Despite current medications, the morbidity and mortality associated with mental illness remain high1-3. An invaluable tool to study brain disorders and develop novel therapies is positron emission tomography (PET), a nuclear imaging technology that allows for the in vivo visualization and quantification of blood flow, metabolism, protein distribution, and drug occupancy using radioactively tagged probes (radioligands). Full quantification of radioligand binding requires invasively obtaining an arterial input function (AIF). A catheter is inserted into the radial artery of a subjects'wrist and blood sampling is done for the duration of the PET scan. If the radioligand metabolizes in the body, the fraction of un-metabolized parent compound must be determined to correct arterial concentration data accordingly - metabolite-corrected AIF (mAIF). Furthermore, the free fraction (fP) of radioligand unbound to blood proteins and vessels walls must be measured to scale the AIF (or mAIF) appropriately. These procedures are invasive, risky, time consuming, uncomfortable for subjects, and extremely costly. However, measuring the AIF, mAIF and fP is key for many radioligands (e.g. harmine4, those targeting opioid receptors5) for which a reference region devoid of the receptor of interest is not available, and for correctly interpreting PET binding, as in the case of [11C]WAY-1006356, where reports using a reference region method suggest a reduction of 5-HT1A receptors in depressed subjects compared to controls7,8, while those using AIF, mAIF and fP conclude the opposite by discovering small amount of specific binding in the reference region9,10. A promising alternative to the AIF is to use an "image-derived" input function (IDIF) that estimates the AIF directly from the PET image data. IDIF methods separate the PET signal into blood and tissue components using different approaches, such as segmentation of internal brain vasculature11-20, independent component analysis21-28, and principal components analysis29. However, these methods have not translated into clinical use because they: (1) still require some blood sampling for IDIF calibration, (2) cannot estimate the mAIF and fP, or (3) require specific acquisition protocols and cannot be implemented and/or used with previously acquired data. Characterization of these methods for usability, and addressing their limitations by combining them with state-of-the-art image processing and statistical techniques will advance the field toward quantitative PET without blood sampling. Therefore, we will (a) compare and validate existing IDIF methods to provide recommendations to the community about appropriateness of IDIF in different PET study scenarios, helping ensure widespread adoption, and (b) advance IDIF methods by combining them with statistical techniques, such as Bayesian predictive modeling, that utilizes supplementary patient data to predict the shape and scale of AIF, mAIF, and fP without blood sampling. This will allow the PET scientific and clinical community to advance towards the ultimate goal of making PET cost-effective and non-invasive to improve the study of brain disorders in terms of diagnosis, treatment prediction and healthcare cost alleviation.
We will progress towards the non-invasive full quantification of positron emission tomography (PET) data. This will allow the PET scientific and clinical community to advance towards the goal of making PET cost-effective and accurate in order to improve the study of brain disorders in terms of determining pathology, aiding in diagnostic assessment, determining brain levels of pharmaceuticals, and developing novel therapies.
|Mikhno, Arthur; Zanderigo, Francesca; Todd Ogden, R et al. (2015) Toward noninvasive quantification of brain radioligand binding by combining electronic health records and dynamic PET imaging data. IEEE J Biomed Health Inform 19:1271-82|
|Mikhno, Arthur; Angelini, Elsa D; Bai, Bing et al. (2013) LOCALLY WEIGHTED TOTAL VARIATION DENOISING FOR RINGING ARTIFACT SUPPRESSION IN PET RECONSTRUCTION USING PSF MODELING. Proc IEEE Int Symp Biomed Imaging 2013:1252-1255|
|Mikhno, Arthur; Zanderigo, Francesca; Naganawa, Mika et al. (2012) Brain tissue selection procedures for image derived input functions derived using independent components analysis. Conf Proc IEEE Eng Med Biol Soc 2012:5987-90|
|Mikhno, Arthur; Nuevo, Pablo Martinez; Devanand, Davangere P et al. (2012) MULTIMODAL CLASSIFICATION OF DEMENTIA USING FUNCTIONAL DATA, ANATOMICAL FEATURES AND 3D INVARIANT SHAPE DESCRIPTORS. Proc IEEE Int Symp Biomed Imaging :606-609|