| Autor: |
Ramyea, R., Kasthuri, N., Kumar, Kandasamy Senthil, Kavivarman, J., Keerthana, K., Keerthana, R. |
| Předmět: |
|
| Zdroj: |
AIP Conference Proceedings; 3/27/2024, Vol. 2966 Issue 1, p1-8, 8p |
| Abstrakt: |
Pulmonary fibrosis is a serious lung disorder occurs when the tissues in lungs become scarred and damaged. The stiff and thickened tissue makes it harder for lungs to work properly. Major causes for this disease are vulnerability to toxins such as coal dust, asbestos and silica. It makes the tissues to scar, thicken and leads to breathing difficulty. The pulmonary fibrosis has no cure so far. Due to the limited number of labelled data and also the presence of outliers in the dataset, the accurate prediction of pulmonary fibrosis is highly challenging. The risk factor and progression of the disease could be reduced when it is predicted well in advance. In this paper, a novel prediction system using natural gradient boosting and light gradient boosting classifier is discussed. Natural gradient boosting is an algorithm that utilizes probabilistic prediction and provides confidence value of -7.27. The algorithm is optimized by Sequential Least SQuares Programming (SLSQP) method, which further reduces the confidence value to -6.37. The light gradient boosting classifier follows leaf wise tree growth methodology and gives confidence value of about -4.95. Furthermore, the presence of noise in CT scan images are detected and removed using canny edge detection. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
