International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 05 | May 2025
p-ISSN: 2395-0072
www.irjet.net
NeuroDepthNet: A Deep Learning-Based System for Brain Tumor Classification and Depth Estimation Mrs. Sumaiya1, Ms. Gagana H P2, Ms. S Bhoomika3, Ms. Srushti Harish4, Mr. Suhas H K5 1Assistant Professor, Dept. of Computer Science and Engineering, Maharaja Institute of Technology
Thandavapura
2345Students, Dept of Computer Science and Engineering, Maharaja Institute of Technology Thandavapura
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Early and accurate detection of brain tumors is critical for improving patient outcomes. This paper presents a dual-module approach for optimized brain tumor detection and classification, integrated with a novel estimation of tumor depth using 3D reconstruction. The system combines the power of deep neural networks (DNNs) for classification with medical image processing using SimpleITK and VTK for depth value estimation through 3D reconstruction. The proposed architecture addresses the critical gap in tumor localization and depth analysis, enhancing surgical planning and clinical diagnostics. The pipeline has been evaluated on MRI datasets with promising results in accuracy, visual quality of reconstructions, and computational efficiency.
advancement of artificial intelligence, especially in medical imaging. Their hierarchical architecture enables the automatic extraction of complex features from highdimensional data such as MRI scans. CNNs have demonstrated superior performance in image classification, segmentation, and detection tasks, making them highly suitable for brain tumor diagnosis. In the context of this research, the motivation lies not only in accurately detecting and classifying brain tumors but also in addressing the critical need for spatial depth analysis—an aspect often overlooked in 2D imaging approaches. By integrating DNNbased classification with 3D reconstruction using tools like VTK and SimpleITK, this work aims to bridge the gap between precise tumor identification and spatial understanding, ultimately contributing to improved clinical decision-making.
Key Words: Brain tumor detection, classification, 3D reconstruction, VTK, SimpleITK, Deep Neural Networks, tumor depth estimation, medical imaging.
1.3 CHALLENGES TO BE ADDRESSED
1.INTRODUCTION
Artificial intelligence (AI) has emerged as a pivotal force in technological advancement, with applications extending into nearly every field—including medicine. Within this landscape, Deep Neural Networks (DNNs) have demonstrated exceptional capability in pattern recognition and classification tasks, particularly in medical imaging. Despite their promise, several challenges must be addressed to effectively integrate DNNs into brain tumor diagnostics. One key issue is the limited availability of annotated medical datasets necessary for training robust models. Additionally, conventional 2D MRI-based analysis often lacks spatial context, making it difficult to estimate tumor depth—a crucial factor for surgical planning. This project seeks to overcome these challenges by combining DNN-based classification with 3D reconstruction techniques using VTK and SimpleITK, thereby offering a more comprehensive and interpretable diagnostic tool for brain tumor detection.
Brain tumors pose a significant risk to human life due to their invasive nature and the complexity of surgical intervention. The medical community heavily relies on Magnetic Resonance Imaging (MRI) for the diagnosis and treatment planning of brain tumors. However, traditional 2D slice-based analysis often lacks depth perception, making it difficult to assess the tumor’s full extent.This paper introduces an optimized framework that not only automates tumor detection and classification using deep learning but also estimates the depth value by reconstructing the tumor in 3D using SimpleITK and VTK libraries. This dual-module system aims to bridge the gap between classification and spatial quantification, offering a more holistic view for radiologists and neurosurgeons. 1.1 OBJECTIVE To develop a dual-module system that Detects and classifies brain tumors using a DNN classifier Estimates tumor depth through 3D reconstruction using SimpleITK and VTK Enhances diagnostic decision-making by integrating classification results with 3D spatial information.
1.4 CNN and DNN Convolutional Neural Networks (CNNs) are a specialized class of deep learning models designed to process grid-like data such as medical images. They utilize multiple layers, including convolutional, pooling, normalization, and fully connected layers, to automatically extract and learn hierarchical features from input data with minimal manual preprocessing. CNNs are often referred to as shift-invariant
1.2 MOTIVATION TO TAKE UP THE PROJECT Deep Neural Networks (DNNs), particularly Convolutional Neural Networks (CNNs), have become foundational in the
© 2025, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 576