sunlabuiuc · ffanzhang · Apr 1, 2026 · Apr 2, 2026 · Apr 8, 2026 · Apr 8, 2026
diff --git a/docs/api/datasets.rst b/docs/api/datasets.rst
@@ -238,6 +238,9 @@ Available Datasets
     datasets/pyhealth.datasets.BMDHSDataset
     datasets/pyhealth.datasets.COVID19CXRDataset
     datasets/pyhealth.datasets.ChestXray14Dataset
+    datasets/pyhealth.datasets.ISIC2018Dataset
+    datasets/pyhealth.datasets.ISIC2018ArtifactsDataset
+    datasets/pyhealth.datasets.PH2Dataset
     datasets/pyhealth.datasets.TUABDataset
     datasets/pyhealth.datasets.TUEVDataset
     datasets/pyhealth.datasets.ClinVarDataset

diff --git a/docs/api/datasets/pyhealth.datasets.ISIC2018ArtifactsDataset.rst b/docs/api/datasets/pyhealth.datasets.ISIC2018ArtifactsDataset.rst
@@ -0,0 +1,34 @@
+pyhealth.datasets.ISIC2018ArtifactsDataset
+======================================
+
+A dataset class for dermoscopy images paired with per-image artifact
+annotations.  The default annotation file is ``isic_bias.csv`` from
+Bissoto et al. (2020), but any CSV following the same column format can be
+supplied via the ``annotations_csv`` parameter.
+
+**Default data sources (Bissoto et al. 2020)**
+
+Using ``ISIC2018ArtifactsDataset`` with the default annotation CSV requires
+**two separate downloads**:
+
+1. **Artifact annotations** (``isic_bias.csv``):
+   https://github.com/alceubissoto/debiasing-skin
+
+   See ``artefacts-annotation/`` in that repository for the annotation files.
+
+   Reference:
+   Bissoto et al. "Debiasing Skin Lesion Datasets and Models? Not So Fast",
+   ISIC Skin Image Analysis Workshop @ CVPR 2020.
+
+2. **ISIC 2018 Task 1/2 images & masks** (~8 GB):
+   https://challenge.isic-archive.com/data/#2018
+
+   * Training images: ``ISIC2018_Task1-2_Training_Input.zip``
+   * Segmentation masks: ``ISIC2018_Task1_Training_GroundTruth.zip``
+
+Both can be fetched automatically by passing ``download=True``.
+
+.. autoclass:: pyhealth.datasets.ISIC2018ArtifactsDataset
+    :members:
+    :undoc-members:
+    :show-inheritance:
diff --git a/docs/api/datasets/pyhealth.datasets.ISIC2018Dataset.rst b/docs/api/datasets/pyhealth.datasets.ISIC2018Dataset.rst
@@ -0,0 +1,21 @@
+pyhealth.datasets.ISIC2018Dataset
+===================================
+
+Unified dataset class for the ISIC 2018 challenge, supporting both
+**Task 1/2** (lesion segmentation & attribute detection) and **Task 3**
+(7-class skin lesion classification) via the ``task`` argument.
+
+For more information see `the ISIC 2018 challenge page <https://challenge.isic-archive.com/data/#2018>`_.
+
+.. note::
+   **Licenses differ by task:**
+
+   * ``task="task1_2"`` — `CC-0 (Public Domain) <https://creativecommons.org/share-your-work/public-domain/cc0/>`_.
+     No attribution required.
+   * ``task="task3"`` — `CC-BY-NC 4.0 <https://creativecommons.org/licenses/by-nc/4.0/>`_.
+     Attribution is required; commercial use is not permitted.
+
+.. autoclass:: pyhealth.datasets.ISIC2018Dataset
+    :members:
+    :undoc-members:
+    :show-inheritance:
diff --git a/docs/api/datasets/pyhealth.datasets.PH2Dataset.rst b/docs/api/datasets/pyhealth.datasets.PH2Dataset.rst
@@ -0,0 +1,47 @@
+pyhealth.datasets.PH2Dataset
+============================
+
+A dataset class for the **PH2 dermoscopy database** — 200 dermoscopic
+images of melanocytic lesions labelled as Common Nevus (80), Atypical
+Nevus (80), or Melanoma (40).
+
+Two source formats are supported automatically:
+
+* **Mirror format** (recommended for quick start): flat JPEG images and a
+  ``PH2_simple_dataset.csv`` from the community mirror
+  `vikaschouhan/PH2-dataset <https://github.com/vikaschouhan/PH2-dataset>`_.
+  Pass ``download=True`` to fetch this automatically.
+
+* **Original format**: nested BMP images and expert annotations from the
+  `official ADDI project release <https://www.fc.up.pt/addi/ph2%20database.html>`_
+  (requires registration).  Place the extracted ``PH2_Dataset_images/``
+  directory inside *root* and the dataset will use it automatically.
+
+**Data source**
+
+Teresa Mendonça, Pedro M. Ferreira, Jorge Marques, Andre R.S. Marcal,
+Jorge Rozeira. "PH² - A dermoscopic image database for research and
+benchmarking", 35th International Conference on Engineering in Medicine
+and Biology Society, EMBC'13, pp. 5437-5440, IEEE, 2013.
+
+License: free for non-commercial research purposes.  See the
+`ADDI project page <https://www.fc.up.pt/addi/ph2%20database.html>`_ for
+full terms.
+
+**Quick start**
+
+.. code-block:: python
+
+    from pyhealth.datasets import PH2Dataset
+    from pyhealth.tasks import PH2MelanomaClassification
+
+    # Download mirror automatically
+    dataset = PH2Dataset(root="~/ph2", download=True)
+
+    # Apply 3-class melanoma classification task
+    samples = dataset.set_task(PH2MelanomaClassification())
+
+.. autoclass:: pyhealth.datasets.PH2Dataset
+    :members:
+    :undoc-members:
+    :show-inheritance:
diff --git a/docs/api/tasks.rst b/docs/api/tasks.rst
@@ -225,6 +225,9 @@ Available Tasks
     Benchmark EHRShot <tasks/pyhealth.tasks.benchmark_ehrshot>
     ChestX-ray14 Binary Classification <tasks/pyhealth.tasks.ChestXray14BinaryClassification>
     ChestX-ray14 Multilabel Classification <tasks/pyhealth.tasks.ChestXray14MultilabelClassification>
+    ISIC 2018 Skin Lesion Classification <tasks/pyhealth.tasks.ISIC2018Classification>
+    ISIC 2018 Dermoscopic Artifacts Classification <tasks/pyhealth.tasks.ISIC2018ArtifactsBinaryClassification>
+    PH2 Melanoma Classification <tasks/pyhealth.tasks.PH2MelanomaClassification>
     Variant Classification (ClinVar) <tasks/pyhealth.tasks.VariantClassificationClinVar>
     Mutation Pathogenicity (COSMIC) <tasks/pyhealth.tasks.MutationPathogenicityPrediction>
     Cancer Survival Prediction (TCGA) <tasks/pyhealth.tasks.CancerSurvivalPrediction>

diff --git a/docs/api/tasks/pyhealth.tasks.ISIC2018ArtifactsBinaryClassification.rst b/docs/api/tasks/pyhealth.tasks.ISIC2018ArtifactsBinaryClassification.rst
@@ -0,0 +1,7 @@
+pyhealth.tasks.ISIC2018ArtifactsBinaryClassification
+=====================================================
+
+.. autoclass:: pyhealth.tasks.ISIC2018ArtifactsBinaryClassification
+    :members:
+    :undoc-members:
+    :show-inheritance:
diff --git a/docs/api/tasks/pyhealth.tasks.ISIC2018Classification.rst b/docs/api/tasks/pyhealth.tasks.ISIC2018Classification.rst
@@ -0,0 +1,7 @@
+pyhealth.tasks.ISIC2018Classification
+=======================================
+
+.. autoclass:: pyhealth.tasks.ISIC2018Classification
+    :members:
+    :undoc-members:
+    :show-inheritance:
diff --git a/docs/api/tasks/pyhealth.tasks.PH2MelanomaClassification.rst b/docs/api/tasks/pyhealth.tasks.PH2MelanomaClassification.rst
@@ -0,0 +1,7 @@
+pyhealth.tasks.PH2MelanomaClassification
+=========================================
+
+.. autoclass:: pyhealth.tasks.PH2MelanomaClassification
+    :members:
+    :undoc-members:
+    :show-inheritance: