Deep-learning algorithms for the interpretation of chest radiographs to aid in the triage of COVID-19 patients: A multicenter retrospective study

The recent medical applications of deep-learning (DL) algorithms have demonstrated their clinical efficacy in improving speed and accuracy of image interpretation. If the DL algorithm achieves a performance equivalent to that achieved by physicians in chest radiography (CR) diagnoses with Coronavirus disease 2019 (COVID-19) pneumonia, the automatic interpretation of the CR with DL algorithms can significantly reduce the burden on clinicians and radiologists in sudden surges of suspected COVID-19 patients. The aim of this study was to evaluate the efficacy of the DL algorithm for detecting COVID-19 pneumonia on CR compared with formal radiology reports. This is a retrospective study of adult patients that were diagnosed as positive COVID-19 cases based on the reverse transcription polymerase chain reaction among all the patients who were admitted to five emergency departments and one community treatment center in Korea from February 18, 2020 to May 1, 2020. The CR images were evaluated with a publicly available DL algorithm. For reference, CR images without chest computed tomography (CT) scans classified as positive for COVID-19 pneumonia were used given that the radiologist identified ground-glass opacity, consolidation, or other infiltration in retrospectively reviewed CR images. Patients with evidence of pneumonia on chest CT scans were also classified as COVID-19 pneumonia positive outcomes. The overall sensitivity and specificity of the DL algorithm for detecting COVID-19 pneumonia on CR were 95.6%, and 88.7%, respectively. The area under the curve value of the DL algorithm for the detection of COVID-19 with pneumonia was 0.921. The DL algorithm demonstrated a satisfactory diagnostic performance comparable with that of formal radiology reports in the CR-based diagnosis of pneumonia in COVID-19 patients. The DL algorithm may offer fast and reliable examinations that can facilitate patient screening and isolation decisions, which can reduce the medical staff workload during COVID-19 pandemic situations.

Download Full-text

Domain-Adaptive Artificial Intelligence-Based Model for Personalized Diagnosis of Trivial Lesions Related to COVID-19 in Chest Computed Tomography Scans

Journal of Personalized Medicine ◽

10.3390/jpm11101008 ◽

2021 ◽

Vol 11 (10) ◽

pp. 1008

Author(s):

Muhammad Owais ◽

Na Rae Baek ◽

Kang Ryoung Park

Keyword(s):

Computed Tomography ◽

Deep Learning ◽

State Of The Art ◽

Area Under The Curve ◽

Ground Glass Opacity ◽

Visual Assessment ◽

Ct Scans ◽

Execution Speed ◽

Ct Data ◽

Personalized Diagnosis

Background: Early and accurate detection of COVID-19-related findings (such as well-aerated regions, ground-glass opacity, crazy paving and linear opacities, and consolidation in lung computed tomography (CT) scan) is crucial for preventive measures and treatment. However, the visual assessment of lung CT scans is a time-consuming process particularly in case of trivial lesions and requires medical specialists. Method: A recent breakthrough in deep learning methods has boosted the diagnostic capability of computer-aided diagnosis (CAD) systems and further aided health professionals in making effective diagnostic decisions. In this study, we propose a domain-adaptive CAD framework, namely the dilated aggregation-based lightweight network (DAL-Net), for effective recognition of trivial COVID-19 lesions in CT scans. Our network design achieves a fast execution speed (inference time is 43 ms on a single image) with optimal memory consumption (almost 9 MB). To evaluate the performances of the proposed and state-of-the-art models, we considered two publicly accessible datasets, namely COVID-19-CT-Seg (comprising a total of 3520 images of 20 different patients) and MosMed (including a total of 2049 images of 50 different patients). Results: Our method exhibits average area under the curve (AUC) up to 98.84%, 98.47%, and 95.51% for COVID-19-CT-Seg, MosMed, and cross-dataset, respectively, and outperforms various state-of-the-art methods. Conclusions: These results demonstrate that deep learning-based models are an effective tool for building a robust CAD solution based on CT data in response to present disaster of COVID-19.

Download Full-text

DIAG a Diagnostic Web Application Based on Lung CT Scan Images and Deep Learning

Studies in Health Technology and Informatics - Public Health and Informatics ◽

10.3233/shti210175 ◽

2021 ◽

Author(s):

Amel Imene Hadj Bouzid ◽

Said Yahiaoui ◽

Anis Lounis ◽

Sid-Ahmed Berrani ◽

Hacène Belbachir ◽

...

Keyword(s):

Deep Learning ◽

Web Application ◽

Patient Management ◽

Image Interpretation ◽

Ct Scans ◽

Diagnostic Tools ◽

The World ◽

Aided Diagnosis ◽

Lung Ct ◽

Lung Ct Scan

Coronavirus disease is a pandemic that has infected millions of people around the world. Lung CT-scans are effective diagnostic tools, but radiologists can quickly become overwhelmed by the flow of infected patients. Therefore, automated image interpretation needs to be achieved. Deep learning (DL) can support critical medical tasks including diagnostics, and DL algorithms have successfully been applied to the classification and detection of many diseases. This work aims to use deep learning methods that can classify patients between Covid-19 positive and healthy patient. We collected 4 available datasets, and tested our convolutional neural networks (CNNs) on different distributions to investigate the generalizability of our models. In order to clearly explain the predictions, Grad-CAM and Fast-CAM visualization methods were used. Our approach reaches more than 92% accuracy on 2 different distributions. In addition, we propose a computer aided diagnosis web application for Covid-19 diagnosis. The results suggest that our proposed deep learning tool can be integrated to the Covid-19 detection process and be useful for a rapid patient management.

Download Full-text

Deep Learning for Diagnosis of Paranasal Sinusitis Using Multi-View Radiographs

Diagnostics ◽

10.3390/diagnostics11020250 ◽

2021 ◽

Vol 11 (2) ◽

pp. 250

Author(s):

Yejin Jeon ◽

Kyeorye Lee ◽

Leonard Sunwoo ◽

Dongjun Choi ◽

Dong Yul Oh ◽

...

Keyword(s):

Deep Learning ◽

Learning Algorithm ◽

Imaging Modality ◽

Maxillary Sinusitis ◽

Image Interpretation ◽

Area Under The Curve ◽

Temporal Separation ◽

Deep Learning Algorithm ◽

Line Imaging ◽

The One

Accurate image interpretation of Waters’ and Caldwell view radiographs used for sinusitis screening is challenging. Therefore, we developed a deep learning algorithm for diagnosing frontal, ethmoid, and maxillary sinusitis on both Waters’ and Caldwell views. The datasets were selected for the training and validation set (n = 1403, sinusitis% = 34.3%) and the test set (n = 132, sinusitis% = 29.5%) by temporal separation. The algorithm can simultaneously detect and classify each paranasal sinus using both Waters’ and Caldwell views without manual cropping. Single- and multi-view models were compared. Our proposed algorithm satisfactorily diagnosed frontal, ethmoid, and maxillary sinusitis on both Waters’ and Caldwell views (area under the curve (AUC), 0.71 (95% confidence interval, 0.62–0.80), 0.78 (0.72–0.85), and 0.88 (0.84–0.92), respectively). The one-sided DeLong’s test was used to compare the AUCs, and the Obuchowski–Rockette model was used to pool the AUCs of the radiologists. The algorithm yielded a higher AUC than radiologists for ethmoid and maxillary sinusitis (p = 0.012 and 0.013, respectively). The multi-view model also exhibited a higher AUC than the single Waters’ view model for maxillary sinusitis (p = 0.038). Therefore, our algorithm showed diagnostic performances comparable to radiologists and enhanced the value of radiography as a first-line imaging modality in assessing multiple sinusitis.

Download Full-text

Assessment of Ankle Fractures using Deep Learning Algorithms and Convolutional Neural Network

10.1101/2021.07.25.21261103 ◽

2021 ◽

Author(s):

Soheil Ashkani-Esfahani ◽

Reze Mojahed Yazdi ◽

Rohan Bhimani ◽

Gino M Kerkhoffs ◽

Mario Maas ◽

...

Keyword(s):

Deep Learning ◽

Imaging Techniques ◽

Image Interpretation ◽

Area Under The Curve ◽

Ground Truth ◽

Ankle Fractures ◽

Deep Convolutional Neural Networks ◽

Radiographic Images ◽

Single View ◽

Occult Fractures

Early and accurate detection of ankle fractures is crucial for reducing future complications. Radiographs are the most abundant imaging techniques for assessing fractures. We believe deep learning (DL) methods, through adequately trained deep convolutional neural networks (DCNNs), can assess radiographic images fast and accurate without human intervention. Herein, we aimed to assess the performance of two different DCNNs in detecting ankle fractures using radiographs compared to the ground truth. In this retrospective study, our DCNNs were trained using radiographs obtained from 1050 patients with ankle fracture and the same number of individuals with otherwise healthy ankles. Inception V3 and Renet50 pretrained models were used in our algorithms. Danis-Weber classification method was used. Out of 1050, 72 individuals were labeled as occult fractures as they were not detected in the primary radiographic assessment. Using single-view radiographs was compared with 3-views (anteroposterior, mortise, lateral) for training the DCNNs. Our DCNNs showed a better performance using 3-views images versus single-view based on greater values for accuracy, F-score, and area under the curve (AUC). The sensitivity and specificity in detection of ankle fractures using 3-views were 97.5% and 93.9% using Resnet50 compared to 98.7% and 98.6 using inception V3, respectively. Resnet50 missed 3 occult fractures while Inception V3 missed only one case. Clinical Significance: The performance of our DCNNs showed a promising potential that can be considered in developing the currently used image interpretation programs or as a separate assistant to the clinicians to detect ankle fractures faster and more precisely.

Download Full-text

COVID-FACT: A Fully-Automated Capsule Network-Based Framework for Identification of COVID-19 Cases from Chest CT Scans

Frontiers in Artificial Intelligence ◽

10.3389/frai.2021.598932 ◽

2021 ◽

Vol 4 ◽

Author(s):

Shahin Heidarian ◽

Parnian Afshar ◽

Nastaran Enshaei ◽

Farnoosh Naderkhani ◽

Moezedin Javad Rafiee ◽

...

Keyword(s):

Health Care ◽

Deep Learning ◽

Early Diagnosis ◽

Data Augmentation ◽

Spatial Information ◽

Area Under The Curve ◽

Spatial Relations ◽

Building Blocks ◽

Ct Scans ◽

Rapid Separation

The newly discovered Coronavirus Disease 2019 (COVID-19) has been globally spreading and causing hundreds of thousands of deaths around the world as of its first emergence in late 2019. The rapid outbreak of this disease has overwhelmed health care infrastructures and arises the need to allocate medical equipment and resources more efficiently. The early diagnosis of this disease will lead to the rapid separation of COVID-19 and non-COVID cases, which will be helpful for health care authorities to optimize resource allocation plans and early prevention of the disease. In this regard, a growing number of studies are investigating the capability of deep learning for early diagnosis of COVID-19. Computed tomography (CT) scans have shown distinctive features and higher sensitivity compared to other diagnostic tests, in particular the current gold standard, i.e., the Reverse Transcription Polymerase Chain Reaction (RT-PCR) test. Current deep learning-based algorithms are mainly developed based on Convolutional Neural Networks (CNNs) to identify COVID-19 pneumonia cases. CNNs, however, require extensive data augmentation and large datasets to identify detailed spatial relations between image instances. Furthermore, existing algorithms utilizing CT scans, either extend slice-level predictions to patient-level ones using a simple thresholding mechanism or rely on a sophisticated infection segmentation to identify the disease. In this paper, we propose a two-stage fully automated CT-based framework for identification of COVID-19 positive cases referred to as the “COVID-FACT”. COVID-FACT utilizes Capsule Networks, as its main building blocks and is, therefore, capable of capturing spatial information. In particular, to make the proposed COVID-FACT independent from sophisticated segmentations of the area of infection, slices demonstrating infection are detected at the first stage and the second stage is responsible for classifying patients into COVID and non-COVID cases. COVID-FACT detects slices with infection, and identifies positive COVID-19 cases using an in-house CT scan dataset, containing COVID-19, community acquired pneumonia, and normal cases. Based on our experiments, COVID-FACT achieves an accuracy of 90.82%, a sensitivity of 94.55%, a specificity of 86.04%, and an Area Under the Curve (AUC) of 0.98, while depending on far less supervision and annotation, in comparison to its counterparts.

Download Full-text

Deep learning combined with radiomics may optimize the prediction in differentiating high-grade lung adenocarcinomas in ground glass opacity lesions on CT scans

European Journal of Radiology ◽

10.1016/j.ejrad.2020.109150 ◽

2020 ◽

Vol 129 ◽

pp. 109150

Author(s):

Xing Wang ◽

Li Zhang ◽

Xin Yang ◽

Lei Tang ◽

Jie Zhao ◽

...

Keyword(s):

Deep Learning ◽

Ground Glass Opacity ◽

Ct Scans ◽

Ground Glass ◽

High Grade ◽

Lung Adenocarcinomas

Download Full-text

Utility and costs of radiologist interpretation of perioperative imaging in patients with traumatic single-level thoracolumbar fractures

Journal of Neurosurgery Spine ◽

10.3171/2017.4.spine16923 ◽

2017 ◽

Vol 27 (5) ◽

pp. 578-583 ◽

Cited By ~ 1

Author(s):

Michael H. Weber ◽

Lojan Sivakumaran ◽

Maryse Fortin ◽

Alisson R. Teles ◽

Jeff D. Golan ◽

...

Keyword(s):

Incidental Findings ◽

Health Centre ◽

Image Interpretation ◽

Ct Scans ◽

Thoracolumbar Fractures ◽

Potential Cost ◽

Single Level ◽

Radiology Reports ◽

The Cost ◽

The Impact

OBJECTIVEThe cost of spine management is rising. As diagnostic imaging accounts for approximately 10% of total patient care spending, there is interest in determining if economies could be made with regard to the routine consultation of radiology for image interpretation. In the context of spine trauma, both the spine surgeon and the radiologist interpret perioperative imaging. Authors of the present study investigated the impact of radiologist interpretation of perioperative imaging from patients with traumatic single-level thoracolumbar fractures given that spine surgeons are expected to be comfortable interpreting pathologies of the musculoskeletal system.METHODSThe authors conducted a retrospective review of all patients presenting with a single-level thoracolumbar fracture treated at the McGill University Health Centre in the period from January 2003 to December 2010. The time between image capture and radiologist interpretation as well as the number of extraskeletal and/or incidental findings was extracted from the radiology reports on all perioperative images including radiographic, fluoroscopic, and CT images. The cost of interpretation was obtained from the provincial health insurance entity of Quebec.RESULTSEighty-two patients met the study inclusion criteria. Radiologists took a median of 1 day (IQR 0–5.5 days) to interpret preoperative radiographs. Intraoperative fluoroscopic images and postoperative radiographs were read by the radiologist a median of 19 days (IQR 4–56.75 days) and 34 days (IQR 1–137.5 days) after capture, respectively (p < 0.05). Preoperative radiologist dictations reported extraskeletal and/or incidental findings for 8.1% of radiographs; there were no intraoperative or postoperative extraskeletal findings beyond those previously reported on the preoperative radiographs. Radiologists took a median of 1 day (IQR 0–1 day) to read both preoperative and postoperative CT scans; extraskeletal and/or incidental findings were present in 46.2% of preoperative reports and 4.5% of postoperative reports. There were no intraoperative or postoperative radiological findings that provoked reoperation. A total of 66 intraoperative fluoroscopy images and 225 postoperative radiographs were read for a cost of $1399.20 and $1867.50 (Canadian dollars), respectively, for radiologist interpretation. This cost amounted to 40.3% of all perioperative image interpretation spending.CONCLUSIONSIn the management of single-level thoracolumbar fractures, radiologists add information to the diagnostic picture when interpreting preoperative radiographs and perioperative CT scans; however, the interpretation of intraoperative fluoroscopic images and postoperative radiographs comes with significant delay, does not add additional information, and represents an area of potential cost and professional-resource reduction.

Download Full-text

Automated Chest Radiographs Triage Reading by a Deep Learning Referee Network

10.1101/2021.06.01.21257399 ◽

2021 ◽

Author(s):

Rafael Lopez-Gonzalez ◽

Jose Sanchez-Garcia ◽

Belen Fos-Guarinos ◽

Fabio Garcia-Castro ◽

Angel Alberich-Bayarri ◽

...

Keyword(s):

Neural Network ◽

Deep Learning ◽

Language Processing ◽

Large Scale ◽

Area Under The Curve ◽

Chest Radiographs ◽

Radiology Reports ◽

Chest X Ray ◽

Diagnosis Tool ◽

Fully Connected

Chest radiographs are often obtained as a screening for early diagnosis tool to rule out abnormalities mainly related to different cardiovascular and respiratory diseases. Reading and reporting numerous chest radiographs is a complex and time-consuming task. This research proposes and evaluates a deep learning (DL) approach based on convolutional neural networks (CNN) combined with a referee fully connected neural network as a computer-aided diagnosis tool in chest X-ray triage and worklist prioritization. The CNN models were trained with a combination of three large scale databases: ChestX-ray14, CheXpert and PadChest. The final database contained 327,176 images labeled with findings obtained by natural language processing (NLP) techniques applied to the radiology reports. The dataset was split in 16 different balanced binary partitions, which were used to train 16 finding-specific classification CNNs. Afterwards, a normal vs abnormal partition of the dataset was created, being abnormal the presence of at least one pathologic change. This final partition was used to train a fully connected neural network as referee that was fed with all the 16 previously trained outcomes. The Area Under the Curve (AUC) analysis evaluated and compared the performance of the models. The system was successfully implemented and evaluated with a test set of 3400 images. The AUC of the normal vs abnormal classification was 0.94. The highest AUC of the finding-specific classifiers was 0.99 for hernia. The proposed system can be used to assist radiologists identifying abnormal exams, allowing a time-efficiency triage approach.

Download Full-text