Pneumonia In Chest X-ray Research Articles

BSD: A multi-task framework for pulmonary disease classification using deep learning

The diagnosis of pulmonary diseases using deep learning on chest X-ray images can be affected by the bone structures, the tissue in regions outside the lungs, and the characteristics of the images as burrs, blurring and complex pulmonary structures. To address these issues, a new pulmonary diseases diagnostic framework named BSD is proposed, the innovations of which are: Firstly, three steps of bone suppression, pulmonary parenchyma extraction and pulmonary diseases diagnosis are included. The first two steps can eliminate the influence of bone structures and other tissues. Therefore, better diagnostic results can be obtained based on the extracted boneless pulmonary parenchyma. Secondly, aiming at the characteristics of the chest X-ray image, a new segmentation network named IU-Net is proposed for the pulmonary extraction, in which the original convolution layers of U-Net are replaced by our new designed function modules. Finally, based on the characteristic of boneless pulmonary parenchyma, an enhanced functional module RCA which can better extract the image features is designed, therefore improved the disease diagnosis. In the experiments on the chest X-ray pneumonia datasets, the framework achieved satisfied results in the binary classification with an accuracy and Kappa of 98.73% and 96.80%, respectively. And in the multi-classification task on the COVID-19 radiology database, it achieved an accuracy and Kappa of 95.76% and 95.75%, respectively, which is better than the state-of-the-art. Therefore, BSD framework can better facilitate physicians to diagnose pulmonary disease and increase the diagnostic capabilities.

Pixel embedding for grayscale medical image classification

In our paper, we present an extension of text embedding architectures for grayscale medical image classification. We introduce a mechanism that combines n-gram features with an efficient pixel flattening technique to preserve spatial information during feature representation generation. Our approach involves flattening all pixels in grayscale medical images using a combination of column-wise, row-wise, diagonal-wise, and anti-diagonal-wise orders. This ensures that spatial dependencies are captured effectively in the feature representations. To evaluate the effectiveness of our method, we conducted a benchmark using 5 grayscale medical image datasets of varying sizes and complexities. 10-fold cross-validation showed that our approach achieved test accuracy score of 99.92 % on the Medical MNIST dataset, 90.06 % on the Chest X-ray Pneumonia dataset, 96.94 % on the Curated Covid CT dataset, 79.11 % on the MIAS dataset and 93.17 % on the Ultrasound dataset. The framework and reproducible code can be found on GitHub at https://github.com/xizhou/pixel_embedding.

CycleGAN generated pneumonia chest x-ray images: Evaluation with vision transformer

The use of generative models in image synthesis has become increasingly prevalent. Synthetic medical imaging data is of paramount importance, primarily because medical imaging data is scarce, costly, and encumbered by legal considerations pertaining to patient confidentiality. Synthetic medical images offer a potential answer to these issues. The predominant approaches primarily assess the quality of images and the degree of resemblance between these images and the original ones employed for their generation.The central idea of the work can be summarized in the question: Do the performance metrics of Frechet Inception Distance(FID) and Inception Score(IS) in the Cycle-consistent Generative Adversarial Networks (CycleGAN) model are adequate to determine how real a generated chest x-ray pneumonia image is? In this study, a CycleGAN model was employed to produce artificial images depicting 3 classes of chest x-ray pneumonia images: general(any type), bacterial, and viral pneumonia. The quality of the images were evaluated assessing and contrasting 3 criteria: performance metric of CycleGAN model, clinical assessment of respiratory experts and the results of classification of a visual transformer(ViT). The overall results showed that the evaluation metrics of the CycleGAN are insufficient to establish realism in generated medical images.

Diagnostic approaches and treatments of systemic sclerosis patient with interstitial lung disease

Background: Systemic sclerosis or scleroderma is a rare disease with multiple complications which may include interstitial lung disease. The findings of diffuse type of systemic sclerosis, moderate restriction in pulmonary function test, interstitial pneumonia in chest x-ray, and ground glass opacity in thoracic CT-scan establish the diagnosis of interstitial lung disease associated with systemic sclerosis. We present a case of interstitial lung disease in a systemic sclerosis case, which was approached and treated comprehensively to slow the progression. Case Presentation: A 40-year-old female was admitted to the rheumatology polyclinic of Dr. Soetomo Hospital Surabaya with persistent dry cough for two years in June 2021. The patient also complained of breathlessness, painful joints, swollen knees, and morning stiffness three months prior to admission. The positive results of antinuclear antibody and rheumatoid factor were identified in 2016, which suggested autoimmune disease of scleroderma and being treated with azathioprine 50 mg twice a day, methylprednisolone 4 mg once a day, chloroquine 200 mg once a day, and calcium lactate once a day. The examinations in 2020 showed moderate restriction from the pulmonary function test, bilateral lung parenchymal infiltrates from the chest x-ray, and ground-glass opacity from the chest computed tomography scan (CT-scan). The patient was monthly prescribed cyclophosphamide 500 mg intravenously for six months as the substitution of azathioprine, methylprednisolone 4 mg once a day, and aspirin 100 mg once a day. The latest worsening condition led the patient to receive the substitution of cyclophosphamide which was mycophenolate mofetil 360 mg twice a day, along with aspirin 100mg and methylprednisolone 4 mg once a day, respectively. Conclusion: As a rare disease, the complications of systemic sclerosis should be anticipated. In this case, interstitial lung disease occurred following nearly five years of systemic sclerosis course which required immunosuppressive agents to enhance the efficacy of treatment. Comprehensive treatment with tight monitoring is necessary to slow the disease progression.

Spontaneously Regressing Retropharyngeal Neuroblastoma – A Rare Cause of Upper Airway Obstruction in a Neonate: A Case Report with Review of Literature

Abstract Background: Retropharyngeal masses may rarely cause persistent respiratory distress, stridor, and feeding difficulties in a neonate. We report a rare case of a primary cervical neuroblastoma presenting as a retropharyngeal mass, which underwent spontaneous regression. Clinical Description: A 29-day-old male baby was referred for recurrent vomiting since day 10 of life followed by severe difficulty in breathing since day 14 of life. At admission, the baby had stridor, cyanosis with respiratory failure, without any facial dysmorphism or cardiovascular involvement. Management and Outcome: After initial stabilization, investigations showed pneumonia in chest X-ray, with no evidence of sepsis. On intubation, a retropharyngeal bulge was noticed which was later confirmed by direct laryngoscopy. Later, computed tomography delineated the soft-tissue mass extending from C2–C5. Endoscopy with marsupialization of the tumor was done, and on histopathological examination, it was diagnosed as a small round blue cell tumor, suggestive of neuroblastoma, along with positive immunohistochemistry. According to the International Neuroblastoma Staging System, the tumor belonged to Stage I. As the baby had started improving clinically, further management by a multidisciplinary team was collectively decided to continue conservative observation. On follow-up, there was spontaneous regression in size of tumor with no recurrence of symptoms till the last follow-up at 11 months. Conclusion: Neuroblastoma, the most common solid tumor malignancy in children, may rarely have a primary cervical presentation, causing breathing and feeding difficulties. Depending on the course of symptoms, a strategy of observation and close monitoring by a multidisciplinary team can be taken, as such tumors may regress spontaneously.

Artificial Intelligence Applied to Chest X-ray: A Reliable Tool to Assess the Differential Diagnosis of Lung Pneumonia in the Emergency Department.

Considering the large number of patients with pulmonary symptoms admitted to the emergency department daily, it is essential to diagnose them correctly. It is necessary to quickly solve the differential diagnosis between COVID-19 and typical bacterial pneumonia to address them with the best management possible. In this setting, an artificial intelligence (AI) system can help radiologists detect pneumonia more quickly. We aimed to test the diagnostic performance of an AI system in detecting COVID-19 pneumonia and typical bacterial pneumonia in patients who underwent a chest X-ray (CXR) and were admitted to the emergency department. The final dataset was composed of three sub-datasets: the first included all patients positive for COVID-19 pneumonia (n = 1140, namely "COVID-19+"), the second one included all patients with typical bacterial pneumonia (n = 500, "pneumonia+"), and the third one was composed of healthy subjects (n = 1000). Two radiologists were blinded to demographic, clinical, and laboratory data. The developed AI system was used to evaluate all CXRs randomly and was asked to classify them into three classes. Cohen's κ was used for interrater reliability analysis. The AI system's diagnostic accuracy was evaluated using a confusion matrix, and 95%CIs were reported as appropriate. The interrater reliability analysis between the most experienced radiologist and the AI system reported an almost perfect agreement for COVID-19+ (κ = 0.822) and pneumonia+ (κ = 0.913). We found 96% sensitivity (95% CIs = 94.9-96.9) and 79.8% specificity (76.4-82.9) for the radiologist and 94.7% sensitivity (93.4-95.8) and 80.2% specificity (76.9-83.2) for the AI system in the detection of COVID-19+. Moreover, we found 97.9% sensitivity (98-99.3) and 88% specificity (83.5-91.7) for the radiologist and 97.5% sensitivity (96.5-98.3) and 83.9% specificity (79-87.9) for the AI system in the detection of pneumonia+ patients. Finally, the AI system reached an accuracy of 93.8%, with a misclassification rate of 6.2% and weighted-F1 of 93.8% in detecting COVID+, pneumonia+, and healthy subjects. The AI system demonstrated excellent diagnostic performance in identifying COVID-19 and typical bacterial pneumonia in CXRs acquired in the emergency setting.

Enhancing the confidence of deep learning classifiers via interpretable saliency maps

This paper quantifies the quality of heatmap-based eXplainable AI (XAI) methods w.r.t image classification problem. Here, a heatmap is considered desirable if it improves the probability of predicting the correct classes. Different XAI heatmap-based methods are empirically shown to improve classification confidence to different extents depending on the datasets, e.g. Saliency works best on ImageNet and Deconvolution on Chest X-ray Pneumonia dataset. The novelty includes a new gap distribution that shows a stark difference between correct and wrong predictions. Finally, the generative augmentative explanation is introduced, a method to generate heatmaps capable of improving predictive confidence to a high level.

Improved Transformer Based Hashing for Cased-based Pneumonia Retrieval

The deep hash method has been proved to be the most effective nearest neighbor search technique in large-scale image retrieval. However, up to now, deep hashing in medical image retrieval has been dominated by convolution network architecture, such as Resnet. Inspired by Transformer, we integrate Transformer into the deep learning framework. Our framework is mainly composed of two main modules: (1) In order to solve the problem of big difference among different subclasses of medical image categories, the Transformer model is improved in this paper. The external attention is used to replace the self-attention in the original Transformer model, and the feature acquisition ability of the model is improved by capturing the correlation between samples. (2) In addition, in order to learn compact binary coding, we adopt Bayesian framework for end-to-end learning. We conduct experiments on ChestX-ray Pneumonia dataset, and compare the proposed algorithm with the commonly used depth hashing method. Experimental results show that the MAP values under different hash bit codes are better than the existing deep hash methods, which proves the effectiveness of the proposed method.

Clinical features and outcomes of renal disease in children with COVID-19: A single-center retrospective study

Abstract Background: The clinical spectrum of coronavirus disease 2019 (COVID-19) ranges from asymptomatic course to severe illness. Children have accounted 1%–5% of diagnosed cases so far. This study aimed to observe the clinical features and outcomes of renal disease in children with COVID-19. Materials and Methods: This retrospective study was conducted in the Department of Pediatric Nephrology, Bangabandhu Sheikh Mujib Medical University, Dhaka, Bangladesh, from November 2021 to April 2022. The data were collected from 24 COVID-19-positive children under 18 years old with preexisting renal disease admitted in BSMMU from March 2020 to September 2021. Results: The mean age of the participants was 8.7 (±4.4) years. Among them, preexisting renal diseases were mainly nephrotic syndrome (50%), chronic kidney diseases (29%), and lupus nephritis (8%). Among the participants, 58% of patients became reverse transcription- polymerase chain reaction negative within 14 days, and the mean duration of hospital stay was 14.5 (±5.9) days. Most frequently presented clinical features were fever (75%), cough (71%), respiratory distress (54%), proteinuria (71%), hematuria (21%), diarrhea and acute kidney injury (17%), rapidly progressive glomerulonephritis (4%), heart failure (33%), and decreased saturation of peripheral oxygen ≤90% in 58% of cases. Mean white blood cell count was 8773.9 (±5178.1)/mm3, raised serum creatinine 62.5%. Pneumonia in chest X-ray was present in 50% of cases. Children getting more immunosuppressive drugs, especially those who got rituximab, had milder symptoms. Overall case-fatality rate was 25% with the highest rate in chronic kidney disease (CKD) patients. Conclusion: Children with preexisting renal disease got affected slightly at a higher percentage than the normal child, and those who got immunosuppressive therapy, especially rituximab, had milder symptoms. Children with CKD had fatal outcomes.

The effect of Gaussian noise on pneumonia detection on chest radiographs, using convolutional neural networks

IntroductionChest X-rays (CXR) with under-exposure increase image noise and this may affect convolutional neural network (CNN) performance. This study aimed to train and validate CNNs for classifying pneumonia on CXR as normal or pneumonia acquired at different image noise levels. MethodsThe study used the curated and publicly available “Chest X-Ray Pneumonia” dataset of 5856 AP CXR classified into 1583 normal, 4273 viral and bacterial pneumonia cases. Gaussian noise with zero mean was added to the images, at 5 image noise variance levels, corresponding to decreasing exposure. Each noise-level dataset was split into 80% for training, 10% for validation, and 10% for test data and then classified using custom trained sequential CNN architecture. Six classification tasks were developed for five Gaussian noise levels and the original dataset. Sensitivity, specificity, predictive values and accuracy were used as evaluation performance metrics. ResultsCNN evaluation on the different datasets revealed no performance drop from the original dataset to the five datasets with different noise levels. Sensitivity, specificity and accuracy for the normal datasets were 98.7%, 76.1% and 90.2%. For the five Gaussian noise levels the sensitivity, specificity and accuracy ranged from 96.9% to 98.2%, 94.4%–98.7% and 96.8%–97.6%, respectively. A heat map was used for visual explanation of the CNNs. ConclusionThe CNNs sensitivity maintained, and the specificity increased in distinguishing between normal and pneumonia CXR with the introduction of image noise. Implications for practiceNo performance drops of CNNs in distinguishing cases with and without pneumonia CXR with different Gaussian noise levels was observed. This has potential for decreasing radiation dose to patients or maintaining exposure parameters for patients that require additional radiographs.

Risk factors of late-onset Neonatal Sepsis in Special Care Neonatal Unit of a Tertiary Care Hospital in Bangladesh

Introduction: Late-onset neonatal sepsis is one of the most common causes of neonatal deaths in hospital as well as community in developing country like Bangladesh. It can present asmeningitis, pneumonia, pyelonephritis, or gastroenteritis. Various risk factors play important role in its causation. In this study our main goal was to evaluate the risk factors of late onset neonatal sepsis in SCANU (Special Care Newborn Unit) of a tertiary care hospital. Methods: This cross-sectional study was done in the SCANU of Dr. M.R. Khan Shishu Hospital and Institute of Child Health, Mirpur-2, Dhaka for 6 months. A total of 59 neonates with late onset sepsis in SCANU were included during the study period. Results: The socio-demographic status of the neonates with late onset sepsis shows that, most of the neonates (44.1%) were between 24 to 28 days. Also, 54.24% patients were male and 57.6% patients belong to middle economic status group. In the study, 79.7% had Cough with respiratory distress followed by 23.7% had cough only, 6.8% had fever, 52.5% had vomiting, abdominal distension, and 8.5% patients had Jaundice. Only 1.7% patients used bottles for feeding. Among neonates, the highest (49.15%) came with the risk factor of poor hygiene. The Majority (95%) of low-birthweight neonates had patchy opacity resembling pneumonia in chest x-ray. Conclusion: Poor hygiene is a high-risk factor for neonates with late onset sepsis. Bottle feeding had comparatively higher risk in developing late onset sepsis than formula feeding with spoon. Poor cord care is another risk factor. J Bangladesh Coll Phys Surg 2022; 40: 257-262

Moderate COVID-19: Clinical Trajectories and Predictors of Progression and Outcomes.

Background: Patients with COVID-19 commonly present at healthcare facilities with moderate disease, i.e., pneumonia without a need for oxygen therapy. Aim: To identify clinical/laboratory characteristics of patients with moderate COVID-19, which could predict disease progression. Methods: 384 adult patients presented with moderate COVID-19 and admitted to two hospitals were retrospectively evaluated. In a multivariate analysis gender, age, BMI, Charlson Comorbidity Index (CCI) and National Early Weaning Score 2 were treated as co-variates. The development of hypoxemic respiratory failure, intubation rate and risk of death were considered as dependent variables. Estimated values are presented as odds-ratio (OR) with 95% confidence interval (CI). Results: Most of the patients were male (63.28%) with a mean (standard deviation) age of 59 (16.04) years. Median (interquartile range) CCI was 2 (1–4). A total of 58.85% of the patients developed respiratory failure; 6.51% were intubated, and 8.85% died. The extent of pneumonia in chest X-ray (involvement of all four quartiles) [OR 3.96 (1.18–13.27), p = 0.026], respiratory rate [OR 1.17 (1.05–1.3), p = 0.004], SatO2 [OR 0.72 (0.58–0.88), p = 0.002], systolic blood pressure [OR 1.02 (1–1.04), p = 0.041] and lymphocyte count [OR 0.9993 (0.9986–0.9999), p = 0.026] at presentation were associated with the development of respiratory failure. The extent of pneumonia [OR 26.49 (1.81–387.18), p = 0.017] was associated with intubation risk. Age [OR 1.14 (1.03–1.26), p = 0.014] and the extent of pneumonia [OR 22.47 (1.59–316.97), p = 0.021] were associated with increased risk of death. Conclusion: Older age, the extent of pneumonia, tachypnea, lower SatO2, higher systolic blood pressure and lymphopenia are associated with dismal outcomes in patients presenting with moderate COVID-19.

Detection of Pneumonia in Chest X-Ray with Bat-Algorithm Selected Deep and Handcrafted Features

Pneumonia is one infectious disease caused by viruses/bacteria, and early screening is necessary for the detection and treatment. Furthermore, pneumonia causes severe problems in children and elderly. The proposed work aims to develop a disease screening scheme for efficiently classifying the chest radiograph (X-ray) pictures into the Normal/Pneumonia group. The proposed process has the below phases; (i) Image collecting and resizing, (ii) Deep-feature extraction, (iii) Handcrafted feature extraction, (iv) Bat-Algorithm based feature selection and (v) Classification. In this work, the VGG16 scheme is considered to extract the deep-features and the necessary handcrafted features are mined using the Weighted Local Binary Pattern (WLBP). The necessary feature is then selected using the bat-algorithm supported feature selection. The experimental result of this study proves the accuracy of KNN is healthier (>98%) than other methods.

Mediflex Diagnostics Detection of COVID, Pneumonia Using CNN

At the end of 2019, a new form of Coronavirus, called COVID-19, has widely spread in the world. To quickly screen patients with the aim to detect this new form of pulmonary disease, in this paper we propose a method aimed to automatically detect the COVID-19 disease by analyzing medical images. We exploit supervised machine learning techniques building a model considering a data-set freely available for research purposes of 85 chest X-rays. The experiment shows the effectiveness of the proposed method in the discrimination between the COVID-19 disease and other pulmonary diseases. Pneumonia remains a threat to human health; the coronavirus disease 2019 (COVID-19) that began at the end of 2019 had a major impact on the world. It is still raging in many countries and has caused great losses to people’s lives and property. In this paper, we present a method based on DeepConv-DilatedNet of identifying and localizing pneumonia in chest X-ray (CXR) images. Two-stage detector Faster R-CNN is adopted as the structure of a network. Feature Pyramid Network (FPN) is integrated into the residual neural network of a dilated bottleneck so that the deep features are expanded to preserve the deep feature and position information of the object. In the case of DeepConv-DilatedNet, the deconvolution network is used to restore high-level feature maps into its original size, and the target information is further retained.

Pneumonia Detection Using an Improved Algorithm Based on Faster R-CNN.

Pneumonia remains a threat to human health; the coronavirus disease 2019 (COVID-19) that began at the end of 2019 had a major impact on the world. It is still raging in many countries and has caused great losses to people's lives and property. In this paper, we present a method based on DeepConv-DilatedNet of identifying and localizing pneumonia in chest X-ray (CXR) images. Two-stage detector Faster R-CNN is adopted as the structure of a network. Feature Pyramid Network (FPN) is integrated into the residual neural network of a dilated bottleneck so that the deep features are expanded to preserve the deep feature and position information of the object. In the case of DeepConv-DilatedNet, the deconvolution network is used to restore high-level feature maps into its original size, and the target information is further retained. On the other hand, DeepConv-DilatedNet uses a popular fully convolution architecture with computation shared on the entire image. Then, Soft-NMS is used to screen boxes and ensure sample quality. Also, K-Means++ is used to generate anchor boxes to improve the localization accuracy. The algorithm obtained 39.23% Mean Average Precision (mAP) on the X-ray image dataset from the Radiological Society of North America (RSNA) and got 38.02% Mean Average Precision (mAP) on the ChestX-ray14 dataset, surpassing other detection algorithms. So, in this paper, an improved algorithm that can provide doctors with location information of pneumonia lesions is proposed.

PNEUMONIA DETECTION ON CHEST X-RAY USING RADIOMIC FEATURES AND CONTRASTIVE LEARNING.

Chest X-ray becomes one of the most common medical diagnoses due to its noninvasiveness. The number of chest X-ray images has skyrocketed, but reading chest X-rays still has been manually performed by radiologists, which creates huge burnouts and delays. Traditionally, radiomics, as a subfield of radiology that can extract a large number of quantitative features from medical images, demonstrates its potential to facilitate medical imaging diagnosis before the deep learning era. With the rise of deep learning, the explainability of deep neural networks on chest X-ray diagnosis remains opaque. In this study, we proposed a novel framework that leverages radiomics features and contrastive learning to detect pneumonia in chest X-ray. Experiments on the RSNA Pneumonia Detection Challenge dataset show that our model achieves superior results to several state-of-the-art models (> 10% in F1-score) and increases the model's interpretability.

COVID-19 in CXR: From Detection and Severity Scoring to Patient Disease Monitoring.

This work estimates the severity of pneumonia in COVID-19 patients and reports the findings of a longitudinal study of disease progression. It presents a deep learning model for simultaneous detection and localization of pneumonia in chest Xray (CXR) images, which is shown to generalize to COVID-19 pneumonia. The localization maps are utilized to calculate a “Pneumonia Ratio” which indicates disease severity. The assessment of disease severity serves to build a temporal disease extent profile for hospitalized patients. To validate the model's applicability to the patient monitoring task, we developed a validation strategy which involves a synthesis of Digital Reconstructed Radiographs (DRRs - synthetic Xray) from serial CT scans; we then compared the disease progression profiles that were generated from the DRRs to those that were generated from CT volumes.

Risk factors for disease progression in COVID-19 patients

BackgroundCoronavirus disease (COVID-19) is rapidly spreading worldwide. Although 10–20% of patients with COVID-19 have severe symptoms, little is known about the risk factors related to the aggravation of COVID-19 symptoms from asymptomatic or mild to severe disease states.MethodsThis retrospective study included 211 patients who were asymptomatic or with mild presentations of COVID-19. We evaluated the differences in demographic and clinical data between the cured (discharged to home) and transferred (aggravated to severe-stage COVID-19) groups.ResultsA multivariate logistic analysis showed that body temperature, chills, initial chest X-ray findings, and the presence of diabetes were significantly associated with predicting the progression to severe stage of COVID-19 (p < 0.05). The odds ratio of transfer in patients with COVID-19 increased by 12.7-fold for abnormal findings such as haziness or consolidation in initial chest X-ray, 6.32-fold for initial symptom of chills, and 64.1-fold for diabetes.ConclusionsEven if patients are asymptomatic or have mild symptoms, clinicians should closely observe patients with COVID-19 presenting with chills, body temperature > 37.5 °C, findings of pneumonia in chest X-ray, or diabetes.

Classification of pneumonia from X-ray images using siamese convolutional network

Pneumonia is one of the highest global causes of deaths especially for children under 5 years old. This happened mainly because of the difficulties in identifying the cause of pneumonia. As a result, the treatment given may not be suitable for each pneumonia case. Recent studies have used deep learning approaches to obtain better classification within the cause of pneumonia. In this research, we used siamese convolutional network (SCN) to classify chest x-ray pneumonia image into 3 classes, namely normal conditions, bacterial pneumonia, and viral pneumonia. Siamese convolutional network is a neural network architecture that learns similarity knowledge between pairs of image inputs based on the differences between its features. One of the important benefits of classifying data with SCN is the availability of comparable images that can be used as a reference when determining class. Using SCN, our best model achieved 80.03% accuracy, 79.59% f1 score, and an improved result reasoning by providing the comparable images.

GeminiNet: Combine Fully Convolution Network With Structure of Receptive Fields for Object Detection

Pneumonia is a relatively common disease that will endanger the lives of patients if left untreated. End-to-end detection of pneumonia using neural networks will be helpful for reducing related workforce. CNN's processing of images shows remarkable performance, naturally, the use of CNN based methods for assisted reading will be a trend in modern medicine. The property of current detection algorithms is not yet satisfactory, so further research is extremely needed. In this article, we design GeminiNet to identify and localize the pneumonia in Chest X-ray (CXR) images. It uses a popular fully convolution architecture with computation shared on the entire image, combining RoI Align and PSRoI Pooling to capture global and local information and output. Our approach introduces DetNet59, a network designed specifically for detection to capture deep features. In the sixth stage of DetNet59, the structure of the retina-like convolutional layers is added to replace the fully connected layer. This structure uses the dilated convolution to extend the reconstructive field, and the convolution kernels of three different scales are used for parallel calculation to collect rich feature information. GeminiNet is validated on the RSNA dataset. We augment dataset by flipping on horizontal and vertical for the small amount of data. At IoU (Intersection over Union) =0.5, AP reached 0.4575, 0.078 higher than ResNet50, and reached 0.7758 on the AUC indicator. GeminiNet achieves 8fps in detection speed, which is better than the 7fps of the popular Faster R-CNN architecture.