Histologic Distinction Between Clear Cell Sarcoma and Melanoma Using Supervised and Deep Learning.

Area of Science:

• Computational pathology
• Digital pathology
• Machine learning in oncology

Background:

• Clear cell sarcoma and melanoma share histological and immunophenotypic similarities, complicating diagnosis.
• Molecular testing for EWSR1 rearrangement is crucial but not always accessible for distinguishing these tumors.

Purpose of the Study:

• To develop and validate machine learning-based classifiers for differentiating clear cell sarcoma from melanoma.
• To explore the utility of nuclear morphometrics and deep learning in diagnostic pathology.

Main Methods:

• Digitized hematoxylin-and-eosin-stained slides from clear cell sarcomas and melanomas.
• Constructed nuclear morphometric and deep learning classifiers (CLAM/ResNet-50, CTransPath, UNI).
• Evaluated classifier performance on an independent external validation set.

Main Results:

• Two nuclear morphometric classifiers achieved 80%-90% accuracy in external validation.
• The optimal deep learning classifier (CLAM/CTransPath) reached 90% accuracy.
• Performance was comparable to that of experienced pathologists.

Conclusions:

• Interpretable nuclear morphometric and deep learning classifiers can reliably distinguish clear cell sarcoma from melanoma.
• Quantitative morphometric analysis with machine learning serves as a valuable diagnostic adjunct.
• These AI tools can improve diagnostic accuracy, especially when molecular testing is limited.