1.6.0

What's Changed

• Update citation for 1.5.1 by @KumoLiu in #8582
• Bump actions/download-artifact from 4 to 5 by @dependabot[bot] in #8590
• Bump actions/checkout from 4 to 5 by @dependabot[bot] in #8588
• Bump actions/setup-python from 5 to 6 by @dependabot[bot] in #8589
• 8564 fourier positional encoding by @NabJa in #8570
• Include more-itertools in build env by @KumoLiu in #8611
• Bump peter-evans/create-or-update-comment from 4 to 5 by @dependabot[bot] in #8612
• Bump actions/download-artifact from 5 to 6 by @dependabot[bot] in #8614
• Bump actions/upload-artifact from 4 to 5 by @dependabot[bot] in #8615
• Bump github/codeql-action from 3 to 4 by @dependabot[bot] in #8616
• added ReduceTrait and FlattenSequence by @lukas-folle-snkeos in #8531
• Fix box_iou returning 0 for floating-point results less than 1. #8369 by @reworld223 in #8553
• 8620 ModuleNotFoundError: No module named 'onnxscript' in test-py3x (3.11) pipeline by @garciadias in #8621
• timestep scheduling with np.linspace by @ytl0623 in #8623
• Correct H&E stain ordering heuristic in ExtractHEStains by @iyassou in #8551
• feat: add activation checkpointing to unet by @ferreirafabio80 in #8554
• Fix index using tuple for image cropping operation by @johnzielke in #8633
• Fix #8599: Add track_meta and weights_only arguments to PersistentDataset for MetaTensor support. by @mccle in #8628
• Update documentation links by @KumoLiu in #8637
• Fix #8350: Clarify LocalNormalizedCrossCorrelationLoss docstring by @engmohamedsalah in #8639
• 8620 modulenotfounderror no module named onnxscript in test py3x 311 pipeline by @garciadias in #8638
• Generate heatmap transforms by @eclipse0922 in #8579
• Added an optional_import check for onnxruntime and applied the @unitt… by @ytl0623 in #8641
• Bump peter-evans/slash-command-dispatch from 4.0.0 to 5.0.0 by @dependabot[bot] in #8650
• Bump al-cheb/configure-pagefile-action from 1.4 to 1.5 by @dependabot[bot] in #8648
• Bump actions/checkout from 4 to 6 by @dependabot[bot] in #8649
• resolves CI's drive-out-of-space by prune caching and use torch fro CPU by @Borda in #8673
• Consolidating Version Bumps by @ericspod in #8681
• Fix Zip Slip vulnerability in NGC private bundle download by @yueyueL in #8682
• Fix channel-first indices buffer for distance_transform_edt (return_indices=True) #8656 by @alexanderjaus in #8657
• Fix missing closing parenthesis in SSIM formula documentation by @Marshall-mk in #8661
• Replace pyupgrade with builtin Ruff's UP rule by @Borda in #8606
• Rename NormalizeLabelsInDatasetd to RemapLabelsToSequentiald and fix label ordering bug by @aymuos15 in #8680
• Adding What's New Page for 1.5.2 Release by @ericspod in #8720
• Update base image to 25.12 by @KumoLiu in #8738
• Fixes typo by @benediktjohannes in #8709
• Fix dead code and logic error by @ytl0623 in #8714
• Fix flake8 format in PatchEmbeddingBlock by @KumoLiu in #8742
• Fix network_type docstring formatting in PerceptualLoss by @aman0311x in #8670
• Update pre-commit hooks to latest versions by @Borda in #8693
• Fixes several typos by @benediktjohannes in #8712
• "Transfroms" -> "Transforms" (typo fix in test_nvtx_transform.py) by @benediktjohannes in #8711
• Fix docstring typos across codebase by @aymuos15 in #8688
• docs: add differentiability warning for Warp with mode='nearest' by @Kheil-Z in #8629
• Complete replace flake8 with ruff and update rules by @Borda in #8694
• Fix AutoencoderKlMaisi forcing CUDA transfer on CPU inputs by @ytl0623 in #8736
• Modernize build commands for setuptools 80+ compatibility by @engmohamedsalah in #8728
• feat(metrics): Add MAPEMetric for regression evaluation. by @akshatsinha0 in #8686
• Fix type hint for image reader by @Kirscher in #8646
• Weights in alpha for FocalLoss by @ytl0623 in #8665
• Replace yesqa hook from pre-commit configuration with Ruff's RUF100 by @Borda in #8692
• fix(metrics): Corrected dimension count in MS-SSIM error message for 3D images. by @akshatsinha0 in #8685
• ENH: support additional dtypes in pad_nd by @shubham-61969 in #8672
• Update the docstrings to match the actual code defaults by @ytl0623 in #8701
• Fix #8366: Add strict shape validation to sliding_window_inference by @mattlin1124 in #8645
• fix: support decollate for numpy scalars by @arthurdjn in #8470
• Update ['pixdim'] after Spacing transform in meta dict. by @slicepaste in #8269
• Generalize TestTimeAugmentation to non-spatial predictions by @ytl0623 in #8715
• docs: clarify backward mapping logic for RandAffine and RandAffined by @ytl0623 in #8730
• Add 3D support and confusion matrix output to PanopticQualityMetric by @aymuos15 in #8684
• Change tuple() to () for slightly improved performance in trt_compiler.py by @benediktjohannes in #8749
• Fix incorrect docstring defaults for DiffusionModelUNet by @eitag-uni in #8571
• fix(losses): add missing docstring entries for default parameters by @yushiran in #8755
• Fix batch size broadcasting bug in GeneralizedWassersteinDiceLoss by @hongjie-qiu in #8744
• Add CalibrationErrorMetric and CalibrationError handler by @theo-barfoot in #8707
• Fix GEGLU docstring: Sigmoid -> GELU by @aymuos15 in #8696
• Fix the computation of the crop region end index by @AndersMollgaard in #7825
• Support MIG UUID and respect CUDA_VISIBLE_DEVICES in nnUNetV2Runner by @ytl0623 in #8716
• Adjust execution order of activation and masking in MaskedDiceLoss by @ytl0623 in #8704
• Fft cleanup/update by @aymuos15 in #8762
• Fix using a non-tuple sequence for multidimensional indexing that wil… by @aarpon in #8732
• Docker Slim Image by @ericspod in #8640
• Remove unused n_dims parameter from Fourier.inv_shift_fourier by @aymuos15 in #8766
• Add package in requirement file resolve the Module Not Found Error by @ytl0623 in #8792
• Deploy Script Update by @ericspod in #8745
• Replace deprecated cuda.cudart with cuda.bindings.runtime. by @ytl0623 in #8790
• Fix Invertd by @eclipse0922 in #8651
• Fixing SSIM Test Method Names by @ericspod in #8746
• Add missing channel_wise parameter to RandScaleIntensityFixedMean by @engmohamedsalah in #8741
• Add GradientAccumulation utility for SupervisedTrainer by @aymuos15 in #8763
• Fix #8340: Remove incorrect minus sign in rotate_range documentation by @yunhaoli24 in #8773
• Prevent implicit conversion of metatensor to numpy array by @sudomakeinstall in #8654
• Enhance ExtractDataKeyFromMetaKeyd to work with MetaTensor by @haoyu-haoyu in #8772
• Add MCC loss by @kakumarabhishek in #8785
• Fixes #8697 GPU memory leak by checking both image and label tensors for CUDA device by @benediktjohannes in #8708
• 8800 potentially wrong device using cuda variable status in monaiauto3dseganalyzerpy by @garciadias in #8801
• Follow-up for Fixing SSIM Test Method Names by @benediktjohannes in #8750
• Use [] and {} instead of list() and dict() by @benediktjohannes in #8748
• Slightly improve performance in engines/utils.py by @benediktjohannes in #8747
• Use {} for better performance in dict() by @benediktjohannes in #8751
• Adding pybind11 explicitly by @ericspod in #8814
• Fix compute_shape_offset non-tuple indexing for PyTorch >=2.9 by @Harrisaint in #8776
• Fix #8775: handle torch.Tensor input in compute_shape_offset by @williams145 in #8812
• Fix make_gaussian_kernel truncated parameter unit mismatch (#8780) by @aymuos15 in #8781
• Fix incorrect truncated parameter in make_gaussian_kernel causing corrupted LocalNormalizedCrossCorrelationLoss by @ytl0623 in #8783
• fix(autoencoderkl): handle proj_attn→out_proj key mapping in load_old_state_dict by @ytl0623 in #8786
• Add spatial shape constraint docs and test for SwinUNETR (#6771) by @Cado87 in #8817
• Replace direct np.random.* calls with np.random.RandomState instances by @haoyu-haoyu in #8798
• Excluded from MANIFEST.in and setup.py tests folder and subfolders. by @ffvogmv in #8807
• Add support for configurable GD-enhancing tumor label in ConvertToMul… by @AbdoMis in #8779
• Fix memory leak in optional_import traceback handling by @aymuos15 in #8782
• Fix TrainableBilateralFilter 3D input validation (#7444) by @getrichthroughcode in #8729
• Fix: swap input_amplitude and target_amplitude in JukeboxLoss.forward by @Zeesejo in #8821
• Docs: add spatial shape constraints to SegResNetDS docstring by @Talhax55z in #8797
• Fix: handle precomputed crops and safe no-grad cleanup in auto3dseg by @bluehyena in #8803
• Refactor download_url test into TestDownloadUrl class by @e-mny in #8829
• Add EmbeddingCollapseMetric: detect representational collapse in medi… by @ekansh-arora0 in #8815
• Fix multi-axis shear transform to compose individual shear matrices by @aymuos15 in #8778
• 7951 anchors are not centered on grid cells by @hachoj in #8475
• fix(apps/nuclick): replace np.random.choice with self.R.choice in Add… by @Zeesejo in #8836
• fix: use register_buffer for kernel and kernel_vol in LocalNormalizedCrossCorrelationLoss by @Zeesejo in #8818
• Add parameter to DiceMetric and DiceHelper classes by @VijayVignesh1 in #8774
• Enable global coordinates in spatial crop transforms by @emanuilo in #8794
• Avoid eager C-order copy in NibabelReader (fixes #8107) by @aymuos15 in #8825
• Refactoring min-dep tests by @ericspod in #8561
• Fix incomplete activation validation in HausdorffDTLoss by @aymuos15 in #8841
• Fix #8442: pass meta_data to EnsureChannelFirst when track_meta is False by @williams145 in #8835
• Fix align_corners mismatch in AffineTransform by @aymuos15 in #8690
• Fix nested Compose map_items in forward and inverse paths by @aymuos15 in #8787
• fix: replace set_track_meta with F.interpolate in RandSimulateLowResolution (#8409) by @chhayankjain in #8837
• Feature added: Get WSI at mpp by @NikolasSchmitz in #7574
• Replace pickle with JSON in Auto3DSeg algo serialization by @aymuos15 in #8695
• docs: add missing docstrings, Raises sections, and fix Args in networ… by @chhayankjain in #8850
• docs: fix docstring parameter name, add missing Args and Raises secti… by @chhayankjain in #8849
• 4609: Add AUC-Margin Loss for AUROC optimization by @shubham-61969 in #8719
• Replace BaseException with Exception across codebase by @AlexanderSanin in #8859
• Add nested dot-notation access to ConfigParser (#6837) by @aymuos15 in #8858
• Fix: make y optional in ignore_background by @Kirscher in #8647
• Fix #8239: Enhance SoftclDiceLoss and SoftDiceclDiceLoss with DiceLoss-compatible API by @aymuos15 in #8703
• 8627 perceptual loss errors out after hitting the maximum number of downloads by @virginiafdez in #8652
• 8587 test erros on pytorch release 2508 on series 50 by @garciadias in #8770
• Further CICD Cleanup by @ericspod in #8867
• fix: only instantiate CrossAttentionBlock when with_cross_attention=True by @chhayankjain in #8848
• Make get_norm_layer repr test tolerant of PyTorch bias= field by @hjmjohnson in #8880
• fix(networks): replace Tensor | None with Optional[Tensor] for TorchScript compatibility by @AlexanderSanin in #8879
• Register GlobalMutualInformationLoss bin_centers as buffer by @ugbotueferhire in #8869
• Fix For GHSA-rghg-q7wp-9767 by @ericspod in #8885
• Adding allow_pickle argument by @ericspod in #8875
• CodeQL and GPU Actions Update by @ericspod in #8883
• Minor Fixes From Testing 1.6 by @ericspod in #8897
• Release 1.6 Doc Updates by @ericspod in #8882

New Contributors

• @lukas-folle-snkeos made their first contribution in #8531
• @reworld223 made their first contribution in #8553
• @iyassou made their first contribution in #8551
• @ferreirafabio80 made their first contribution in #8554
• @mccle made their first contribution in #8628
• @engmohamedsalah made their first contribution in #8639
• @eclipse0922 made their first contribution in #8579
• @yueyueL made their first contribution in #8682
• @alexanderjaus made their first contribution in #8657
• @Marshall-mk made their first contribution in #8661
• @aymuos15 made their first contribution in #8680
• @benediktjohannes made their first contribution in #8709
• @aman0311x made their first contribution in #8670
• @Kheil-Z made their first contribution in #8629
• @akshatsinha0 made their first contribution in #8686
• @Kirscher made their first contribution in #8646
• @shubham-61969 made their first contribution in #8672
• @mattlin1124 made their first contribution in #8645
• @arthurdjn made their first contribution in #8470
• @eitag-uni made their first contribution in #8571
• @yushiran made their first contribution in #8755
• @hongjie-qiu made their first contribution in #8744
• @theo-barfoot made their first contribution in #8707
• @AndersMollgaard made their first contribution in #7825
• @aarpon made their first contribution in #8732
• @yunhaoli24 made their first contribution in #8773
• @sudomakeinstall made their first contribution in #8654
• @haoyu-haoyu made their first contribution in #8772
• @kakumarabhishek made their first contribution in #8785
• @Harrisaint made their first contribution in #8776
• @williams145 made their first contribution in #8812
• @Cado87 made their first contribution in #8817
• @ffvogmv made their first contribution in #8807
• @AbdoMis made their first contribution in #8779
• @getrichthroughcode made their first contribution in #8729
• @Zeesejo made their first contribution in #8821
• @Talhax55z made their first contribution in #8797
• @bluehyena made their first contribution in #8803
• @e-mny made their first contribution in #8829
• @ekansh-arora0 made their first contribution in #8815
• @hachoj made their first contribution in #8475
• @VijayVignesh1 made their first contribution in #8774
• @emanuilo made their first contribution in #8794
• @chhayankjain made their first contribution in #8837
• @NikolasSchmitz made their first contribution in #7574
• @AlexanderSanin made their first contribution in #8859
• @ugbotueferhire made their first contribution in #8869

Full Changelog: 1.5.2...1.6.0