SIME: Simultaneous Estimation of Non-Displaceable Binding (SIME)

In mathesong/kinfitr: Kinetic Modelling of PET Time Activity Curves

Simultaneous Estimation of Non-Displaceable Binding (SIME)

Description

Function to fit the SIME Model of Ogden et al (2015) to data to estimate Vnd for a set of TACs.

Usage

SIME(
  t_tac,
  tacdf,
  input,
  Vndgrid,
  weights = NULL,
  roiweights = NULL,
  inpshift = 0,
  vB = NULL,
  twotcmstart = NULL,
  frameStartEnd = NULL,
  k2.start = 0.1,
  k2.lower = 0,
  k2.upper = 0.5,
  k3.start = 0.1,
  k3.lower = 0,
  k3.upper = 0.5,
  k4.start = 0.1,
  k4.lower = 0,
  k4.upper = 0.5,
  success_cutoff = 0.5
)

Arguments

t_tac	Numeric vector of times for each frame in minutes. We use the time halfway through the frame as well as a zero. If a time zero frame is not included, it will be added.
tacdf	Named dataframe of TACs in wide format, i.e. each TAC should be a column.
input	Data frame containing the blood, plasma, and parent fraction concentrations over time. This can be generated using the blood_interp function.
Vndgrid	The grid of Vnd values which will be tested to see which one has the best fit.
weights	Optional. Numeric vector of the weights assigned to each frame in the fitting. We include zero at time zero: if not included, it is added. If not specified, uniform weights will be used.
roiweights	Optional. Numeric vector of the weights assigned to each ROI in the fitting. If not specified, uniform weights will be used.
inpshift	Optional. The number of minutes by which to shift the timing of the input data frame forwards or backwards. If not specified, this will be set to 0. This can be fitted using 1TCM or 2TCM.
vB	Optional. The blood volume fraction. If not specified, this will be set to 0.05. This can be fitted using 1TCM or 2TCM.
twotcmstart	Optional. The function can fit a 2TCM model to one of the ROIs and use the estimated k2, k3 and k4 as starting parameters for the rest of the fits. If left alone, these parameters will be specified as below. If one wishes to run the 2TCM to start off, use a numeric value to specify which column of tacdf to use for fitting this: best to use the largest ROI.
frameStartEnd	Optional: This allows one to specify the beginning and final frame to use for modelling, e.g. c(1,20). This is to assess time stability.
k2.start	Optional. Starting parameter for fitting of k2. Default is 0.1.
k2.lower	Optional. Lower bound for the fitting of k2. Default is 0.
k2.upper	Optional. Upper bound for the fitting of k2. Default is 0.5.
k3.start	Optional. Starting parameter for fitting of k3. Default is 0.1.
k3.lower	Optional. Lower bound for the fitting of k3. Default is 0.
k3.upper	Optional. Upper bound for the fitting of k3. Default is 0.5.
k4.start	Optional. Starting parameter for fitting of k4. Default is 0.1.
k4.lower	Optional. Lower bound for the fitting of k4. Default is 0.
k4.upper	Optional. Upper bound for the fitting of k4. Default is 0.5.
success_cutoff	Optional. Should values of Vnd for which a certain proportion of ROIs failed be included? Default is 0.5, i.e. 50 successfully fitted.

Value

A list with a data frame of the fitted parameter out$par, the dataframe containing the times and TACs out$tacs, the mean cost values after fitting (after ROI weighting) out$fitvals, the ROI cost values after fitting (before ROI weighting) out$roifits, the blood input data frame after time shifting input, a vector of the weights out$weights, a vector of the ROI weights out$roiweights, the inpshift value used inpshift and the vB value used out$vB, and the success cutoff success_cutoff.

Author(s)

Granville J Matheson, [e-mail]

References

Ogden RT, Zanderigo F, Parsey RV. Estimation of in vivo nonspecific binding in positron emission tomography studies without requiring a reference region. NeuroImage. 2015 Mar 31;108:234-42.

Examples

## Not run: 
data(pbr28)

t_tac <- pbr28$tacs[[2]]$Times / 60
tacdf <- dplyr::select(pbr28$tacs[[2]], FC:CBL)
weights <- pbr28$tacs[[2]]$Weights

input <- blood_interp(
  pbr28$procblood[[2]]$Time / 60, pbr28$procblood[[2]]$Cbl_dispcorr,
  pbr28$procblood[[2]]$Time / 60, pbr28$procblood[[2]]$Cpl_metabcorr,
  t_parentfrac = 1, parentfrac = 1
)

Vndgrid <- seq(from = 0, to = 3, by = 0.5)
SIMEout <- SIME(t_tac, tacdf, input, Vndgrid,
  weights = weights,
  inpshift = 0.1, vB = 0.05
)

## End(Not run)

mathesong/kinfitr documentation built on Jan. 15, 2024, 11:07 p.m.