Article Details
  • Published Online:
    April  2025
  • Product Name:
    The IUP Journal of Electrical & Electronics Engineering
  • Product Type:
    Article
  • Product Code:
    IJEEE030425
  • DOI:
    10.71329/IUPJEEE/2025.18.2.54-68
  • Author Name:
    Dushyantha N D, Kavya R Naik, Chaithanya H G, Amrutha L and Dinesh C
  • Availability:
    YES
  • Subject/Domain:
    Engineering
  • Download Format:
    PDF
  • Pages:
    54-68
Volume 18, Issue 2, April 2025
Brain Tumor Detection and Survival Prediction Using 3D U-Net and Random Forest: A Streamlit-Based Interactive System for MRI Analysis
Abstract

The paper presents a smart, end-to-end framework for automated brain tumor detection, segmentation, and survival estimation that synergizes machine learning (ML) and deep learning (DL). The proposed method conducts volumetric segmentation of preprocessed Magnetic Resonance Imaging (MRI) images with a 3D U-Net structure. Considering the age, tumor size, and how much of the tumor was excised, the patient’s survival time after tumor diagnosis is predicted using a Random Forest (RF) regression model. Furthermore, the strategy includes interpretability in survival forecasting using SHAP (SHapley Additive exPlanations) plots. The tool, which is built on Streamlit platform, is an interactive and intuitive user interface supporting MRI uploads, axial, sagittal, and coronal images, dynamic overlays of the tumor, and PDF reports that are downloadable. With interpretable ML together with reliable segmentation, the study demonstrates an effective and computationally and clinically viable strategy for assisting oncologists and radiologists in the diagnosis and forecasting of brain tumor patients.

Introduction

Brain tumor is one of the most grievous and deadly medical conditions whose unchecked growth would lead to very severe neurological compromise or even fatality.