diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index b108119cf..725ab6562 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -10,7 +10,7 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python-version: [3.9, '3.10', '3.11']
+        python-version: [3.8, 3.9, '3.10', '3.11']
 
     steps:
       - uses: actions/checkout@v3
diff --git a/environment.yml b/environment.yml
index 98f49b9eb..fc602fa48 100644
--- a/environment.yml
+++ b/environment.yml
@@ -11,17 +11,18 @@ dependencies:
   - click>=8.1
   - numpy~=1.23.0
   - pillow~=9.2.0
-  - scipy~=1.11.0
+  - scipy~=1.10.0
   - jinja2~=3.0
   - conda-forge::torchvision-cpu>=0.5.0
   - conda-forge::pytorch-cpu>=1.11,<1.14
   - jsonschema
   - scikit-learn~=1.2.1
-  - scikit-image~=0.21.0
+  - scikit-image~=0.22.0
   - shapely~=1.8.5
   - pyvips
   - imagemagick>=7.1.0
   - pyarrow
+  - importlib-resources>=1.3.0
   - conda-forge::pytorch-lightning~=2.0.0
   - conda-forge::torchmetrics>=1.1.0
   - conda-forge::threadpoolctl~=3.2
diff --git a/environment_cuda.yml b/environment_cuda.yml
index f492c0443..ca671c7de 100644
--- a/environment_cuda.yml
+++ b/environment_cuda.yml
@@ -11,18 +11,19 @@ dependencies:
   - click>=8.1
   - numpy~=1.23
   - pillow>=9.2.0
-  - scipy~=1.11.0
+  - scipy~=1.10.0
   - jinja2~=3.0
   - conda-forge::torchvision
   - conda-forge::pytorch>=1.11,<1.14
   - cudatoolkit>=9.2
   - jsonschema
   - scikit-learn~=1.2.1
-  - scikit-image~=0.21.0
+  - scikit-image~=0.22.0
   - shapely~=1.8.5
   - pyvips
   - imagemagick>=7.1.0
   - pyarrow
+  - importlib-resources>=1.3.0
   - conda-forge::pytorch-lightning~=2.0.0
   - conda-forge::torchmetrics>=1.1.0
   - conda-forge::threadpoolctl~=3.2
diff --git a/kraken/blla.py b/kraken/blla.py
index cd9486de9..9c9054af4 100644
--- a/kraken/blla.py
+++ b/kraken/blla.py
@@ -25,7 +25,7 @@
 import torch
 import logging
 import numpy as np
-import importlib.resources
+import importlib_resources
 import shapely.geometry as geom
 import torch.nn.functional as F
 import torchvision.transforms as tf
@@ -310,7 +310,7 @@ def segment(im: PIL.Image.Image,
     """
     if model is None:
         logger.info('No segmentation model given. Loading default model.')
-        model = vgsl.TorchVGSLModel.load_model(importlib.resources.files(__name__).joinpath('blla.mlmodel'))
+        model = vgsl.TorchVGSLModel.load_model(importlib_resources.files(__name__).joinpath('blla.mlmodel'))
 
     if isinstance(model, vgsl.TorchVGSLModel):
         model = [model]
diff --git a/kraken/ketos/repo.py b/kraken/ketos/repo.py
index 17b4ee8cc..d977d947d 100644
--- a/kraken/ketos/repo.py
+++ b/kraken/ketos/repo.py
@@ -43,7 +43,7 @@ def publish(ctx, metadata, access_token, private, model):
     Publishes a model on the zenodo model repository.
     """
     import json
-    import importlib.resources
+    import importlib_resources
 
     from jsonschema import validate
     from jsonschema.exceptions import ValidationError
@@ -52,7 +52,7 @@ def publish(ctx, metadata, access_token, private, model):
     from kraken.lib import models
     from kraken.lib.progress import KrakenDownloadProgressBar
 
-    ref = importlib.resources.files('kraken').joinpath('metadata.schema.json')
+    ref = importlib_resources.files('kraken').joinpath('metadata.schema.json')
     with open(ref, 'rb') as fp:
         schema = json.load(fp)
 
diff --git a/kraken/kraken.py b/kraken/kraken.py
index d269cc7c8..8e31d29ad 100644
--- a/kraken/kraken.py
+++ b/kraken/kraken.py
@@ -23,7 +23,7 @@
 import warnings
 import logging
 import dataclasses
-import importlib.resources
+import importlib_resources
 
 from PIL import Image
 from pathlib import Path
@@ -45,7 +45,7 @@
 install(suppress=[click])
 
 APP_NAME = 'kraken'
-SEGMENTATION_DEFAULT_MODEL = importlib.resources.files(APP_NAME).joinpath('blla.mlmodel')
+SEGMENTATION_DEFAULT_MODEL = importlib_resources.files(APP_NAME).joinpath('blla.mlmodel')
 DEFAULT_MODEL = ['en_best.mlmodel']
 
 # raise default max image size to 20k * 20k pixels
diff --git a/kraken/lib/dataset/utils.py b/kraken/lib/dataset/utils.py
index 7e5d808ff..b690fdabf 100644
--- a/kraken/lib/dataset/utils.py
+++ b/kraken/lib/dataset/utils.py
@@ -18,7 +18,7 @@
 import json
 import torch
 import numbers
-import importlib.resources
+import importlib_resources
 import torch.nn.functional as F
 
 from functools import partial
@@ -320,7 +320,7 @@ def compute_confusions(algn1: Sequence[str], algn2: Sequence[str]):
         script substitutions.
     """
     counts: Dict[Tuple[str, str], int] = Counter()
-    ref = importlib.resources.files(__name__).joinpath('scripts.json')
+    ref = importlib_resources.files(__name__).joinpath('scripts.json')
     with ref.open('rb') as fp:
         script_map = json.load(fp)
 
diff --git a/setup.cfg b/setup.cfg
index de5ed343a..3a802ada3 100644
--- a/setup.cfg
+++ b/setup.cfg
@@ -14,6 +14,7 @@ classifier =
         Intended Audience :: Science/Research
 	License :: OSI Approved :: Apache Software License
         Operating System :: POSIX
+	Programming Language :: Python :: 3.8
 	Programming Language :: Python :: 3.9
 	Programming Language :: Python :: 3.10
 	Programming Language :: Python :: 3.11
@@ -37,7 +38,7 @@ max_line_length = 160
 exclude = tests/*
 
 [options]
-python_requires = >=3.9,<=3.11.99
+python_requires = >=3.8,<=3.11.99
 install_requires =
     jsonschema
     lxml
@@ -46,7 +47,7 @@ install_requires =
     numpy~=1.23.0
     Pillow>=9.2.0
     regex
-    scipy~=1.11.0
+    scipy~=1.10.0 # bump up to latest release with py3.8 EOL
     protobuf>=3.0.0
     coremltools~=6.0
     jinja2~=3.0
@@ -54,12 +55,13 @@ install_requires =
     torchvision>=0.5.0
     torch~=2.0.1
     scikit-learn~=1.2.1
-    scikit-image~=0.21.0
+    scikit-image~=0.22.0
     shapely~=1.8.5
     pyarrow
     pytorch-lightning~=2.0.0
     torchmetrics>=1.1.0
     threadpoolctl~=3.2.0
+    importlib-resources>=1.3.0
     rich
 
 [options.extras_require]