This will be difficult to come up with a reproducible example for as there is no open source data as yet, and I'm not sure I'm allowed to share the data I have. I will try my best to explain it, and if this doesn't work, I can maybe take some time to simulate some data at a later point. Hopefully it's an easy solution though...
Background
I am busy creating an R package for kinetic modelling in the field that I work in (https://github.com/mathesong/kinfitr). I am trying as best as I can to make everything amenable to tidyverse tooling. However, there is a particular use case for which I can't figure out how to do it as it involves pulling data from several different formats in rather different structures, and pulling them together in the model.
In the README on the page, I present a solution for Reference Region models, where all inputs are of the same length and I can work with the following workflow:
data %>%
gather() %>%
group_by() %>%
do()
The Issue
However, for arterial models, the input arguments are as follows:
Brain kinetic data: times, values, weights - each vectors of the same length, in this case 38
Blood kinetic data: bloodinput - data frame of 4096 rows x 4 columns. For the sake of convenience, all models read this in as a data frame with all the information already interpolated.
Each of the models requires inputs of all three vectors, as well as the bloodinput data frame.
I currently have all the data stored in a list, with an element for each measurement. Each element of the list contains 1. a data frame with the brain kinetic data (each region of the brain, let's say 3 regions), as well as times and weights, and 2. a data frame containing the bloodinput data. Thus I create my final data frame
datdf <- map(dat, 'braindf') %>% # Extract the brain data
bind_rows(.id = "id") %>% # Add an id column
select(PET = id, Times = Times, Weights=weights, R1 = Region1, R2 = Region2, R3 = Region3) %>% # Rename and select columns
group_by(PET) %>% # Group by each measurement
nest() %>% # Nest everything
rename(braindata=data) %>% # Rename
mutate(Subjname = stringr::str_extract(....)), # Add subject acronym
PETNo = as.numeric(stringr::str_extract(....)), # Add measurement number
input=map(dat, 'bloodinput')) # Add blood input data frame as a nested column
This leaves me with the following
# A tibble: 6 × 5
PET braindata Subjname PETNo bloodinput
<chr> <list> <chr> <dbl> <list>
1 s1_1 <tibble [38 × 6]> s1 1 <data.frame [4,096 × 4]>
2 s1_2 <tibble [38 × 6]> s1 2 <data.frame [4,096 × 4]>
3 s2_1 <tibble [38 × 6]> s2 1 <data.frame [4,096 × 4]>
4 s2_2 <tibble [38 × 6]> s2 2 <data.frame [4,096 × 4]>
5 s1_1 <tibble [38 × 6]> s3 1 <data.frame [4,096 × 4]>
6 s2_2 <tibble [38 × 6]> s3 2 <data.frame [4,096 × 4]>
where each brain data contains the following:
head(datdf[1,]$braindata[[1]])
# A tibble: 6 × 6
Times Weights R1 R2 R3
<dbl> <dbl> <dbl> <dbl> <dbl>
1 0 0 0.00E+00 0.00E+00 0.00E+00
2 22 0.3 1.12E-03 4.14E-03 4.78E-04
3 32 0.5 5.61E-01 4.08E-01 7.38E-01
4 42 0.7 4.53E+01 4.50E+01 5.61E+01
5 52 0.7 8.12E+01 8.07E+01 1.02E+02
6 62 0.9 1.03E+02 1.04E+02 1.31E+02
From this point, I cannot figure out how to fit the model for each row.
This is what I have tried:
R1_outcomes <- datdf %>%
group_by(PET) %>% # or rowwise()
mutate(onetcmout = onetcm(t_tac=.$braindata[[1]]$Times/60,
tac=.$braindata[[1]]$R1,
input=.$bloodinput,
weights=.$braindata[[1]]$Weights))
R1_outcomes <- datdf %>%
rowwise() %>%
do(onetcmout = onetcm(t_tac=.$braindata[[1]]$Times/60,
tac=.$braindata[[1]]$R1,
input=.$bloodinput,
weights=.$braindata[[1]]$Weights))
I'm sure there's a way of doing this with the map functions, but I can't quite figure out how.
I would really appreciate any advice on how I might be able to do this. Thank you to anyone in advance!
