Sample-Efficient Deep Learning for COVID-19 Diagnosis Based on CT Scans

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2020, Volume and Issue: unknown

Published: April 17, 2020

Abstract Coronavirus disease 2019 (COVID-19) has infected more than 1.3 million individuals all over the world and caused 106,000 deaths. One major hurdle in controlling spreading of this is inefficiency shortage medical tests. There have been increasing efforts on developing deep learning methods to diagnose COVID-19 based CT scans. However, these works are difficult reproduce adopt since data used their studies not publicly available. Besides, require a large number CTs train accurate diagnosis models, which obtain. In paper, we aim address two problems. We build publicly-available dataset containing hundreds scans positive for develop sample-efficient that can achieve high accuracy from even when training images limited. Specifically, propose Self-Trans approach, synergistically integrates contrastive self-supervised with transfer learn powerful unbiased feature representations reducing risk overfitting. Extensive experiments demonstrate superior performance our proposed approach compared several state-of-the-art baselines. Our achieves an F1 0.85 AUC 0.94 diagnosing scans, though just few hundred.

Language: Английский

---

COVID-ResNet: A Deep Learning Framework for Screening of COVID19 from Radiographs

arXiv (Cornell University), Journal Year: 2020, Volume and Issue: unknown

Published: Jan. 1, 2020

In the last few months, novel COVID19 pandemic has spread all over world. Due to its easy transmission, developing techniques accurately and easily identify presence of distinguish it from other forms flu pneumonia is crucial. Recent research shown that chest Xrays patients suffering depicts certain abnormalities in radiography. However, those approaches are closed source not made available community for re-producibility gaining deeper insight. The goal this work build open access datasets present an accurate Convolutional Neural Network framework differentiating cases cases. Our utilizes state art training including progressive resizing, cyclical learning rate finding discriminative rates fast residual neural networks. Using these techniques, we showed results on open-access COVID-19 dataset. This presents a 3-step technique fine-tune pre-trained ResNet-50 architecture improve model performance reduce time. We call COVIDResNet. achieved through progressively re-sizing input images 128x128x3, 224x224x3, 229x229x3 pixels fine-tuning network at each stage. approach along with automatic selection enabled us achieve accuracy 96.23% (on classes) COVIDx dataset only 41 epochs. presented computationally efficient highly multi-class classification three different infection types Normal individuals. can help early screening burden healthcare systems.

Language: Английский

---

CoroDet: A deep learning based classification for COVID-19 detection using chest X-ray images

Chaos Solitons & Fractals, Journal Year: 2020, Volume and Issue: 142, P. 110495 - 110495

Published: Nov. 23, 2020

Language: Английский

---

Blockchain-Federated-Learning and Deep Learning Models for COVID-19 Detection Using CT Imaging

IEEE Sensors Journal, Journal Year: 2021, Volume and Issue: 21(14), P. 16301 - 16314

Published: April 30, 2021

With the increase of COVID-19 cases worldwide, an effective way is required to diagnose patients. The primary problem in diagnosing patients shortage and reliability testing kits, due quick spread virus, medical practitioners are facing difficulty identifying positive cases. second real-world share data among hospitals globally while keeping view privacy concerns organizations. Building a collaborative model preserving major for training global deep learning model. This paper proposes framework that collects small amount from different sources (various hospitals) trains using blockchain based federated learning. Blockchain technology authenticates organization. First, we propose normalization technique deals with heterogeneity as gathered having kinds CT scanners. Secondly, use Capsule Network-based segmentation classification detect Thirdly, design method can collaboratively train privacy. Additionally, collected real-life data, which is, open research community. proposed utilize up-to-date improves recognition computed tomography (CT) images. Finally, our results demonstrate better performance

Language: Английский

---

The Role of Imaging in the Detection and Management of COVID-19: A Review

IEEE Reviews in Biomedical Engineering, Journal Year: 2020, Volume and Issue: 14, P. 16 - 29

Published: April 27, 2020

Coronavirus disease 2019 (COVID-19) caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is spreading rapidly around world, resulting in a massive death toll. Lung infection or pneumonia common complication of COVID-19, and imaging techniques, especially computed tomography (CT), have played an important role diagnosis treatment assessment disease. Herein, we review characteristics computing models that been applied for management COVID-19. CT, positron emission - CT (PET/CT), lung ultrasound, magnetic resonance (MRI) used detection, treatment, follow-up. The quantitative analysis data using artificial intelligence (AI) also explored. Our findings indicate typical their changes can play crucial roles detection In addition, AI other image methods are urgently needed to maximize value

Language: Английский

---