DISCUSSION
A literature search showed that there were no clinical studies on this
subject. Our study is the first one conducted on this subject. The
closest to our study was the one made by Alakuş et
al.11 where they developed clinical models predicting
COVID-19 using a hypothetical deep learning model based on the
laboratory data of COVID-19 patients. They tested 18 laboratory findings
of 600 patients and found that their experimental estimation model
detected the presence of COVID-19 with 86.6% accuracy, 86.7%
specificity and 62.5% AUC.
Using the WEKA Linear Regression Formula, the formulas found in our
study estimated the presence of involvement on thoracic CT at a rate of
78.92%, the extent of involvement on thoracic CT at 71.69%, and the
need for intensive care in patients with thoracic CT involvement at
91%. Using the J48 Decision Tree Formula, the formulas found in the
study estimated the presence of involvement on thoracic CT at a rate of
95.95%, the extent of involvement on thoracic CT at 84.39%, and the
need for intensive care at 93.06%. A major limitation of this study is
that the comorbidities of patients could not be assessed.
In the COVID-19 pandemic regions, it is important to diagnose the
disease quickly and obtain medical resources. Limited medical resources
have become a huge problem in pandemic regions. It is also quite
important to determine the severity of the disease and identify
priorities in treatment. The WEKA Linear Regression Formulas and J48
Decision Trees that were constructed with the help of the deep learning
method in our study were able to estimate the presence of lung
involvement, the extent of such involvement and the need for intensive
care in COVID-19 patients at a high rate. We think that the results of
this study will facilitate revealing the presence of lung involvement
and identification of critical patients in the COVID-19 pandemic and
particularly under circumstances where cases exponentially increase and
resources are restricted, and can be used effectively to ensure triage.