Quantitative Analysis Of Brain-Dedicated All-Digital PET

Positron emission tomography(PET)is a clinical medical imaging technology with extremely high sensitivity to biology activities.Benefiting from its in vivo,dynamic and quantitative detection ability,PET has great potential in understanding the human brain by assessing several aspects of cerebral biology and biochemistry.However,due to the variety of factors affecting quantitative analysis,it is difficult to develop robust results in clinical applications,and the application of quantitative brain PET imaging cannot fully meet the needs of clinical diagnosis and treatment of brain diseases,which hinders the development of brain PET imaging applications.Facing this challenge,this work seeks breakthroughs from the two aspects: developing a quantitative analysis method and a brain-dedicated alldigital PET image analysis platform.Firstly,to address the challenge that traditional parameters such as standard uptake values cannot fully meet the needs of clinical quantitative analysis,a brain regions quantitative analysis method for PET images is proposed to deeply mine the plenty of underlying information from PET images.By using the high-throughput feature extraction method,the quantitative parameters are extracted from PET images.Then the quantitative relationship is established between the extracted features and clinical parameters,so as to be applied to clinical diagnosis,auxiliary treatment,curative effect evaluation,and prognosis prediction.The main steps are designed for the brain region PET image quantitative analysis,including image pre-processing,brain region extraction,feature extraction,analysis,and modeling.Based on the characteristics of brain PET images,the above steps are specialized.For instance,the methods of image pre-processing are specialized according to the characteristics of the PET images,especially the signal-to-noise ratio and resolution.Tailored brain region extraction methods based on the features of different tracers.In the step of feature extraction,the adaptive feature extraction method based on PET images is investigated.The neighborhood grey-level variance matrix with high sensitivity to brain PET images is proposed in order to improve the accuracy of subsequent modeling and analysis.Then,a quantitative analysis platform for the brain-dedicated all-digital PET is designed and established.In particular,the aforementioned new method is instantiated as a subsystem in the analysis platform.The core functional modules of the brain-dedicated alldigital PET analysis platform are achieved,including the cloud-based extension of the plugand-play imaging software system,and the development of functional modules for medical image processing and quantitative analysis of PET images of brain regions.Benefiting from the high-quality original data provided by the brain-dedicated all-digital PET system developed by our group,the cloud-based brain-dedicated all-digital PET software platform is designed and established to transfer the data collected from the PET imaging system to the analysis platform for subsequent processing and analysis.This analysis platform enables the optimal signal and data processing method can be selected to obtain high-quality PET images,according to the needs of clinical quantitative analysis.In addition,the platform establishes a complete labeled clinical database for brain disease research;provides a comprehensive and intuitive cloud-based analysis tool;supports the rapid expansion of new algorithms for quantitative cerebral disease analysis.The brain-dedicated all-digital PET analysis platform builds a pipeline for brain PET imaging quantitative analysis.Finally,hundreds of cases including more than ten kinds of brain diseases are collected based on the brain-dedicated all-digital PET imaging platform.And the new method for quantitative analysis of brain regions proposed above was used to conduct an applied study using Alzheimer’s disease(AD)and Parkinson’s disease(PD)as examples.The difference between early-onset AD(EOAD)and late-onset AD(LOAD)is studied.The feature extraction is performed on the [18F]florbetaben PET images and 2-[18F]FDG PET images.Then use the principal component analysis method for dimensionality reduction.We identified four principal components in the frontal and temporal lobes of AD patients that exhibited a significant correlation with age of onset.In EOAD patients,the clinical course was more aggressive than in LOAD patients.This finding provides a basis for revealing the underlying pathological mechanisms.Moreover,the identified textural features can be used as quantitative biomarkers for the diagnosis of EOAD and LOAD.The diagnosis method of PD based on PET images is studied.The image features sensitive to PD were extracted from the [18F] DOPA PET images and the 2-[18F]FDG PET images.There are differences between the PD group and the control group in the putamen brain region in the [18F]DOPA PET images.And also there are differences in the caudate nucleus,frontal lobe,and parietal lobe in the 2-[18F]FDG PET images.The identified texture features extracted from [18F]DOPA and 2-[18F]FDG PET images can be used as quantitative biomarkers for the early diagnosis of PD.To sum up,this paper addresses the challenges in the clinical application of quantitative brain PET image analysis,promotes the development of new methods and new platforms on the basis of clinical needs,and instructs the clinical application through the results obtained by the development of new methods and new platforms.This paper can lead to the innovation of clinical application of quantitative brain PET analysis and boost the development of quantitative medical imaging diagnosis.