Artificial intelligence (AI) has emerged as a transformative tool in surgical neuropathology, particularly for enhancing frozen section interpretation during brain tumor resections. This review aims to evaluate AI’s current applications and performance in intraoperative frozen section diagnosis, with emphasis on gliomas. A focused literature review was conducted using PubMed and Google Scholar, including peer-reviewed articles from 2018–2024
that reported on the use of deep learning (DL), convolutional neural networks (CNNs), and generative adversarial
networks (GANs) in neuropathologic imaging.
Frozen section slides often suffer from processing artifacts that obscure diagnostic features, yet recent AI models show promise in mitigating these limitations. CNNs have demonstrated high accuracy—up to 94% in some studies—in classifying tumor subtypes, while GANs have successfully transformed frozen images to mimic formalin-fixed paraffin-embedded (FFPE) slides. The Cryosection Histopathology Assessment and Review Machine (CHARM) has achieved near pathologist-level performance in tumor classification and prediction of key genetic alterations, such as IDH mutation and 1p/19q codeletion.
Despite these advances, challenges remain in clinical validation, interpretability, and integration into existing workflows. Nevertheless, AI holds the potential to reduce misdiagnosis, enhance intraoperative decision-making, and improve patient outcomes. Further research is needed to optimize these technologies for routine clinical application.
Frozen section slides often suffer from processing artifacts that obscure diagnostic features, yet recent AI models show promise in mitigating these limitations. CNNs have demonstrated high accuracy—up to 94% in some studies—in classifying tumor subtypes, while GANs have successfully transformed frozen images to mimic formalin-fixed paraffin-embedded (FFPE) slides. The Cryosection Histopathology Assessment and Review Machine (CHARM) has achieved near pathologist-level performance in tumor classification and prediction of key genetic alterations, such as IDH mutation and 1p/19q codeletion.
Despite these advances, challenges remain in clinical validation, interpretability, and integration into existing workflows. Nevertheless, AI holds the potential to reduce misdiagnosis, enhance intraoperative decision-making, and improve patient outcomes. Further research is needed to optimize these technologies for routine clinical application.