diff --git a/scilpy/segment/streamlines.py b/scilpy/segment/streamlines.py
index 788455ea2..1eb81856e 100644
--- a/scilpy/segment/streamlines.py
+++ b/scilpy/segment/streamlines.py
@@ -76,8 +76,10 @@ def filter_grid_roi_both(sft, mask_1, mask_2):
     # For endpoint filtering, we need to keep 2 separately
     # Could be faster for either end, but the code look cleaner like this
     line_based_indices = []
-    voxel_beg = np.asarray([s[0] for s in streamline_vox]).astype(np.int16).transpose(1, 0)
-    voxel_end = np.asarray([s[-1] for s in streamline_vox]).astype(np.int16).transpose(1, 0)
+    voxel_beg = np.asarray([s[0] for s in streamline_vox],
+                           dtype=np.int16).transpose(1, 0)
+    voxel_end = np.asarray([s[-1] for s in streamline_vox],
+                           dtype=np.int16).transpose(1, 0)
 
     map1_beg = map_coordinates(mask_1, voxel_beg, order=0, mode='nearest')
     map2_beg = map_coordinates(mask_2, voxel_beg, order=0, mode='nearest')
@@ -154,7 +156,7 @@ def filter_grid_roi(sft, mask, filter_type, is_exclude):
     if is_exclude:
         line_based_indices = np.setdiff1d(range(len(sft)),
                                           np.unique(line_based_indices))
-    line_based_indices = np.asarray(line_based_indices).astype(np.int32)
+    line_based_indices = np.asarray(line_based_indices, dtype=np.int32)
 
     # From indices to sft
     streamlines = sft.streamlines[line_based_indices]
@@ -196,7 +198,7 @@ def pre_filtering_for_geometrical_shape(sft, size,
 
     # Create relevant info about the ellipsoid in vox/world space
     if is_in_vox:
-        center = np.asarray(apply_affine(transfo, center))
+        center = np.asarray(apply_affine(transfo, center), dtype=float)
     bottom_corner = center - size
     top_corner = center + size
     x_val = [bottom_corner[0], top_corner[0]]
@@ -253,7 +255,8 @@ def filter_ellipsoid(sft, ellipsoid_radius, ellipsoid_center,
 
     if is_in_vox:
         ellipsoid_center = np.asarray(apply_affine(transfo,
-                                                   ellipsoid_center))
+                                                   ellipsoid_center),
+                                      dtype=float)
     selected_by_ellipsoid = []
     line_based_indices_1 = []
     line_based_indices_2 = []
@@ -262,8 +265,8 @@ def filter_ellipsoid(sft, ellipsoid_radius, ellipsoid_center,
     # The result won't be identical to MI-Brain since I am not using the
     # vtkPolydata. Also it won't be identical to TrackVis either,
     # because TrackVis is point-based for Spherical ROI...
-    ellipsoid_radius = np.asarray(ellipsoid_radius)
-    ellipsoid_center = np.asarray(ellipsoid_center)
+    ellipsoid_radius = np.asarray(ellipsoid_radius, dtype=float)
+    ellipsoid_center = np.asarray(ellipsoid_center, dtype=float)
 
     for i, line in enumerate(pre_filtered_streamlines):
         if filter_type in ['any', 'all']:
@@ -307,7 +310,7 @@ def filter_ellipsoid(sft, ellipsoid_radius, ellipsoid_center,
     if is_exclude:
         selected_by_ellipsoid = np.setdiff1d(range(len(sft)),
                                              np.unique(selected_by_ellipsoid))
-    line_based_indices = np.asarray(selected_by_ellipsoid).astype(np.int32)
+    line_based_indices = np.asarray(selected_by_ellipsoid, dtype=np.int32)
 
     # From indices to sft
     streamlines = sft.streamlines[line_based_indices]
@@ -359,8 +362,8 @@ def filter_cuboid(sft, cuboid_radius, cuboid_center,
     line_based_indices_2 = []
     # Also here I am not using a mathematical intersection and
     # I am not using vtkPolyData like in MI-Brain, so not exactly the same
-    cuboid_radius = np.asarray(cuboid_radius)
-    cuboid_center = np.asarray(cuboid_center)
+    cuboid_radius = np.asarray(cuboid_radius, dtype=float)
+    cuboid_center = np.asarray(cuboid_center, dtype=float)
     for i, line in enumerate(pre_filtered_streamlines):
         if filter_type in ['any', 'all']:
             # Resample to 1/10 of the voxel size
@@ -404,7 +407,7 @@ def filter_cuboid(sft, cuboid_radius, cuboid_center,
     if is_exclude:
         selected_by_cuboid = np.setdiff1d(range(len(sft)),
                                           np.unique(selected_by_cuboid))
-    line_based_indices = np.asarray(selected_by_cuboid).astype(np.int32)
+    line_based_indices = np.asarray(selected_by_cuboid, dtype=np.int32)
 
     # From indices to sft
     streamlines = sft.streamlines[line_based_indices]
diff --git a/scilpy/tracking/tools.py b/scilpy/tracking/tools.py
index 287436360..d21e19532 100644
--- a/scilpy/tracking/tools.py
+++ b/scilpy/tracking/tools.py
@@ -41,7 +41,8 @@ def filter_streamlines_by_length(sft, min_length=0., max_length=np.inf):
     else:
         filter_stream = []
 
-    filtered_streamlines = list(np.asarray(sft.streamlines)[filter_stream])
+    filtered_streamlines = list(np.asarray(sft.streamlines,
+                                           dtype=object)[filter_stream])
     filtered_data_per_point = sft.data_per_point[filter_stream]
     filtered_data_per_streamline = sft.data_per_streamline[filter_stream]
 
@@ -80,7 +81,7 @@ def get_subset_streamlines(sft, max_streamlines, rng_seed=None):
     ind = np.arange(len(sft.streamlines))
     rng.shuffle(ind)
 
-    subset_streamlines = list(np.asarray(sft.streamlines)[
+    subset_streamlines = list(np.asarray(sft.streamlines, dtype=object)[
                               ind[:max_streamlines]])
     subset_data_per_point = sft.data_per_point[ind[:max_streamlines]]
     subset_data_per_streamline = sft.data_per_streamline[ind[:max_streamlines]]
@@ -205,7 +206,7 @@ def smooth_line_gaussian(streamline, sigma):
     x3 = gaussian_filter1d(x, sigma)
     y3 = gaussian_filter1d(y, sigma)
     z3 = gaussian_filter1d(z, sigma)
-    smoothed_streamline = np.asarray([x3, y3, z3]).T
+    smoothed_streamline = np.asarray([x3, y3, z3], dtype=float).T
 
     # Ensure first and last point remain the same
     smoothed_streamline[0] = streamline[0]
diff --git a/scilpy/utils/bvec_bval_tools.py b/scilpy/utils/bvec_bval_tools.py
index e29a5ae8c..252354853 100644
--- a/scilpy/utils/bvec_bval_tools.py
+++ b/scilpy/utils/bvec_bval_tools.py
@@ -250,7 +250,7 @@ def identify_shells(bvals, threshold=40.0, roundCentroids=False, sort=False):
     # Finding centroids
     bval_centroids = [bvals[0]]
     for bval in bvals[1:]:
-        diffs = np.abs(np.asarray(bval_centroids) - bval)
+        diffs = np.abs(np.asarray(bval_centroids, dtype=float) - bval)
         if not len(np.where(diffs < threshold)[0]):
             # Found no bval in bval centroids close enough to the current one.
             # Create new centroid (i.e. new shell)
diff --git a/scilpy/utils/metrics_tools.py b/scilpy/utils/metrics_tools.py
index aa9b754bf..c4a097cbb 100644
--- a/scilpy/utils/metrics_tools.py
+++ b/scilpy/utils/metrics_tools.py
@@ -134,7 +134,7 @@ def _get_profile_one_streamline(streamline, metrics_files):
     # Each tuple has S elements, where S is the number of streamlines.
     # We then convert each tuple to a numpy array
     for metric_values in zip(*metrics_per_strl):
-        converted.append(np.asarray(metric_values))
+        converted.append(np.asarray(metric_values, dtype=float))
 
     return converted