Cerebral blood flow measurements with ¹⁵O-water PET using a non-invasive machine-learning-derived arterial input function

Abstract

Cerebral blood flow (CBF) can be measured with dynamic positron emission tomography (PET) of 15O-labeled water by using tracer kinetic modelling. However, for quantification of regional CBF, an arterial input function (AIF), obtained from arterial blood sampling, is required. In this work we evaluated a novel, non-invasive approach for input function prediction based on machine learning (MLIF), against AIF for CBF PET measurements in human subjects. Twenty-five subjects underwent two 10 min dynamic 15O-water brain PET scans with continuous arterial blood sampling, before (baseline) and following acetazolamide medication. Three different image-derived time-activity curves were automatically segmented from the carotid arteries and used as input into a Gaussian process-based AIF prediction model, considering both baseline and acetazolamide scans as training data. The MLIF approach was evaluated by comparing AIF and MLIF curves, as well as whole-brain grey matter CBF values estimated by kinetic modelling derived with either AIF or MLIF. The results showed that AIF and MLIF curves were similar and that corresponding CBF values were highly correlated and successfully differentiated before and after acetazolamide medication. In conclusion, our non-invasive MLIF method shows potential to replace the AIF obtained from blood sampling for CBF measurements using 15O-water PET and kinetic modelling.

Prediction of in‐hospital hypokalemia using machine learning and first hospitalization day records in patients with traumatic brain injury

“…Four subsets were used as training data and the remaining one as the validation set. The cross‐validation process was repeated five times, and each of the five folds was used once as validation data 27,28 . The results were then averaged to produce a single estimation.…”

“…The cross‐validation process was repeated five times, and each of the five folds was used once as validation data. 27 , 28 The results were then averaged to produce a single estimation. To overcome overfitting due to group imbalance when predicting hypokalemia of different severity, a propensity score matching approach was used to balance the gender and age of patients in a 1:1 ratio.…”

Mathematical Models for FDG Kinetics in Cancer: A Review

“…An example of how this perspective may be helpful in compartmental analysis is illustrated in Reference [ 49 ], where an optimization scheme inspired by ant colony behavior is utilized to determine the kinetic parameters. However, most optimization algorithms belonging to this group of methods rely on neural networks that are formulated within the framework of machine and deep learning theory [ 64 , 65 , 66 ].…”

Noninvasive Quantification of Cerebral Blood Flow Using Hybrid PET/MR Imaging to Extract the [15O]H2O Image‐Derived Input Function Free of Partial Volume Errors

“…However, quantifying CBF with [ 15 O]H 2 O PET requires arterial blood sampling to measure the arterial input function (AIF). Efforts to avoid this invasive procedure have primarily focused on methods of obtaining an image‐derived input function (IDIF) 4–12 . In general, the accuracy of these methods requires careful correction of partial volume effects (PVEs; i.e., correcting for spill‐in and spill‐out activity), which are typically performed by measuring the point‐spread‐function of the PET scanner and/or using calibration factors, such as obtained by acquiring a few blood samples 13 .…”

“…Efforts to avoid this invasive procedure have primarily focused on methods of obtaining an image-derived input function (IDIF). [4][5][6][7][8][9][10][11][12] In general, the accuracy of these methods requires careful correction of partial volume effects (PVEs; i.e., correcting for spill-in and spill-out activity), which are typically performed by measuring the point-spread-function of the PET scanner and/or using calibration factors, such as obtained by acquiring a few blood samples. 13 MRI angiography enables more robust vessel segmentation for calculating correction factors [4][5][6] ; however, it is challenging to obtain accurate estimates due to the relatively small size of internal carotid and vertebral arteries and misalignments between PET and MR images.…”