<![CDATA[Brain cancer has a high mortality rate due to delayed diagnosis, making accurate early detection systems an urgent necessity. This study proposes a two-stage transfer learning approach (initial training and fine-tuning) using VGG16 as a feature extractor, combined with three classification architectures—CNN, FNN, and LSTM—for brain cancer detection in MRI images. The novelty of this study lies in the systematic comparison of the three architectures within a transfer learning framework on a small-scale MRI dataset (818 images with an 80:20 ratio) enhanced through data augmentation. The VGG16+LSTM model achieved the highest accuracy (96.38 percent), followed by VGG16+FNN (96.21 percent) and VGG16+CNN (94.74 percent). The best-performing model was integrated into a web application as a clinical decision support system for early screening. These results confirm the effectiveness of the two-stage transfer learning approach in overcoming data limitations while improving MRI-based classification performance.]]>
Copyrights © 2026
