From 4f3ab87bcc63e72f07733fcd4a7795a2951545bc Mon Sep 17 00:00:00 2001
From: Joeran Bosma <joeran@bosma.co>
Date: Thu, 3 Oct 2024 17:13:00 +0200
Subject: [PATCH] Account for new keyword in sklearn v1.5+ (#19)

---
 setup.py                  |  2 +-
 src/picai_eval/metrics.py | 20 +++++++++++++++-----
 2 files changed, 16 insertions(+), 6 deletions(-)

diff --git a/setup.py b/setup.py
index 04a3bdb..c9b347f 100644
--- a/setup.py
+++ b/setup.py
@@ -63,7 +63,7 @@ def run(self):
         long_description = fh.read()
 
     setuptools.setup(
-        version='1.4.6',  # also update version in metrics.py -> version
+        version='1.4.7',  # also update version in metrics.py -> version
         author_email='Joeran.Bosma@radboudumc.nl',
         long_description=long_description,
         long_description_content_type="text/markdown",
diff --git a/src/picai_eval/metrics.py b/src/picai_eval/metrics.py
index 2d99e00..0ec9bbe 100644
--- a/src/picai_eval/metrics.py
+++ b/src/picai_eval/metrics.py
@@ -19,6 +19,8 @@
 from typing import Any, Dict, Hashable, List, Optional, Tuple, Union
 
 import numpy as np
+import sklearn
+from packaging import version
 from sklearn.metrics import auc, precision_recall_curve, roc_curve
 
 try:
@@ -265,11 +267,19 @@ def calculate_precision_recall(self, subject_list: Optional[List[str]] = None) -
         y_pred: "npt.NDArray[np.float64]" = np.array([pred for _, pred, *_ in lesion_y_list])
 
         # calculate precision-recall curve
-        precision, recall, thresholds = precision_recall_curve(
-            y_true=y_true,
-            probas_pred=y_pred,
-            sample_weight=self.get_lesion_weight_flat(subject_list=subject_list)
-        )
+        if version.parse(sklearn.__version__) >= version.parse("1.5"):
+            # in the future this if/else block can be removed, then set 1.5 as minimum in requirements.txt
+            precision, recall, thresholds = precision_recall_curve(
+                y_true=y_true,
+                y_score=y_pred,
+                sample_weight=self.get_lesion_weight_flat(subject_list=subject_list)
+            )
+        else:
+            precision, recall, thresholds = precision_recall_curve(
+                y_true=y_true,
+                probas_pred=y_pred,
+                sample_weight=self.get_lesion_weight_flat(subject_list=subject_list)
+            )
 
         # set precision to zero at a threshold of "zero", as those lesion
         # candidates are included just to convey the number of lesions to