Data Harmonization with StyleTransfer-GANs: Enhancing Non-Invasive IDH Classification in Brain Tumors

Isocitrate dehydrogenase (IDH) mutation status has emerged as an important prognostic marker in brain gliomas. Accurate non-invasive determination of IDH mutation status is crucial for effective therapy and prognosis. However, the variability in imaging protocols across institutions hinders the reliability of deep learning (DL) models used for IDH classification. To address data heterogeneity, a StyleTransfer-GAN ( ST-net ) was developed to harmonize multi-site MRI data. ST-net was applied to harmonize MR images from diverse databases while preserving critical imaging features. A high performing deep-learning network ( MC-net ) was used to assess the impact of style transfer on IDH classification. The results demonstrated that ST-net effectively aligned style features across datasets and minimized discrepancies caused by varied imaging protocols. The IDH classification accuracy improved post-harmonization, with enhancements in both sensitivity and specificity depending on the style transfer reference. Our study highlights the potential of data harmonization using GANs to improve the generalizability and clinical utility of DL models in neuro-oncology. It also provides a scalable solution adaptable to various neuroimaging tasks.