[MAINT] update parameters for scipy.sparse.linalg.cg (#4394)

* changed cg parameters

* fix compat support scipy < 1.12 [test nightly]

nilearn/_utils/segmentation.py

@@ -11,10 +11,12 @@
 import warnings
 
 import numpy as np
-from scipy import ndimage as ndi, sparse
+from scipy import __version__, ndimage as ndi, sparse
 from scipy.sparse.linalg import cg
 from sklearn.utils import as_float_array
 
+from nilearn._utils.helpers import compare_version
+
 
 def _make_graph_edges_3d(n_x, n_y, n_z):
     """Return a list of edges for a 3D image.
@@ -355,7 +357,14 @@ def _solve_cg(lap_sparse, B, tol):
     lap_sparse = lap_sparse.tocsc()
     X = []
     for i in range(len(B)):
-        x0 = cg(lap_sparse, -B[i].todense(), tol=tol, atol="legacy")[0]
+        # TODO Python 3.8
+        # consider remove if/else block when dropping support for 3.8
+        # would require pinning scipy to >= 1.12
+        # See https://github.com/nilearn/nilearn/pull/4394
+        if compare_version(__version__, ">=", "1.12"):
+            x0 = cg(lap_sparse, -B[i].todense(), rtol=tol, atol=0)[0]
+        else:
+            x0 = cg(lap_sparse, -B[i].todense(), tol=tol, atol="legacy")[0]
         X.append(x0)
 
     X = np.array(X)