Title : Ensemble based ild patterns classification using hybrid features set and multi level segmentation
Abstract:
The respiratory disease causing the death by wrong analysis might have an impressive effect on the patient’s health. In the pandemic resource limitation has significantly delayed the routine diagnosis of lung diseases patients including Interstitial Lung Disease (ILD). Early detection may improve the lives and health of patients. This paper proposed a methodology based on the Ensemble Classifier using hybrid feature (statistical and structural) set extracted by super-pixel image processing and fusion of K-means clusters (SPFKMC) on ILD images. Segmented super pixel images overcome the limitations of multiclass belongings. Semi-adaptive wavelet-based fusion is applied over selected K-means clusters. Segmentation output is then classified by Support Vector Machines (SVM) and Ensemble classifier. Based on the accuracy comparison of different classifiers, notice that ensemble learning methods such as Ensemble Boosted Tree (EBOT) give the accuracy score satisfactory 99.7% compared to Ensemble Bagged Tree (EBT).
Audience Take Away Notes:
- Super-pixel image processing and fusion of K-means clusters (SPFKMC) segmentation methodology
- Comparison of different classifiers for ILD images.
- Performance of classifier with hybrid feature sets.