Triplet Contents based Medical Image Retrieval System for Lung Nodules CT Images Retrieval and Recognition Application

Abstract

 Abstract: Content Based Medical Image Retrieval (CBMIR) has found its relevance in medical diagnosis by processing massive medical databases based on visual and semantic features and user preferences. In this paper we address two issues such as retrieval and recognition. We present a novel method called Triplet-CBMIR for lung nodules CT images retrieval and recognition application. A Triplet CBMIR is a combination of three properties: Visual Features (Shape and Texture), Semantic Features and Relevance Feedback. Dataset training is done using: Preprocessing, Feature Extraction, Selection, Nodules Sign Detection and Clustering. In preprocessing we perform image scaling, denoising and normalization. In feature extraction, two methods are presented such as Hybrid Discrete Wavelet Transform (DWT) and Discrete Cosine Transform (DCT), Bounding-Box based Convolutional Network (CNN) for visual and semantic features extraction. Then optimum set of feature vectors are selected using Mutual Information based Neighborhood Entropy (MINε). Based on selected features, lung nodule sign is detected using K-nearest Neighbor (KNN) algorithm in which Hassanat Distance used and similar images are grouped using Multi-Self organizing Map (SOM). For similarity measurement, d_1 distance metric is used. Benchmark dataset such as LISS and LIDC are used for the study. Performance matrices such as Average Precision Rate (APR), Average Retrieval Rate (ARR), Average Recognition Rate (ArR), Running Time found in the simulation results are compared with some other already present state-of-the-art works. The proposed method shows a significant improvement as compared to other existing methods.

A Novel Outlier-Robust Accuracy Measure for Machine Learning Regression Based on Hassanat Distance Metric

“…Hassanat distance metric is widely used for different Machine learning applications 26,27 , such as Epilepsy Diagnosis 28 , Alert systems 29 , Computing expectiles 30 , Content-based image retrieval 31,32 , Face Detection/Recognition 33,34 , Biometrics 35,36 , cancer classification 37 , fraudulent detection 38 , Validation of oversampling approach 39,40 , etc.…”