Deep Learning in Mass Spectrometry Imaging

Lead Research Organisation: University of Surrey
Department Name: Computing Science

Abstract

The research project aims to develop new and innovative machine learning algorithms to analyse the data from the new 3D OrbiSIMS instrument in a time and memory efficient manner. The work will concentrate upon machine learning, data-mining, statistical and pattern recognition techniques.

Publications

10 25 50

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/R512126/1 01/10/2017 30/09/2022
2075545 Studentship EP/R512126/1 02/07/2018 30/06/2022 Foivos Ntelemis
 
Title Information Maximization Clustering via Representation Learning Technique 
Description This newly developed algorithm, based on deep learning, extends a representation learning technique by operating simultaneously to provide effective clustering performance and dimensionality reduction (representation learning). Additionally, the computational time of the algorithm has been improved. This work has been submitted to a conference and is under review. 
Type Of Material Computer model/algorithm 
Year Produced 2021 
Provided To Others? No  
Impact This new algorithm presents a single-phase framework for data clustering and reduction. By training in a single phase, it provides an advantage in term of computation cost. The algorithm has shown strong robustness and effectiveness in various types of data. 
 
Title Visual Clustering via Generative Adversarial Networks and Information Maximisation 
Description The subject algorithm is an innovative deep learning technique which is developed during the past period for grouping visual and high dimensional data into a given number of categories based on their similarities. The algorithm has been evaluated successfully in Mass Spectrometry data. At current stage, this method is under review for publication. 
Type Of Material Computer model/algorithm 
Year Produced 2020 
Provided To Others? No  
Impact This new algorithm provides the facility for quick clustering high dimensional datasets including visual or spectra data. Currently, the algorithm has been used to remove the background in Mass Spectrometry Imaging.