Question
What is the Research and Null Hypothesis?
I was given this:
Fusing texture, shape and deep model-learned information at decision level for automated classification of lung nodules on chest CT Cross Mar Zhang Jianpengb, Xia Yongbt, Michael Fulhambde Biseedial a Muhineda Enginestin Technology mtm Research Gong, Schoolers arestion Techekges, Universi yef Syde. NSW 2006, Aastdi ART1CLE 1NFO ABSTRACT The separation o malignant from benign Jung nodules on chest computed tomegraphy (CT)impetant for the early detection o lung cancer, since early desection and managment offer the best chice for cure. Although deep learning methods have roeently peoduced a marked impeovemen in image elassification there are stil challenges as these methods contain myriad parameters and equire large scale trainin sets that are not usually available for most routine medical imaging studies thi‘ paper, we peopuse an algorithm for lung-dule classification that fuses the testure, shape and deep model-learned information (Fuse.TSD)the desion level. This algorithm employs a gray level co-oocurrence matrix (GLCMI-based texture descriptor, a Fourier shape descriptor to characterize the heterogeneity of nodules and deep convolutional neural network (DCNN) to automatically learn the feature representation of nodules on a slice-by sice basis. It trains an AdaBoosbed back propagation neural netwock (BPNN) using each feature type and fuses the decisions made by three classifiers to differentiate nodules We evaluated this algorithm agwinst three approaches on the LIDG-IDRI dataset. When the nodeles with a composite malignancy rate 3 were discarded, regarded as benign or regarded as malignant, our Fua-TSD ilgtrithm achieved an AUC of 9665%, 94.49% and 81.24%, respectively, which was substantially Lang rodule clasificat Back propagation neurel neework (PN AdaBeost, infornation fusion Computer-aided diagnosis (CAD), however, avoids many of these issues and is increasingly being investigated as an alternative and The 2015 global cancer statistics showed that there are approxi- complementary approach to conventional reading [4]. Mamy automated mately 14.1 million new cancer cases each year. Lung cancer has an ung nodule classification approaches have been proposed in the lit- incidence of 13% and death rate of 19.5%, which are the highest rates erature and most of them consist of image preprocessing, nodule de- across all cancers [1]. Early diagnosis and treatment are the most ef tction, nodule segmentation, feature extraction and classification fective means to improve survival of lung cancer patients the 5-year Among them, feature extraction is a critical step. The features used for survival for those with an early diagnosis is approximately 54 % when lung nodule classification can be divided into handerafted features and compared to 4% if the diagnosis is late when the patient has stage IV features leaned by deep neural networks (DNNs). Hand-crafted fea- disease [2). A "spo" on the lung, detected by chest computed tomo tures include texture and shape descriptors, since there is a high cor graphy (CT), which measures less than 3 em in diameter is defined as a respondence between nodule malignancy and heterogeneity in voxel lung nodule and may be benign or malignant. The National Lung values and shape IS). Once hand-crafted features are extracted, a Screening Trial showed that screening with CT will result in a 20% variety of classification techniques can be used including the support reduction in lung cancer deaths, by detecting early disease [3. Radi vector machine (SVM) [6,7, decision tree [8,9), K-nearest neighbor ologists globally typically visually analyze chest CT scans on a slice-by-(KNN) [10), back propagation neural network (BPNN) [1,12], random slice basis, which is time-consuming, expensive and prone to reader forest (RF) [13] and Adaboost [14,15] bias and requires a high degree of skill and concentration The most commonly used usual texture descriptor is based on the Raceived 22 October 303 Received in ried fom 12 Oetober 3037 Accepoed 17 October 2017 1566-2535/ © 2017 Ebevier B.V. All righs reserved
Explanation / Answer
Research
in order to get the advancement of knowledge research will be conducted. It involves collecting of data, information and facts.
Research is divided into 2 categories
A. Basic research to increase scientific knowledge
B. Applied research for solving problems, developing new product, process or tecnique. It uses the badic resesrch.
When the researchers tries to reject the assumption i.e. hypothesis is called the null hypothesis.
Null hypothesis are used to verify statistical assumption
1. To verify multiple experiments producing consistent result null hypothesis of homogenity is used.
2. Scientific null hypothesis are used to directly advance a theory.
