Common computational methods of mass spectrometry imaging and their applications in tumors
Objective To introduce machine learning algorithms in mass spectrometry imaging(MSI)statistical analysis sce-narios and summarize the recent research progress in its application in tumors,promoting the application of MSI technolo-gy in tumor research.Methods Using"mass spectrometry imaging,image segmentation,classification,cancer"as key-words,relevant literature from the China National Knowledge Infrastructure(CNKI)and PubMed databases was re-trieved.Inclusion criteria include:(1)Literature related to the application of MSI in tumor image segmentation;(2)Lit-erature related to the application of MSI in tumor subtype classification;(3)Literature related to machine learning algo-rithms applied to tumor MSI data.Exclusion criteria include:(1)Literature with unclear research mechanisms;(2)Con-ference-related,review-type documents;(3)Literature that is similar or redundant.A total of 41 eligible documents were included.Results MSI is a rapidly developing technology for the spatial imaging of various biomolecules in recent years.This technique segments the surface of a sample into an array of pixel points.Each pixel is undergone mass spectrometry analysis,and the resulting spectral data are transformed into images showing the spatial distribution of compounds within the sample.Compared with traditional mass spectrometry techniques,MSI contains information about the spatial localiza-tion of molecules and has been widely used in tumor heterogeneity studies recently.A multitude of imaging algorithms have been applied to MSI research in tumors,aiming to delineate the heterogeneity of the tumor microenvironment through data analysis and visualization.With the ongoing development of soft ionization techniques and advancements in mass spectrometry equipment,the complexity of MSI data has increased.Proper statistical analysis of this data,focu-sing on dimension reduction and clustering,has become a pivotal research area.The main steps in the statistical analysis of tumor MSI data are dimensionality reduction and clustering.Various machine learning methods are employed to explore tumor-related molecular mechanisms through these analyses.Conclusions Existing computational methods have success-fully achieved tumor image segmentation,constructed tumor subtype classification models,and screened for biomarkers.These fundamental research achievements provide essential technical support for clinical studies focused on tumor diagno-sis,treatment,and prognosis.
mass spectrometry imagingcancerdimensionality reductionclustermachine learningreview literature
NETL
NSTL
万方数据
