Train NNN model on lung CT scan of normal people and introducing new dataset

Background and Aim: The coronavirus has presented a challenge for medical staff worldwidein recent years. The need for rapid and accurate diagnosis, due to the working limits of medicalteams and the vast number of patients admitted, has created an opportunity for artificial intelligence-based diagnosis. Models used in this area require databases with sufficient informationduring the training process to be more accurate and reliable. While there are several normal CTscan databases available for comparison with COVID-۱۹ patient CT scans, none of them have alarge number of CT images. Additionally, a prepared normal CT scan model can help researchersachieve better results through transfer learning.Method: ۱۰,۰۰۰ chest CT scans with DICOM files that were reported as normal in radiologistresident reports will be extracted from the Rasoul Akram HIS database and used to train a normalCNN model. The model will be trained on the Pytorch Library on Python. The research is inprogress, but after completion of training, the model and DICOM File will be shared on GitHubwithout any personal information.Results: This data processing is in progress. The main goal of this article is to release a NormalChest CT Database for future research on Github (DICOM File and The Model).Conclusion: Artificial intelligence is a useful tool in helping medical staff reach faster and moreaccurate diagnoses in a shorter amount of time. We believe that providing a large dataset optimizedfor training these models will be beneficial to the future of medical diagnosis. A well-preparedmodule can help researchers get better results with fewer data and significantly decreaseprocessing time.