NCI PDQ
Incidence & mortality snapshot
Official PDQ summary with U.S. and global incidence/mortality estimates and SEER-based rates.
Read sourceMove your cursor
What is NeuroDetect?
NeuroDetect is an educational, AI-powered brain tumor detection demo that analyzes MRI images and highlights regions the model focuses on. It’s designed to help users understand how machine learning can support medical imaging workflows.
The model behind NeuroDetect was trained on 7,000+ labeled MRI scans across multiple tumor categories, enabling it to learn patterns that are difficult to spot at a glance. In testing, it achieved strong performance (about 89.7% accuracy with a 0.9879 AUC), which demonstrates how deep learning can provide fast, consistent second‑look insights.
Brain tumors can affect people of any age and impact patients, families, and care teams worldwide. AI tools like NeuroDetect can help scale access to early review and triage support, especially in regions with limited specialists, by surfacing patterns quickly and consistently for clinicians to verify.
This experience is for learning and research only—not medical advice. Always consult qualified healthcare professionals for diagnosis and treatment decisions.
High Accuracy
89.7% accuracy on multi-class brain tumor classification with 0.9879 AUC score.
Explainable AI
Grad-CAM heatmaps show exactly where the model focuses for transparent decision-making.
Multi-Class Detection
Classifies: Glioma, Meningioma, Pituitary Tumors, and No Tumor cases.
Real-Time Analysis
Fast inference powered by MobileNetV2 transfer learning architecture.
Confidence Scoring
Get probability scores for all classes to understand model confidence levels.
Privacy First
Your images are processed securely and not stored or retained.
Built with
Python
FastAPI
TensorFlow
Keras
NumPy
HTML5
CSS3
JavaScript
Uvicorn
Gunicorn
SciPy
Pillow
Brain Tumor Analysis
Idle
Preview & Results
Learn
Curated research and clinical resources to explore brain tumors, classification, treatment, and outcomes. Educational use only.
CBTRUS
Population-level tumor statistics
The CBTRUS report aggregates U.S. registry data and details tumor histologies and incidence patterns.
Read sourceWHO 2021
Modern CNS tumor classification
WHO’s 5th edition integrates molecular diagnostics to refine CNS tumor families and types.
Read sourceMayo Clinic
Diagnosis & treatment options
Overview of surgery, radiation, chemotherapy, targeted therapy, and care planning.
Read sourceACS
Key statistics & risk context
Annual U.S. estimates and high‑level context about how common brain tumors are.
Read sourceACS
Survival rates & prognosis
How 5‑year relative survival rates vary by tumor type and age group.
Read sourceReferences & Dataset Information
Dataset Source
Brain Tumor MRI Dataset
Dataset: 7,023 Brain MRI images across 4 classes (Glioma, Meningioma, Pituitary, No Tumor)
Training: 5,712 images | Testing: 1,311 images
Model Architecture
MobileNetV2 Transfer Learning
Pre-trained on ImageNet for efficient feature extraction
Custom dense layers for multi-class classification
Explainability Method: Grad-CAM
Gradient-weighted Class Activation Mapping for visual explanations
Reference: Selvaraju et al., 2017 - "Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization"
Technologies Used
Backend: FastAPI, TensorFlow/Keras, Python
Frontend: HTML5, CSS3, Vanilla JavaScript
ML Libraries: NumPy, SciPy, Pillow
Disclaimer
Research & Educational Use Only
This system is designed for research and educational purposes. It is NOT intended to replace professional medical diagnosis or clinical decision-making. All results should be validated by qualified medical professionals. Always consult with healthcare providers before making any medical decisions.
Research & Clinical Sources
Model Performance
Accuracy: 89.70%
AUC Score: 0.9879
Training Method: Transfer Learning + Fine-tuning (30 epochs)
Test Set: 1,311 images from Kaggle dataset