| Autor: |
Alexandru, Pîrvu Cătălin, Nica, Cristian, Dragoş, Mărgăritescu, Ştefan, Pătraşcu, Şurlin, Valeriu, Sapalidis, Konstantions, Georgescu, Eugen, Georgescu, Ion, Pantea, Stelian |
| Předmět: |
|
| Zdroj: |
Journal of Surgical Sciences; Apr-Jun2021, Vol. 8 Issue 2, p42-49, 8p |
| Abstrakt: |
Acute intestinal ischemia (AMI) is a life-threatening surgical emergency where more than half of the affected patients do not survive. In spite of the medical advance, mortal-ity rates remain high due to late diagnosis, when proper surgical management and reperfusion techniques do not conclude to a successful outcome. The current study aims to find a proper diagnosis method with a high-reliability rate using machine learning (ML) algorithms. Methods: In this prospective cross-sectional study, we have collected and evaluated over the course of two years a total of 147 patients with a clini-cal presentation resembling acute mesenteric ischemia. Five ML algorithms, including Random Forest, Logistic Regression, Gradient Boosted Trees, Naive Bayes, and Multi-ple Layer Perceptron, were compared for their reliability in diagnosing acute intestinal ischemia by using regular blood tests performed in the emergency room (ER), on top of the main clinical characteristics of the researched condition. An algorithm score using Gradient Boosted Trees and Logistic Regression proved good diagnostic performance with an AUROC 0.784, p<0.001, with a sensitivity of 83.8%, specificity of 58.2%, 70.5% positive predictive value, and 75% negative predictive value. The ML algorithm is useful in detecting AMI using only anamnesis data and regular laboratory blood tests available in the ER, although it was not internally validated. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
