running time.

P3.13-037 Deep Learning System for Lung Nodule Detection

Ever since its first success in large scale image recognition problem, deep convolutional neural networks (DCNNs) have shown their capabilities in solving many challenging visual perceptual tasks, such as image classification, segmentation, object detection. In some cases, DCNNs have already achieved near-human performance. In medical image analysis, DCNNs have also been successfully applied to lesion detection, segmentation, and diagnosis. The power of DCNN lies in its ability to learn a hierarchical representation of raw input data, without hand-crafted features. The focus of this study was to improve the performance of DCNN in automatic detection of pulmonary nodule on CT scan. In particular, nodules whose size is less than 4mm were considered. A total of 171 scans were collected at Zhongshan Hospital Fudan University, which was first process by the proposed DCNN system (12Sigma). The detection results were then carefully reviewed by an experienced physician, all false positives and missed nodules were manually labeled. To refine the system, all false positives and true nodules of sizes 4mm and under were selected. The data set was randomly split into training and test sets, where the training set consists of 90% (154 scans) and the test set consists of 10% of data (17 scans). Fig 1 shows the FROC curves on the test set before and after re-training. Overall, the detection sensitivity improved at all false positive levels, but the improvement was most significant at low FP rate region. For example, when FP is 0.5 per scan, the detection sensitivity increased from 0.36 to 0.49. This improvement suggests that, even with limited data, the deep neural network can learn from its mistakes and be easily tuned to be more sensitive to small nodules. We believe that when more data is collected, more significant improve in high FP rate region will be observed.