NeuroDEEP-CNN: A Multi-Task Attention Framework for Joint Classification, Localization, and Segmentation of Brain Tumors on MRI

Öz

Brain tumors require rapid and accurate differential diagnosis across many radiological subtypes. Existing classifiers rarely combine classification, lesion localization, and segmentation in a single interpretable system. NeuroDEEP-CNN is a multi-task ConvNeXtTiny backbone with three decoder heads for classification, coordinate regression, and mask segmentation, fused through a hybrid coordinate-mask attention gate. The model was trained on 12,643 T1, T1C+, and T2 MRI images across 39 classes using a two-phase focal warm-up and progressive fine-tuning strategy with an 80/10/10 stratified split. At the best checkpoint (epoch 33, selected on validation accuracy), the model achieved 92.81% validation classification accuracy and 98.26% validation top-3 accuracy. The detection branch reached a validation mean distance error of 18.56 pixels, and the segmentation branch reached a validation Dice coefficient of 0.664. A five-method explainability suite confirmed predictions are grounded in pathological MRI regions. NeuroDEEP-CNN closely approaches the strongest of five widely used transfer-learning baselines on a related three-class benchmark, while covering thirteen times the number of classes and simultaneously providing validated tumor detection and segmentation outputs suitable for radiologist-facing deployment.

Anahtar Kelimeler

• artificial intelligence
• brain disease
• tumor

Beyin Tumorlerinin MR Goruntulerinde Birlikte Siniflandirma, Lokalizasyon ve Segmentasyonu icin Cok Gorevli Dikkat Mekanizmasi Cercevesi: NeuroDEEP-CNN

Öz

Beyin tümörleri, çok sayıda radyolojik alt tip arasında hızlı ve doğru ayırıcı tanı gerektirmektedir. Mevcut sınıflandırıcılar, sınıflandırma, lezyon lokalizasyonu ve segmentasyonu tek bir yorumlanabilir sistem içinde nadiren birleştirmektedir. NeuroDEEP-CNN, sınıflandırma, koordinat regresyonu ve maske segmentasyonu için üç ayrı kod çözücü başlığa sahip, çok görevli ConvNeXtTiny omurgasına dayanan bir modeldir ve hibrit koordinat-maske dikkat kapısı aracılığıyla birleştirilmiştir. Model, 39 sınıfa ait 12.643 T1, T1C+ ve T2 MRG görüntüsü üzerinde; iki aşamalı odak kaybı ısınma stratejisi ve kademeli ince ayar yaklaşımı kullanılarak, %80/%10/%10 oranında tabakalı veri bölünmesiyle eğitilmiştir. En iyi kontrol noktasında, doğrulama doğruluğuna göre seçilen 33. epokta, model %92,81 doğrulama sınıflandırma doğruluğu ve %98,26 doğrulama ilk-3 doğruluğu elde etmiştir. Tespit dalı, doğrulama kümesinde 18,56 piksel ortalama mesafe hatasına ulaşırken, segmentasyon dalı 0,664 doğrulama Dice katsayısına ulaşmıştır. Beş yöntemli açıklanabilirlik paketi, model tahminlerinin patolojik MRG bölgelerine dayandığını doğrulamıştır. NeuroDEEP-CNN, ilişkili üç sınıflı bir kıyaslama veri setinde yaygın olarak kullanılan beş transfer öğrenme temel modelinin en güçlüsüne yakın performans göstermekte; aynı zamanda sınıf sayısını on üç kat artırmakta ve radyolog odaklı klinik kullanıma uygun şekilde doğrulanmış tümör tespiti ve segmentasyon çıktıları sunmaktadır.

Anahtar Kelimeler

• artificial intelligence
• grad cam
• brain tumor

Kaynakça

1. [1] Z. O. Jagun, O. J. Adetunji, M. B. Olajide, and O. Onafuye, “Deep Learning Based Brain Tumor Diagnosis,” Sakarya University Journal of Computer and Information Sciences, vol. 9, no. 2, pp. 306-324, 2026, doi:10.35377/saucis.1628374.
2. [2] F. Özger, I. Pacal, and D. Sökmen, “Cutting Edge Deep Learning Models for Brain Tumor Classification,” Gazi University Journal of Science, vol. 39, no. 1, pp. 394-413, 2026, doi:10.35378/gujs.1684696.
3. [3] Ç. Saraç, S. Arıkan Arıbal, and Y. A. Üncü, “Hybrid Deep Learning and Reinforcement Learning Approach for Brain Tumor Classification from MRI Images,” Süleyman Demirel University Faculty of Arts and Science Journal of Science, vol. 20, no. 2, 2025.
4. [4] M. Coşkun, “Enhancing Reliability in Deep Learning Diagnosis of Brain Tumors Using Grad-CAM,” Turkish Journal of Nature and Science, vol. 15, no. 2, pp. 39-48, 2026, doi:10.46810/tdfd.1749282.
5. [5] B. Gencer, “A Comparative Analysis of EfficientNetB0 and EfficientNetV2 Variants for Brain Tumor Classification Using MRI Images,” International Journal of Innovative Engineering Applications, vol. 9, no. 1, 2025.
6. [6] Ö. Akboyraz and S. Kızıloluk, “Improving CNN-based brain tumor classification via differential learning rate and data augmentation strategies,” International Journal of Pure and Applied Sciences, vol. 12, no. 1, 2026.
7. [7] B. Ulu, “Comparison of deep learning methods in brain tumor diagnosis: High-performance classification with MRI data,” Communications Faculty of Sciences University of Ankara Series A2-A3, vol. 67, no. 1, pp. 59-73, 2025, doi:10.33769/aupse.1619837.
8. [8] A. Sevinç, B. Kaya, and M. Gül, “Multiple classification of brain tumor images using a new and efficient convolutional neural network-based model,” Dicle University Journal of Engineering, vol. 16, no. 2, 2025.

9. [9] E. Aslan and Y. Özüpak, “Performance Comparison of Deep Learning Models in Brain Tumor Classification,” Balkan Journal of Electrical and Computer Engineering, vol. 13, no. 2, 2025.
10. [10] S. Bayaral, E. Gül, and D. Avcı, “Classification of Brain Tumors Using Artificial Intelligence,” International Journal of Innovative Engineering Applications, vol. 9, no. 1, pp. 8-22, 2025, doi:10.46460/ijiea.1563426.
11. [11] H. Aslım and A. Çalışkan, “Improving Brain Tumor Detection with Deep Learning Models: A Comparative Analysis Using MRI and RO-SVM Classification,” Bitlis Eren University Journal of Science, vol. 14, no. 2, 2025.
12. [12] A. Kalender et al., “New Results in Brain Tumor Classification Using Hybrid Transfer Learning Techniques,” Gazi University Journal of Science Part C: Design and Technology, vol. 13, no. 1, 2025.
13. [13] M. Özer, “Brain Tumor Detection using Deep CNNs and Ensemble Algorithms over MRI Images,” Bilgisayar Bilimleri, vol. 9, no. 2, pp. 142-150, 2024, doi:10.53070/bbd.1455902.
14. [14] F. D. Eryılmaz Baran and M. Çetin, “Brain Tumor Classification Using Convolutional Neural Networks,” Researcher, vol. 4, no. 2, 2024.
15. [15] Ş. Öztürk and O. Katar, “A Deep Learning Model Collaborates with an Expert Radiologist to Classify Brain Tumors from MR Images,” Turkish Journal of Science and Technology, vol. 17, no. 2, 2022.
16. [16] A. Doğanay and A. Yıldız, “Deep Learning Based Computer Aided Diagnostic System for Brain Tumor Detection,” Düzce University Journal of Science and Technology, vol. 10, no. 4, 2022.
17. [17] B. Şener and B. Ergen, “Enhancing Brain Tumor Detection on MRI Images Using an Innovative VGG-19 Model-Based Approach,” Sakarya University Journal of Science, vol. 27, no. 5, 2023, doi:10.16984/saufenbilder.1302803.
18. [18] T. Çay, “Deep learning-based brain tumor segmentation: A comparison of U-Net and SegNet algorithms,” Bozok Journal of Engineering and Architecture, vol. 3, no. 2, pp. 99-109, 2024, doi:10.70700/bjea.1581404.
19. [19] M. U. Ateş, R. T. Günlü, E. Ekinci, and Z. Garip, “Using 3D U-Net for Brain Tumour Segmentation from Magnetic Resonance Images,” Academic Platform Journal of Engineering and Smart Systems, vol. 12, no. 3, pp. 81-87, 2024, doi:10.21541/apjess.1508913.
20. [20] E. Güvenç, M. Ersoy, and G. Çetin, “Brain Tumor Segmentation on FLAIR MR Images with U-Net,” Mugla Journal of Science and Technology, vol. 9, no. 1, pp. 34-41, 2023, doi:10.22531/muglajsci.1244322.
21. [21] A. Bayram, S. Gürkan, A. Budak, and M. Karataş, “CNNs for MRI-Based Brain Tumor Segmentation: Comparative Analysis,” Turkish Journal of Forecasting, vol. 6, no. 2, pp. 61-66, 2022, doi:10.34110/forecasting.1190289.
22. [22] S. H. H. Abdalgadir and M. Öztürk, “Image Enhancement and U-Net Based Brain Tumor Segmentation Using MRI,” European Journal of Engineering and Applied Sciences, vol. 8, no. 1, pp. 31-38, 2025, doi:10.55581/ejeas.1703274.
23. [23] E. Cengil, Y. Eroğlu, A. Çınar, and M. Yıldırım, “Detection and Localization of Glioma and Meningioma Tumors in Brain MR Images using Deep Learning,” Sakarya University Journal of Science, vol. 27, no. 3, pp. 550-563, 2023, doi:10.16984/saufenbilder.1067061.
24. [24] G. Tüzün and M. Özdemir, “Classification of Brain Tumors with Deep Learning Models,” Journal of Scientific Reports-A, no. 54, pp. 296-306, 2023, doi:10.59313/jsr-a.1293119.
25. [25] H. Demir, A. Arı, and F. Demir, “Detection of brain tumor with a pre-trained deep learning model based on feature selection using MR images,” Fırat University Journal of Engineering and Computational Engineering, vol. 2, no. 1, pp. 23-31, 2023, doi:10.5505/fujece.2023.36844.
26. [26] M. Yıldırım et al., “Automated Grading of Glioma Using Deep Neural Networks,” NATURENGS, vol. 4, no. 2, pp. 15-22, 2023, doi:10.46572/naturengs.1397010.
27. [27] A. Alzahrani and A. Qahtani, “Knowledge distillation in transformers with tripartite attention: Multiclass brain tumor detection in highly augmented MRIs,” Journal of King Saud University - Computer and Information Sciences, vol. 36, no. 1, 101907, 2024, doi:10.1016/j.jksuci.2023.101907.
28. [28] M. Rutoh et al., “GAIR-U-Net: 3D guided attention inception residual U-Net for brain tumor segmentation using multimodal MRI images,” Journal of King Saud University - Computer and Information Sciences, vol. 36, no. 6, 102086, 2024, doi:10.1016/j.jksuci.2024.102086.
29. [29] E. Çetiner and S. Metlek, “DenseUNet+: A novel hybrid segmentation approach based on multi-modality images for brain tumor segmentation,” Journal of King Saud University - Computer and Information Sciences, vol. 35, no. 8, 101663, 2023, doi:10.1016/j.jksuci.2023.101663.
30. [30] R. Almnawer, Z. Alhaj, and M. Ozturk, “CrossTaskFormer: Bidirectional cross-task attention for joint brain tumor segmentation and classification,” Journal of King Saud University - Computer and Information Sciences, 2026, doi:10.1007/s44443-026-00973-3.
31. [31] Z. Liu et al., “A ConvNet for the 2020s,” in Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 11976-11986, 2022, doi:10.1109/CVPR52688.2022.01167.
32. [32] R. R. Selvaraju et al., “Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization,” International Journal of Computer Vision, vol. 128, pp. 336-359, 2020, doi:10.1007/s11263-019-01228-7.
33. [33] F. Isensee, P. F. Jaeger, S. A. A. Kohl, J. Petersen, and K. H. Maier-Hein, “nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation,” Nature Methods, vol. 18, pp. 203-211, 2021, doi:10.1038/s41592-020-01008-z.
34. [34] F. Rad, “Brain Tumor 12K MRI Images with Masks, Metadata and Bounding Boxes,” Kaggle dataset, accessed Aug. 7, 2026.