On the Implementation of a Post-Pandemic Deep Learning Algorithm Based on a Hybrid CT-Scan/X-ray Images Classification Applied to Pneumonia Categories.

The identification and characterization of lung diseases is one of the most interesting research topics in recent years. They require accurate and rapid diagnosis. Although lung imaging techniques have many advantages for disease diagnosis, the interpretation of medial lung images has always been a major problem for physicians and radiologists due to diagnostic errors. This has encouraged the use of modern artificial intelligence techniques such as deep learning. In this paper, a deep learning architecture based on EfficientNetB7, known as the most advanced architecture among convolutional networks, has been constructed for classification of medical X-ray and CT images of lungs into three classes namely: common pneumonia, coronavirus pneumonia and normal cases. In terms of accuracy, the proposed model is compared with recent pneumonia detection techniques. The results provided robust and consistent features to this system for pneumonia detection with predictive accuracy according to the three classes mentioned above for both imaging modalities: radiography at 99.81% and CT at 99.88%. This work implements an accurate computer-aided system for the analysis of radiographic and CT medical images. The results of the classification are promising and will certainly improve the diagnosis and decision making of lung diseases that keep appearing over time.

SARS-Cov 2 infection and anti-tuberculosis immunity: temporal association or real protective role? (preprint)

Introduction: According to the literature consulted to date, there is epidemiological heterogeneity of covid 19 between countries depending on their vaccination policy, in particular BCG vaccination. These findings have led to several hypotheses, including the protective role of immunity induced by the BCG tuberculosis vaccine against Covid-19 infection. The immunity induced by the BCG vaccine significantly increases the secretion of pro-inflammatory cytokines, in particular IL-1B, which has been shown to play an essential role in antiviral immunity. This cross-immunity, although not specific, if highlighted, is a real providence that must be taken advantage of in the face of this pandemic. The main objective of this study is to rule out or confirm that anti-tuberculosis immunity protects against SARS-COV 2 in our context. Materiel and methods Two groups will be compared: cases infected with the virus and controls who have never been infected with the virus. Both case and control groups will undergo a tuberculin skin test: the intra dermal tuberculin reaction (IDR). Results: We found that our control group had a high IDR immunity value, with an IDR tuberculin positive percentage of 67.2%. This suggests that immunity to IDR is a protective factor against coronavirus disease. Conclusion: The hypothesis of nonspecific anti-tuberculosis protection deserves further verification studies; it would have large positive repercussions for developing countries.