Gradient Boosting Ensembled Method for in-Vivo Brain Tumour Classification Using Hyperspectral Images

Abstract

In the past few years, the use of hyperspectral imaging for medicinal purposes has increased. Hyperspectral imaging is a trending topic in the remote sensing field. It is used to collect the spectral information present in the scene. Additionally, the combination of spectral and spatial data offers valuable data for classifying brain tumours. When combined with Machine Learning (ML) algorithms, HSI (Hyperspectral Imaging) can be utilised as a non-intrusive medical diagnosis tool. By using hyperspectral images in the medical field, we can classify cancers, tissues, blood vessels etc. This paper offers a gradient boosting based ensembled classification (MCGC) scheme for In-Vivo brain cancer classification. And also, a multi-scale CNN method feature extraction and graph-based clustering method for feature selection are applied to get accurate classification results. The dataset used for brain cancer classification is the In-Vivo brain cancer dataset. These dataset images are captured when the real-time brain tumour surgery was going on. Finally, we performed the experiment results of gradient boosting ensembled classification methods. Support vector machine (SVM) and Random Forest (RF) classification methods were used to compare the outcomes of classification. And in comparison, with the other existing methods, we got good outcomes.

Blind nonlinear unmixing using nonnegative matrix factorization based bi-objective autoencoder

“…NMF is a popular unmixing method used to extract the endmember and the abundance details together. Given a nonnegative matrix X which represents the hyperspectral image, we can find the matrix factors Y and Z based on classical nonnegative factorization as in (6).…”

“…where, E x A x is the nonlinear term. But to unmix using NMF, the format of the equation needs to be as in (6). So, the terms are modified as ( 8)- (10).…”

“…Identifying the pure endmembers and their abundance is an important challenge in most of the hyperspectral image analysis processes. The general hyperspectral image analysis includes image acquisition, feature extraction [2], band selection [3], classification [4], [5], and prediction [6] as per various applications [7]. To attain better image classification and prediction, the above said problem of mixed pixels have to be solved by the hyperspectral unmixing process, which decomposes the different endmembers present in pixels with their fractional abundance.…”

Enhancing Hyperspectral Image Classification for Land Use Land Cover With Dilated Neighborhood Attention Transformer and Crow Search Optimization

“…Its non-invasive method has revolutionized medical imaging, leading to more precise and time-saving cancer diagnoses. Hyperspectral imaging is an essential technology for military operations such as target detection, reconnaissance, and situational awareness, which helps our troops to remain ahead of enemy advances [4]. New advancements in sensor technology continue to expand the capabilities of hyperspectral imaging.…”

A diagnosis model for brain atrophy using deep learning and MRI of type 2 diabetes mellitus

“…These results suggest that there may be a complex relationship between brain health and cognitive performance in diabetic patients. The study in Tejasree and Agilandeeswari ( 2022 ) presents a gradient boosting-based ensembled classification technique to classify brain cancer in vivo using hyperspectral images. It employs a graph-based clustering approach for feature selection and a multi-scale CNN method for feature extraction.…”