Skip to content
DraconiticResearch
Search Sources API
203,952 studies31 sourcesUpdated 16 min ago
PreprintCC BY 4.0Hybrid CNN–Transformer–LSTM Architectures for Explainable Brain Tumor MRI Classification

Hybrid CNN–Transformer–LSTM Architectures for Explainable Brain Tumor MRI Classification

  • Viraj Wankhade1
  • Vrushali Paratkar1
  • Bhuvan Bhiogade1
  • Tushar Baghele1
  • Prof. Dr. U. P. Akare Akare1
  1. 1Dept. of Artificial Intelligence and Data Science Priyadarshini College of Engineering Nagpur, India
Open at sourceDOICite

Abstract

Automated classification of brain tumors from magnetic resonance imaging (MRI) can support radiologists by narrowing down differential diagnoses, but clinical usefulness depends on a model being both accurate and interpretable. In this work we design and empirically compare hybrid deep-learning architectures that combine a convolutional neural network (CNN), a Swin-style Transformer sequence refiner, and a bidirectional long short-term memory (BiLSTM) network for four-class brain tumor classification: Glioma, Meningioma, No Tumor, and Pituitary. A full hybrid model (ResNet-50 → Transformer refiner → BiLSTM) trained for fifteen epochs reaches 94.42% test accuracy, a macro F1-score of 94.44%, and a macro one-vs-rest ROC-AUC of 99.63%. To understand which architectural component drives this performance, three two-component variants — CNN+Swin, CNN+LSTM, and a genuine Swin Transformer encoder paired with a BiLSTM (Swin+LSTM) — were trained under an identical three-epoch budget and evaluated on the same held-out test set. Swin+LSTM was consistently the strongest of the three, reaching 90.25% accuracy and a macro F1-score of 90.28%, ahead of CNN+Swin (84.42%) and CNN+LSTM (81.33%). Grad-CAM heatmaps are provided to visualize which regions of an MRI slice drove a given prediction, and the resulting pipeline is deployed through an interactive Streamlit application. The results suggest that pairing a windowed-attention Transformer encoder with a sequence-aggregating recurrent head is a particularly effective and comparatively lightweight combination for this task.

Related

  1. An explainable hybrid transformer–CNN framework for multi-class brain tumor classification from MRI images
    Bukkapatnam Rakesh Babu, Vullanki Rajesh, Pardhu Thottempudi +1
    Scientific ReportsPeer-reviewedCC BY 4.0Full textNeurologyCrossref
  2. Comparative and explainable deep learning for cherry (Prunus avium L.) cultivar classification using convolutional neural networks (CNN) and transformer architectures
    Bünyamin Demir, Fatih Ünal, Ömer Türk +5
    Scientific ReportsPeer-reviewedCC BY 4.0Crossref
  3. An explainable hybrid CNN-transformer framework for multi-representation leukaemia classification
    Yogita Nikhare, Sagarkumar Badhiye
    Scientific ReportsPeer-reviewedCC BY-NC-ND 4.0Full textCrossref
  4. An explainable vision transformer approach for automated neurological disorder classification from brain MRI scans
    Abdulrahman Alahmadi, Nayab Asim, Mohammad Zubair Khan +2
    Scientific ReportsPeer-reviewedCC BY 4.0Full textNeurologyCrossref
  5. An explainable swarm-optimized hybrid GRU–Transformer framework with wavelet-based spatial–temporal feature modeling and activation function analysis for rainfall classification across Pakistan’s climatic regimes
    Hira Farman, Muhammad Arif Hussain, Umme Laila +4
    Scientific ReportsPeer-reviewedCC BY 4.0Full textEnvironmental science & pollutionCrossref

Cite

BibTeXRIS
BibTeX
@misc{wankhade2026hybrid,
  title = {{Hybrid CNN–Transformer–LSTM Architectures for Explainable Brain Tumor MRI Classification}},
  author = {Viraj Wankhade and Vrushali Paratkar and Bhuvan Bhiogade and Tushar Baghele and Prof. Dr. U. P. Akare Akare},
  howpublished = {Hybrid CNN–Transformer–LSTM Architectures for Explainable Brain Tumor MRI Classification},
  year = {2026},
  month = oct,
  doi = {10.5281/zenodo.23642879},
  url = {https://zenodo.org/records/23642879},
  note = {CC BY 4.0},
}
RIS
TY  - UNPB
TI  - Hybrid CNN–Transformer–LSTM Architectures for Explainable Brain Tumor MRI Classification
AU  - Viraj Wankhade
AU  - Vrushali Paratkar
AU  - Bhuvan Bhiogade
AU  - Tushar Baghele
AU  - Prof. Dr. U. P. Akare Akare
PY  - 2026
DA  - 2026/10/25
PB  - Hybrid CNN–Transformer–LSTM Architectures for Explainable Brain Tumor MRI Classification
DO  - 10.5281/zenodo.23642879
UR  - https://zenodo.org/records/23642879
AB  - Automated classification of brain tumors from magnetic resonance imaging (MRI) can support radiologists by narrowing down differential diagnoses, but clinical usefulness depends on a model being both accurate and interpretable. In this work we design and empirically compare hybrid deep-learning architectures that combine a convolutional neural network (CNN), a Swin-style Transformer sequence refiner, and a bidirectional long short-term memory (BiLSTM) network for four-class brain tumor classification: Glioma, Meningioma, No Tumor, and Pituitary. A full hybrid model (ResNet-50 → Transformer refiner → BiLSTM) trained for fifteen epochs reaches 94.42% test accuracy, a macro F1-score of 94.44%, and a macro one-vs-rest ROC-AUC of 99.63%. To understand which architectural component drives this performance, three two-component variants — CNN+Swin, CNN+LSTM, and a genuine Swin Transformer encoder paired with a BiLSTM (Swin+LSTM) — were trained under an identical three-epoch budget and evaluated on the same held-out test set. Swin+LSTM was consistently the strongest of the three, reaching 90.25% accuracy and a macro F1-score of 90.28%, ahead of CNN+Swin (84.42%) and CNN+LSTM (81.33%). Grad-CAM heatmaps are provided to visualize which regions of an MRI slice drove a given prediction, and the resulting pipeline is deployed through an interactive Streamlit application. The results suggest that pairing a windowed-attention Transformer encoder with a sequence-aggregating recurrent head is a particularly effective and comparatively lightweight combination for this task.
ER  -