Skip to content

Refactor: Move inline Stan model definitions to .stan files for improved modularity #1

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
aviralsaxena16 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Refactor: Move inline Stan model definitions to .stan files for improved modularity #1

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
254 changes: 15 additions & 239 deletions src/bbb_exchange/fitting_multi_te.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,21 @@
from model_multi_te import deltaM_multite_model
import stan

import os

def load_stan_model(filename):
base_dir = os.path.dirname(os.path.abspath(__file__))
stan_path = os.path.join(base_dir, "models", "stan", filename)

if not os.path.exists(stan_path):
raise FileNotFoundError(f"Stan file not found: {stan_path}")

with open(stan_path, "r") as f:
return f.read()


STAN_MODEL_CODE = load_stan_model("multite_full.stan")

import json
with open("config.json", "r") as file:
config = json.load(file)
Expand Down Expand Up @@ -40,245 +55,6 @@ def bayesian_fit_voxel_multite(tis, tes, ntes, signal, M0a, taus,
cbf_prior_std = 20.0



STAN_MODEL_CODE = """
// multi_te_full.stan
data {
// Measurements
int<lower=1> n_measurements; // total number of TE points over all TI
int<lower=1> n_ti; // number different TI
array[n_measurements] real signal; // measured DeltaM (normalised)

// Timing
array[n_ti] real tis; // TI[j]
array[n_ti] int<lower=1> ntes; // number TE for each TI
array[n_measurements] real tes;
array[n_ti] real tau_per_ti; // labelling time

real<lower=0> t1;
real<lower=0> t1b;
real<lower=0> t2;
real<lower=0> t2b;
real<lower=0> texch;
real<lower=0> itt;
real<lower=0> lambd;
real<lower=0> alpha;
real<lower=0> M0a;

// Priors
real att_prior_mean;
real<lower=0> att_prior_std;
real cbf_prior_mean;
real<lower=0> cbf_prior_std;
}

parameters {
real<lower=0.1, upper=3.0> att; // s
real<lower=10.0, upper=250.0> cbf; // ml/min/100g
real<lower=1e-6> sigma; // noise
}

transformed parameters {
// conversion: cbf [ml/min/100g] -> f [ml/s/g]
real f = (cbf / 100.0) * 60.0 / 6000.0;
vector[n_measurements] mu;

{
int te_index = 1; // Stan-index starts at 1 (not 0)
vector[n_measurements] S_bl1_final = rep_vector(0.0, n_measurements);
vector[n_measurements] S_bl2_final = rep_vector(0.0, n_measurements);
vector[n_measurements] S_ex_final = rep_vector(0.0, n_measurements);

for (j in 1:n_ti) {
real tau = tau_per_ti[j];
real ti = tis[j];

// === Case 1: 0 < ti < att ===
if ((0 < ti) && (ti < att)) {
for (k in 1:ntes[j]) {
S_bl1_final[te_index] = 0;
S_bl2_final[te_index] = 0;
S_ex_final[te_index] = 0;
te_index += 1;
}
}

// === Case 2: att <= ti < (att + itt) ===
else if ((att <= ti) && (ti < att + itt)) {
for (k in 1:ntes[j]) {
real te = tes[te_index];
if ((0 <= te) && (te < (att + itt - ti))) {
S_bl1_final[te_index] =
(2 * f * t1b * exp(-att / t1b) * exp(-ti / t1b)
* (exp(ti / t1b) - exp(att / t1b)) * exp(-te / t2b));
}
else if (((att + itt - ti) <= te) && (te < itt)) {
real base_term = 2 * f * t1b * exp(-att / t1b) * exp(-ti / t1b)
* (exp(ti / t1b) - exp(att / t1b));
real transition_factor = (te - (att + itt - ti)) / (ti - att);
S_bl1_final[te_index] = (base_term - transition_factor * base_term) * exp(-te / t2b);
S_bl2_final[te_index] = (transition_factor * base_term) * exp(-te / t2b) * exp(-te / texch);
S_ex_final[te_index] = (transition_factor * base_term) * (1 - exp(-te / texch)) * exp(-te / t2);
}
else {
S_bl2_final[te_index] =
(2 * f * t1b * exp(-att / t1b) * exp(-ti / t1b)
* (exp(ti / t1b) - exp(att / t1b)) * exp(-te / t2b) * exp(-te / texch));
S_ex_final[te_index] =
(2 * f * t1b * exp(-att / t1b) * exp(-ti / t1b)
* (exp(ti / t1b) - exp(att / t1b)) * (1 - exp(-te / texch)) * exp(-te / t2));
}
te_index += 1;
}
}

// === Case 3: (att+itt) <= ti < (att + tau) ===
else if (((att + itt) <= ti) && (ti < (att + tau))) {
for (k in 1:ntes[j]) {
real te = tes[te_index];
real term1 = 2 * f * t1b * exp(-att / t1b) * exp(-ti / t1b)
* (exp(ti / t1b) - exp(att / t1b));
real term2 = 2 * f * t1b * exp(-(att + itt) / t1b) * exp(-ti / t1b)
* (exp(ti / t1b) - exp((att + itt) / t1b));
real base_diff = term1 - term2;
if ((0 <= te) && (te < itt)) {
real transition_factor = te / itt;
S_bl1_final[te_index] = (base_diff - transition_factor * base_diff) * exp(-te / t2b);
S_bl2_final[te_index] =
((2 * f * exp(-(1 / t1b) * (att + itt)) / ((1 / t1b) + (1 / texch))
* exp(-((1 / t1b) + (1 / texch)) * ti)
* (exp(((1 / t1b) + (1 / texch)) * ti) - exp(((1 / t1b) + (1 / texch)) * (att + itt)))
+ transition_factor * base_diff) * exp(-te / t2b) * exp(-te / texch));
S_ex_final[te_index] =
(((2 * f * exp(-(1 / t1b) * (att + itt))) / (1 / t1) * exp(-(1 / t1) * ti)
* (exp((1 / t1) * ti) - exp((1 / t1) * (att + itt)))
- (2 * f * exp(-(1 / t1b) * (att + itt))) / ((1 / texch) + (1 / t1))
* exp(-((1 / t1) + (1 / texch)) * ti)
* (exp(((1 / texch) + (1 / t1)) * ti) - exp(((1 / texch) + (1 / t1)) * (att + itt))))
* exp(-te / t2))
+ (((2 * f * exp(-(1 / t1b) * (att + itt))) / ((1 / t1b) + (1 / texch))
* exp(-((1 / t1b) + (1 / texch)) * ti)
* (exp(((1 / t1b) + (1 / texch)) * ti) - exp(((1 / t1b) + (1 / texch)) * (att + itt)))
+ base_diff) * (1 - exp(-te / texch)) * exp(-te / t2));
} else {
S_bl2_final[te_index] =
((2 * f * exp(-(1 / t1b) * (att + itt)) / ((1 / t1b) + (1 / texch))
* exp(-((1 / t1b) + (1 / texch)) * ti)
* (exp(((1 / t1b) + (1 / texch)) * ti) - exp(((1 / t1b) + (1 / texch)) * (att + itt)))
+ base_diff) * exp(-te / t2b) * exp(-te / texch));
S_ex_final[te_index] =
(((2 * f * exp(-(1 / t1b) * (att + itt))) / (1 / t1) * exp(-(1 / t1) * ti)
* (exp((1 / t1) * ti) - exp((1 / t1) * (att + itt)))
- (2 * f * exp(-(1 / t1b) * (att + itt))) / ((1 / texch) + (1 / t1))
* exp(-((1 / t1) + (1 / texch)) * ti)
* (exp(((1 / texch) + (1 / t1)) * ti) - exp(((1 / texch) + (1 / t1)) * (att + itt))))
* exp(-te / t2))
+ (((2 * f * exp(-(1 / t1b) * (att + itt))) / ((1 / t1b) + (1 / texch))
* exp(-((1 / t1b) + (1 / texch)) * ti)
* (exp(((1 / t1b) + (1 / texch)) * ti) - exp(((1 / t1b) + (1 / texch)) * (att + itt)))
+ base_diff) * (1 - exp(-te / texch)) * exp(-te / t2));
}
te_index += 1;
}
}

// === Case 4: (att + tau) <= ti < (att + itt + tau) ===
else if (((att + tau) <= ti) && (ti < (att + itt + tau))) {
for (k in 1:ntes[j]) {
real te = tes[te_index];
real term1 = 2 * f * t1b * exp(-att / t1b) * exp(-ti / t1b)
* (exp((att + tau) / t1b) - exp(att / t1b));
real term2 = 2 * f * t1b * exp(-(att + itt) / t1b) * exp(-ti / t1b)
* (exp(ti / t1b) - exp((att + itt) / t1b));
real base_diff = term1 - term2;
if ((0 <= te) && (te < (itt - (ti - (att + tau))))) {
real transition_factor = te / (itt - (ti - (att + tau)));
S_bl1_final[te_index] = (base_diff - transition_factor * base_diff) * exp(-te / t2b);
S_bl2_final[te_index] =
((2 * f * exp(-(1 / t1b) * (att + itt)) / ((1 / t1b) + (1 / texch))
* exp(-((1 / t1b) + (1 / texch)) * ti)
* (exp(((1 / t1b) + (1 / texch)) * ti) - exp(((1 / t1b) + (1 / texch)) * (att + itt)))
+ transition_factor * base_diff) * exp(-te / t2b) * exp(-te / texch));
S_ex_final[te_index] =
(((2 * f * exp(-(1 / t1b) * (att + itt))) / (1 / t1) * exp(-(1 / t1) * ti)
* (exp((1 / t1) * ti) - exp((1 / t1) * (att + itt)))
- (2 * f * exp(-(1 / t1b) * (att + itt))) / ((1 / texch) + (1 / t1))
* exp(-((1 / t1) + (1 / texch)) * ti)
* (exp(((1 / texch) + (1 / t1)) * ti) - exp(((1 / texch) + (1 / t1)) * (att + itt))))
* exp(-te / t2))
+ (((2 * f * exp(-(1 / t1b) * (att + itt))) / ((1 / t1b) + (1 / texch))
* exp(-((1 / t1b) + (1 / texch)) * ti)
* (exp(((1 / t1b) + (1 / texch)) * ti) - exp(((1 / t1b) + (1 / texch)) * (att + itt)))
+ transition_factor * base_diff) * (1 - exp(-te / texch)) * exp(-te / t2));
} else {
S_bl2_final[te_index] =
((2 * f * exp(-(1 / t1b) * (att + itt)) / ((1 / t1b) + (1 / texch))
* exp(-((1 / t1b) + (1 / texch)) * ti)
* (exp(((1 / t1b) + (1 / texch)) * ti) - exp(((1 / t1b) + (1 / texch)) * (att + itt)))
+ base_diff) * exp(-te / t2b) * exp(-te / texch));
S_ex_final[te_index] =
(((2 * f * exp(-(1 / t1b) * (att + itt))) / (1 / t1) * exp(-(1 / t1) * ti)
* (exp((1 / t1) * ti) - exp((1 / t1) * (att + itt)))
- (2 * f * exp(-(1 / t1b) * (att + itt))) / ((1 / texch) + (1 / t1))
* exp(-((1 / t1) + (1 / texch)) * ti)
* (exp(((1 / texch) + (1 / t1)) * ti) - exp(((1 / texch) + (1 / t1)) * (att + itt))))
* exp(-te / t2))
+ (((2 * f * exp(-(1 / t1b) * (att + itt))) / ((1 / t1b) + (1 / texch))
* exp(-((1 / t1b) + (1 / texch)) * ti)
* (exp(((1 / t1b) + (1 / texch)) * ti) - exp(((1 / t1b) + (1 / texch)) * (att + itt)))
+ base_diff) * (1 - exp(-te / texch)) * exp(-te / t2));
}
te_index += 1;
}
}

// === Case 5: ti >= (att + itt + tau) ===
else {
for (k in 1:ntes[j]) {
real te = tes[te_index];
S_bl2_final[te_index] =
(2 * f * exp(-(1 / t1b) * (att + itt)) / ((1 / t1b) + (1 / texch))
* exp(-((1 / t1b) + (1 / texch)) * ti)
* (exp(((1 / t1b) + (1 / texch)) * (att + itt + tau)) - exp(((1 / t1b) + (1 / texch)) * (att + itt)))
* exp(-te / t2b) * exp(-te / texch));
S_ex_final[te_index] =
(((2 * f * exp(-(1 / t1b) * (att + itt))) / (1 / t1) * exp(-(1 / t1) * ti)
* (exp((1 / t1) * (att + itt + tau)) - exp((1 / t1) * (att + itt)))
- (2 * f * exp(-(1 / t1b) * (att + itt))) / ((1 / texch) + (1 / t1))
* exp(-((1 / t1) + (1 / texch)) * ti)
* (exp(((1 / texch) + (1 / t1)) * ti) - exp(((1 / texch) + (1 / t1)) * (att + itt))))
* exp(-te / t2))
+ (((2 * f * exp(-(1 / t1b) * (att + itt))) / ((1 / t1b) + (1 / texch))
* exp(-((1 / t1b) + (1 / texch)) * ti)
* (exp(((1 / t1b) + (1 / texch)) * (att + itt + tau)) - exp(((1 / t1b) + (1 / texch)) * (att + itt))))
* (1 - exp(-te / texch)) * exp(-te / t2));
te_index += 1;
}
}
}

mu = (S_bl1_final + S_bl2_final + S_ex_final) * (M0a * alpha / lambd);
}
}

model {
// Priors
att ~ normal(att_prior_mean, att_prior_std);
cbf ~ normal(cbf_prior_mean, cbf_prior_std);
sigma ~ exponential(1.0);

// Likelihood
signal ~ normal(mu, sigma);
}

generated quantities {
vector[n_measurements] mu_ppc;
for (i in 1:n_measurements) {
mu_ppc[i] = normal_rng(mu[i], sigma);
}
}
"""

# Prepare data for Stan
data = {
"n_measurements": len(signal),
Expand Down
Loading