Brain Image Analysis using Intel® AI Software and Intel® Tiber™ AI Cloud: Developer Spotlight

Nikita_Shiledarbaxi · ‎03-17-2025

In the evolving landscape of AI and quantum computing, the ability to enhance medical imaging through advanced neural networks is becoming increasingly crucial. Precise segmentation and evaluation of brain images are essential for diagnosing neurological conditions such as tumours, strokes, and neurodegenerative diseases. However, traditional models often lack the required precision and efficiency for early detection and treatment planning. At the 2024 Intel® Student Ambassador Fall Hackathon, a team of Intel Student Ambassadors developed Quantum SegNet, a quantum-enhanced neural network for brain image analysis to address this challenge.

The Quantun SegNet solution leverages Intel’s AI tools and framework optimizations available on the Intel® Tiber™ AI Cloud. This blog will give you an overview of the project.

About The Intel® Student Ambassadors

The following were the student ambassadors who built the Quantum SegNet project:

Dike Peter, currently pursuing a master’s degree in IT, Digitalization, and Sustainability, along with Business Informatics at the University of Applied Arts and Sciences in Switzerland,
Sreeneha Samudrala Snighdha (Neha), pursuing her master’s degree in Software Engineering in Vienna, and
Nwankwo Chijioke, a computer science master’s student in Northeastern Illinois University.

Each team member leverages Intel Tiber AI Cloud and oneAPI tools to develop optimized AI solutions as part of the Intel Student Ambassador Program.

About the Quantum SegNet Project

Quantum SegNet is an AI-powered solution designed to enhance brain image analysis using Quantum Neural Networks (QNNs). The project integrates quantum computing techniques with AI-driven image segmentation models, significantly improving accuracy and efficiency in medical diagnostics.

The model features Quantum-Enhanced Segmentation, which utilizes a Quanvolution Filter - a novel quantum feature extraction technique that enhances traditional convolutional layers.

Model Architecture: An Overview

The deep learning framework consists of multiple processing layers:

Feature Extraction: Uses a quantum-inspired filtering technique to enhance image clarity.
Segmentation Processing: Optimizes classification of brain structures.
Inference Acceleration: Utilizes Intel® Distribution of OpenVINO™ Toolkit to optimize real-time image segmentation.

Fig.1: Quantum SegNet Structure

Model Optimization and Deployment with OpenVINO™ Toolkit

The team converted the PyTorch* model to ONNX and subsequently to OpenVINO for deployment. The conversion process was crucial for ensuring that the model could run efficiently on Intel’s architecture, minimizing latency and maximizing inference speed. Here’s how the conversion looked:

import torch
import onnx
from openvino.runtime import Core

# Convert PyTorch model to ONNX
model = torch.load('model.pth')
torch.onnx.export(model, dummy_input, "model.onnx", opset_version=11)

# Load ONNX model and optimize with OpenVINO
core = Core()
model_onnx = core.read_model("model.onnx")
compiled_model = core.compile_model(model_onnx, "CPU")

The OpenVINO Toolkit’s ability to optimize and compress the model helped significantly reduce processing time, making it feasible for real-time brain image analysis. This phase of the project required detailed tuning and testing.

Intel® Tools and Technologies Used

The Quantum SegNet team leveraged Intel’s AI and quantum computing ecosystem to optimize their model:

Intel Tiber AI Cloud: Provided computational resources to run quantum-enhanced models efficiently.
Intel Distribution of OpenVINO Toolkit: Enabled real-time inference optimization for faster segmentation.
Intel® Extension for PyTorch and Intel® Extension for TensorFlow*: Assisted in model training and deployment.

The team also used TorchQuantum & ONNX Runtime for quantum neural network development and cross-platform compatibility.

Check out the complete project on GitHub.

What’s Next?

Quantum SegNet has the potential to revolutionize AI-driven medical imaging. The team plans to further optimize the model and explore real-world applications in hospitals and research institutions.

We encourage you to jump-start AI development with our oneAPI-powered developer resources. Sign up to Intel Tiber AI Cloud to leverage our optimized software portfolio on the latest accelerated hardware for high-performance AI and accelerated computing.

Useful Resources

Additional References

Abdelruhman Essam (2024) Brain Tumor Classification Using CNN. (Kaggle)
Hussain, A. et al. (2024) Brain Tumor Diagnosis Using Quantum Convolutional Neural Networks. (ResearchGate)