TimeDistributed-CNN-LSTM: A Hybrid Approach Combining CNN and LSTM to Classify Brain Tumor on 3D MRI Scans Performing Ablation Study

Identification of brain tumors and accurate grading at an early stage are crucial in cancer diagnosis, as a timely diagnosis can increase the chances survival. Considering challenges risks tumor biopsies, noninvasive imaging procedures such Magnetic Resonance Imaging (MRI) extensively used analyzing tumors. Recent advances field medical with deep learning using three dimensional (3D) MRI is aiding clinical experts significantly tumor. In this study, BraTS datasets named 2018, 2019 2020 employed to classify into high-grade glioma (HGG) low-grade (LGG) where each contains four different sequences 3D images T1-weighted (T1), contrast enhancement (T1ce), T2-weighted (T2), Fluid Attenuated Inversion Recovery (FLAIR) for single patient. This research composed two approaches where, first part, we propose hybrid model TimeDistributed-CNN-LSTM (TD-CNN-LSTM) combining Convolutional Neural Network (CNN) Long Short Term Memory (LSTM) layer architecture wrapped TimeDistributed function. The objective developing consider all patient input data because every sequence necessary information on that have impact improving detection performance. However, interpreting together optimal performance especially quite challenging. Therefore, developed configuration based highest accuracy performing ablation study hyper-parameters. second CNN trained respectively compare TD-CNN-LSTM model. regard, both models, 2018 combined number train dataset test dataset. Moreover, before training models preprocessed ensure Our results these demonstrate network outperforms achieving 98.90%. Later, evaluate consistency, evaluated K-fold cross validation k values. approach putting time CNN-LSTM good generalization capability be effectively future Computer Aided Diagnosis (CAD) which aid radiologists diagnostics.