|
Prediction of Glioma Tumor Grade Using Blood Markers with Artificial Neural Network
|
|
| Author:
|
HAMID REZA KHAYAT KASHANI, MD, SHIRZAD AZHARI, MD, HASAN REZA MOHAMMADI, MD, KHALIL KOMLAKH, MD
|
| Abstract:
|
Background: Blood markers are used as non-invasive predictive factor for tumor grade. They vary in different grades of tumor and therefore, they can be used to predict the tumor grade.
Material and Methods: In this retrospective study, blood markers including the leukocyte, lymphocyte, neutrophil, monocyte, Neutrophil-Lymphocyte Ratio (NLR), Platelet-Lymphocyte Ratio (PLR), Lymphocyte-to-Monocyte Ratio (LMR) and derived Neutrophil-to-Lymphocyte Ratio (dNLR) were extracted. Firstly, the blood markers were statistically analyzed and then, an artificial neural network was used to assess the predictive value of blood markers. The network was trained with the data and then, its characteristics such as accuracy and reliability were determined.
Results: Totally, 139 patients were evaluated in this study. NLR and dNLR were significantly and positively correlated with the tumor grade and LMR and lymphocytes were significantly and negatively correlated with the tumor grade. LMR had the best diagnostic value among the blood markers in prediction of tumor grade (AUC = 0.655).
Artificial neural network in two separate designs represented higher accuracy and efficiency in differentiating the low and high grade glioma compared to the four grades of tumor and it could differentiate between the low and high grade glioma tumors with 74.8% of accuracy These results are reliable considering the specifications of the artificial neural network (AUC = 0.78).
Conclusion: This study revealed that the artificial neural network can be used to differentiate between the low and high grade glioma tumor with the accuracy of 74.8%.
|
| Keyword:
|
Glioma, Glioblastoma, Neutrophil to Lymphocyte ratio (NLR), Artificial Neural Network (ANN).
|
| EOI:
|
-
|
| DOI:
|
https://doi.org/10.31838/ijpr/2021.13.01.754
|
| Download:
|
Request For Article
|
|
|