Detection of COVID-19 using AI - Deep learning and multiple medical imaging modalities

SARS-CoV2是一种新的冠状病毒，被确定为2019年冠状病毒病(新冠肺炎)的病原体，该病毒于2019年末在武汉、中国开始流行，并已在世界各地传播。这场大流行正在影响世界各地每个人的日常生活。目前，对新冠肺炎最常见的筛查和诊断测试是一种名为逆转录聚合酶链式反应的实验室测试。这项检测使用了从患者身上收集的样本(用鼻咽和/或喉咙拭子)。这项检测的特异性被认为是很高的。然而，其敏感度可低至60%-70%，导致大量假阴性，并增加社区传播的风险。其他令人担忧的问题是检测结果周转时间长，以及全球试剂和棉签短缺，这限制了可以进行的检测数量。专家表示，要想安全解除社会疏远措施，我们需要进行大量的测试。 医学成像手段也被用来检测新冠肺炎的迹象。最近，一个国际专家小组评估了影像技术在新冠肺炎管理中的应用，特别是胸部X线摄影(CXR)和计算机断层摄影(CT)。另一种有趣的方式显示了快速分诊新感染的前景，那就是胸部超声(CUS)。最近的研究表明，所有这些成像方式在抗击这一大流行方面都是有价值的，可以与RT-PCR实验室测试一起使用。该项目的主要目标是开发一种基于人工智能的工具，能够使用胸部的医学成像方式(X线摄影、超声检查和计算机断层扫描)来检测新冠肺炎。此外，我们将开发一种可解释的算法，以给出显示疾病迹象位置的视觉反馈。这种视觉反馈将非常有助于健康从业者对疾病的确认和分期。

SARS-CoV2 is a new coronavirus identified as the cause of the 2019 coronavirus disease (COVID-19) which started in Wuhan, China in late 2019 and has spread around the world. This pandemic is impacting the daily lives of everyone around the world. Currently, the most common screening and diagnostic test for COVID-19 is a laboratory test called reverse transcription polymerase chain reaction (RT-PCR). This test uses samples collected from the patient (with a nasopharyngeal and or throat swab). The specificity of this test is considered high. However, its sensitivity can be as low as 60-70%, leading to a significant number of false negatives and increasing the risk of community transmission. Other concerns are the test result long turnaround time and the worldwide shortage of reagents and swabs, which limits the number of tests that can be conducted. Experts say that for social distancing measures to be safely lifted, we will need to run a large number of tests. Medical imaging modalities have also been used to detect signs of COVID-19. More recently, an international panel of experts evaluated the utility of imaging technologies in the management of COVID-19, especially chest radiography (CXR) and computed tomography (CT). Another interesting modality that shows a lot of promise for fast triage of new infections is chest ultrasound (CUS). Recent research shows that all these imaging modalities can be valuable in the fight against this pandemic and can be used along with RT-PCR laboratory tests. The main objective of this project is to develop an AI-based tool capable of detecting COVID-19 using medical imaging modalities of the chest (radiography (CXR), ultrasonography (CUS) and computerized tomography (ChCT)). In addition, we will develop an explainability algorithm to give a visual feedback showing the location of the signs of the disease. This visual feedback will be very useful in helping the confirmation and the stage of the disease by a health practitioner.