fix suggestions from PR

move swapping of D and D* to general wrapper
stretch exclusion criteria

Author: oliverchampion

src/standardized/TCML_TechnionIIT_lsqlm.py

@@ -78,4 +78,4 @@ def ivim_fit(self, signals, bvalues, **kwargs):
         results["f"] = fit_results[2]
         results["Dp"] = fit_results[1]
 
-        return self.D_and_Ds_swap(results)
+        return results

src/wrappers/OsipiBase.py

@@ -106,7 +106,7 @@ def osipi_fit(self, data, bvalues=None, **kwargs):
 
         for ijk in tqdm(np.ndindex(data.shape[:-1]), total=np.prod(data.shape[:-1])):
             args = [data[ijk], use_bvalues]
-            fit = self.ivim_fit(*args, **kwargs) # For single voxel fits, we assume this is a dict with a float value per key.
+            fit = self.D_and_Ds_swap(self.ivim_fit(*args, **kwargs)) # For single voxel fits, we assume this is a dict with a float value per key.
             for key in list(fit.keys()):
                 results[key][ijk] = fit[key]
 
@@ -282,7 +282,7 @@ def osipi_simple_bias_and_RMSE_test(self, SNR, bvalues, f, Dstar, D, noise_reali
             noised_signal = np.array([norm.rvs(signal, sigma) for signal in signals])
 
             # Perform fit with the noised signal
-            f_estimates[i], Dstar_estimates[i], D_estimates[i] = self.ivim_fit(noised_signal, bvalues)
+            f_estimates[i], Dstar_estimates[i], D_estimates[i] = self.D_and_Ds_swap(self.ivim_fit(noised_signal, bvalues))
 
         # Calculate bias
         f_bias = np.mean(f_estimates) - f

tests/IVIMmodels/unit_tests/test_ivim_fit.py

@@ -63,7 +63,7 @@ def to_list_if_needed(value):
         "atol": tolerances["atol"]
     }
     record_property('test_data', test_result)
-    if (data['f'] == 1 or data['f'] == 0) and not fit.use_bounds: #in these cases there are multiple solutions (D can become D*, f will be 0 and D* can be anything. This will be a good description of the data, so technically not a fail
+    if (data['f'] > 0.99 or data['f'] < 0.01) and not fit.use_bounds: #in these cases there are multiple solutions (D can become D*, f will be 0 and D* can be anything. This will be a good description of the data, so technically not a fail
         return
     npt.assert_allclose(fit_result['f'],data['f'], rtol=tolerances["rtol"]["f"], atol=tolerances["atol"]["f"])
     if data['f']<0.80: # we need some signal for D to be detected