changes made as requested

Author: oliverchampion

src/standardized/ETP_SRI_LinearFitting.py

@@ -57,6 +57,7 @@ def ivim_fit(self, signals, bvalues=None, linear_fit_option=False, **kwargs):
         Returns:
             _type_: _description_
         """
+        signals[signals<0.0000001]=0.0000001
         if bvalues is None:
             bvalues = self.bvalues
 

src/standardized/OJ_GU_seg.py

@@ -61,7 +61,7 @@ def ivim_fit(self, signals, bvalues=None):
             bthr = 200
         else:
             bthr = self.thresholds[0]
-                    
+        signals[signals<0.00001]=0.00001
         fit_results = seg(signals, bvalues, bthr)
 
         f = fit_results['f']

src/standardized/PvH_KB_NKI_IVIMfit.py

@@ -52,6 +52,7 @@ def ivim_fit(self, signals, bvalues=None):
         #bvalues = np.array(bvalues)
         bvalues = bvalues.tolist() #NKI code expects a list instead of nparray
         # reshape signal as the NKI code expects a 4D array
+        signals[signals<0.00001]=0.00001
         signals = np.reshape(signals, (1, 1, 1, len(signals)))  # assuming that in this test the signals are always single voxel
         fit_results = self.NKI_algorithm(signals,bvalues)
 

tests/IVIMmodels/unit_tests/algorithms.json

@@ -62,15 +62,13 @@
             "dynamic_rtol": {
                 "offset": 0.001,
                 "noise": 200,
-                "scale_f": false,
                 "f": 2,
                 "D": 1,
                 "Dp": 3
             },
             "dynamic_atol": {
                 "offset": 5e-2,
                 "noise": 10,
-                "scale_f": false,
                 "f": 1,
                 "D": 0.001,
                 "Dp": 2

tests/IVIMmodels/unit_tests/test_ivim_fit.py

@@ -114,29 +114,22 @@
 
 def signal_helper(signal):
     signal = np.asarray(signal)
-    signal = np.abs(signal) #Oliver: do we need this for unit testing? It will bias the fits by forcing positive signals.
+    #signal[signal < 0] = 0.00001
     signal /= signal[0] # this scales noise differenly. I have now made sure signal is produced with amplitude 1 if right flags are selected. Then  SNR is fixed. Then, this sentence may be redundant
-    #ratio = 1 / signal[0] #this is 1 per definition (See sentence above)
     return signal
 
 def tolerances_helper(tolerances, data):
-    epsilon = 0.0001
+    epsilon = 0.001
     if "dynamic_rtol" in tolerances:
         dyn_rtol = tolerances["dynamic_rtol"]
         scale = dyn_rtol["offset"] + dyn_rtol["noise"]*data["noise"]
-        if dyn_rtol["scale_f"]:
-            tolerances["rtol"] = {"f": scale*dyn_rtol["f"], "D": scale*dyn_rtol["D"]/(1-data['f']+epsilon), "Dp": scale*dyn_rtol["Dp"]/(data['f']+epsilon)}
-        else:
-            tolerances["rtol"] = {"f": scale * dyn_rtol["f"], "D": scale * dyn_rtol["D"]/0.9,"Dp": scale * dyn_rtol["Dp"]/0.1}
+        tolerances["rtol"] = {"f": scale*dyn_rtol["f"], "D": scale*dyn_rtol["D"]/(1-data['f']+epsilon), "Dp": scale*dyn_rtol["Dp"]/(data['f']+epsilon)}
     else:
         tolerances["rtol"] = tolerances.get("rtol", {"f": 0.1, "D": 0.1, "Dp": 0.3})
     if "dynamic_atol" in tolerances:
         dyn_atol = tolerances["dynamic_atol"]
         scale = dyn_atol["offset"] + dyn_atol["noise"]*data["noise"]
-        if dyn_atol["scale_f"]:
-            tolerances["atol"] = {"f": scale*dyn_atol["f"], "D": scale*dyn_atol["D"]/(1-data['f']+epsilon), "Dp": scale*dyn_atol["Dp"]/(data['f']+epsilon)}
-        else:
-            tolerances["atol"] = {"f": scale*dyn_atol["f"], "D": scale*dyn_atol["D"]/0.9, "Dp": scale*dyn_atol["Dp"]/0.1}
+        tolerances["atol"] = {"f": scale*dyn_atol["f"], "D": scale*dyn_atol["D"]/(1-data['f']+epsilon), "Dp": scale*dyn_atol["Dp"]/(data['f']+epsilon)}
     else:
         tolerances["atol"] = tolerances.get("atol", {"f": 2e-1, "D": 5e-4, "Dp": 4e-2})
     return tolerances