TY - GEN
T1 - Combining brain imaging data with electronic health records to non-invasively quantify [11C]DASB binding
AU - Mikhno, Arthur
AU - Zanderigo, Francesca
AU - Ogden, R. Todd
AU - Mikhno, Michelle
AU - Nagendra, Harry
AU - Mann, J. John
AU - Laine, Andrew F.
AU - Parsey, Ramin V.
PY - 2014
Y1 - 2014
N2 - Quantitative analysis of PET data requires a metabolite-corrected arterial input function (AIF) for estimation of distribution volume and related outcome measures. Collecting arterial blood samples adds risk, cost, and patient discomfort to PET studies. Minimally invasive AIF estimation is possible with simultaneous estimation (SIME), but one arterial blood sample is necessary to be used as an anchor value to ensure identifiability of each individuals AIF. For [11C]DASB, a widely used serotonin transporter PET tracer, this blood sample is optimally taken 50 minutes after injection. We present here an approach for replacing such a single time-point anchor with a predicted value using brain imaging and electronic health record (EHR) data. Average bootstrap R2 > 0.8 in training data suggest that up to 80% of the variance in [11C]DASB SIME anchor may be explained by a model including heart rate, blood pressure, tracer dose, body size and cerebellar gray matter uptake. Preliminary results show that these models generalize well to a small test dataset. This may allow for quantitative analysis with no blood sampling.
AB - Quantitative analysis of PET data requires a metabolite-corrected arterial input function (AIF) for estimation of distribution volume and related outcome measures. Collecting arterial blood samples adds risk, cost, and patient discomfort to PET studies. Minimally invasive AIF estimation is possible with simultaneous estimation (SIME), but one arterial blood sample is necessary to be used as an anchor value to ensure identifiability of each individuals AIF. For [11C]DASB, a widely used serotonin transporter PET tracer, this blood sample is optimally taken 50 minutes after injection. We present here an approach for replacing such a single time-point anchor with a predicted value using brain imaging and electronic health record (EHR) data. Average bootstrap R2 > 0.8 in training data suggest that up to 80% of the variance in [11C]DASB SIME anchor may be explained by a model including heart rate, blood pressure, tracer dose, body size and cerebellar gray matter uptake. Preliminary results show that these models generalize well to a small test dataset. This may allow for quantitative analysis with no blood sampling.
KW - image-derived input function
KW - minimally/non-invasive PET
KW - simultaneous estimation
UR - https://www.scopus.com/pages/publications/84906861829
U2 - 10.1109/BHI.2014.6864468
DO - 10.1109/BHI.2014.6864468
M3 - Conference contribution
SN - 9781479921317
T3 - 2014 IEEE-EMBS International Conference on Biomedical and Health Informatics, BHI 2014
SP - 732
EP - 735
BT - 2014 IEEE-EMBS International Conference on Biomedical and Health Informatics, BHI 2014
PB - IEEE Computer Society
T2 - 2014 IEEE-EMBS International Conference on Biomedical and Health Informatics, BHI 2014
Y2 - 1 June 2014 through 4 June 2014
ER -