ABSTRACT
[This corrects the article DOI: 10.1371/journal.pone.0250952.].
ABSTRACT
The development of medical assisting tools based on artificial intelligence advances is essential in the global fight against COVID-19 outbreak and the future of medical systems. In this study, we introduce ai-corona, a radiologist-assistant deep learning framework for COVID-19 infection diagnosis using chest CT scans. Our framework incorporates an EfficientNetB3-based feature extractor. We employed three datasets; the CC-CCII set, the MasihDaneshvari Hospital (MDH) cohort, and the MosMedData cohort. Overall, these datasets constitute 7184 scans from 5693 subjects and include the COVID-19, non-COVID abnormal (NCA), common pneumonia (CP), non-pneumonia, and Normal classes. We evaluate ai-corona on test sets from the CC-CCII set, MDH cohort, and the entirety of the MosMedData cohort, for which it gained AUC scores of 0.997, 0.989, and 0.954, respectively. Our results indicates ai-corona outperforms all the alternative models. Lastly, our framework's diagnosis capabilities were evaluated as assistant to several experts. Accordingly, We observed an increase in both speed and accuracy of expert diagnosis when incorporating ai-corona's assistance.
Subject(s)
COVID-19/diagnosis , Deep Learning , Thorax/diagnostic imaging , Tomography, X-Ray Computed , Area Under Curve , COVID-19/virology , Databases, Factual , Humans , Pneumonia/diagnosis , Pneumonia/pathology , RNA, Viral/analysis , RNA, Viral/metabolism , ROC Curve , Radiologists/psychology , Reverse Transcriptase Polymerase Chain Reaction , SARS-CoV-2/genetics , SARS-CoV-2/isolation & purification , Sensitivity and SpecificityABSTRACT
Designation of representative watersheds (RWs) as a reference area representing key behavior of the whole region is an essential tool to provide a time and cost-effective basis for monitoring watershed performance against different driving forces. It is more important in developing countries facing lack of necessary investments in one hand and ever-increasing human interventions and need to assess the outcome behavior of the system in another hand. However, this serious affair has been less considered worldwide, in general, and in developing countries, in particular. Therefore, in the present study, a quantitative-based method of Representative Watershed Index (RWI) with potential range from 0 to 100 has been formulated using four important criteria and available national-wide raster data of elevation (meter), slope (%), rainfall erosivity factor (t m ha-1 cm h-1), and land use. The approach was then applied to the data prepared for the unique and invaluable global water ecosystem of the Urmia Lake Basin (ULB), north-western Iran, as a case study. The input raster was overlaid via matrices programming in the MATrix LABoratory (MATLAB) 2016 and Geographic Information System (GIS) 9.3 software environments. The RWIs were accordingly computed for 61 sub-watersheds of the ULB. The RWIs resulted from quadri-partite dimensional matrices that varied from 5.54 to 53.46 with respective maximum dissimilarity and resemblance with the entire 61 study sub-watersheds in the region. However, the sub-watershed with RWI of 40.65 (No. 57) was proposed as the final RW for the whole ULB due to hydrological independency, appropriate locality, and existence of functioning meteorological and hydrometric stations. The identified RW would be suggested to be considered as the basis for future insight monitoring and assessing environmental issues for the region eventually leading to an appropriate adaptive watershed management. Graphical abstract á .
