Breakthrough in Brain Cancer Diagnosis: AI-Powered Classification and Segmentation of MRI Images

Strategic Insights -

A new research study, conducted by a team at Yonsei University, presents a significant advancement in the diagnosis of brain cancer through the development of a Convolutional Neural Network (CNN) model dubbed MediAI. This AI-powered method leverages transfer learning with ResNet50 to classify MRI images of primary brain tumors, achieving an impressive accuracy of 97.6%. The study highlights the potential of MediAI in enhancing diagnosis and treatment assessment by precisely detecting and delineating tumor regions, particularly in glioblastomas.

Key Takeaways:

  • MediAI, a CNN model leveraging transfer learning with ResNet50, was developed for the classification and segmentation of MRI images of primary brain tumors, particularly gliomas.
  • The dataset comprised 414 MRI images of primary brain tumors (86 adenomas, 84 epithelial tumors, 82 gliomas, 80 meningiomas, and 82 schwannomas) and 39 normal MR images.
  • Experimental results demonstrated that MediAI achieved a classification accuracy of 97.6% for brain tumor classification, outperforming previous methods with the highest reported accuracy.
  • Tumor regions were refined through morphological operations, and the final tumor contours were extracted and overlaid on the original images.
  • Results indicated that gliomas were classified with a precision of 91%, recall of 99%, and F1-score of 95%, demonstrating the robustness of MediAI.
  • The proposed methodology enhances the monitoring of treatment responses by tracking changes in segmented tumor regions.
  • Future work will focus on utilizing Radiomics to map tumor, necrotic, and edema regions for advancing diagnostic and therapeutic paradigms in glioma treatment.
  • Hwunjae Lee, a researcher from Yonsei University, is involved in this study, emphasizing the institution's role in the development of AI tools for medical applications.

Statistics:

  • Accuracy of MediAI model for brain tumor classification: 97.6%
  • Precision of MediAI for glioma classification: 91%
  • Recall of MediAI for glioma classification: 99%
  • F1-score of MediAI for glioma classification: 95%
  • Number of MRI images used in the dataset: 453 (414 primary brain tumors and 39 normal MR images)
  • Types of brain tumors included in the dataset: adenomas, epithelial tumors, gliomas, meningiomas, and schwannomas

Sources:

  • A Study On Classification and Segmentation of Brain Tumor Mri Using Mediai. Journal of Magnetics, 2025;30(1). Korean Magnetics Soc, Korea Sciences & Technol Ctr, Rm 905, Yeoksam-Dong 635-4, Kangnam-Ku, Seoul, 135-703, South Korea.
  • NewsRx. Reports from Yonsei University Add New Data to Findings in Brain Cancer (A Study On Classification and Segmentation of Brain Tumor Mri Using Mediai). Health & Medicine Week. June 13, 2025; p 4706.