Fully Quantum U-Net Segmentation With Quantum Convolution Layers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing quantum-classical deep learning hybrid models for image segmentation are limited by scarce or inadequate quantum operations, and mapping a classical U-Net architecture into a quantum domain is challenging due to unitary nature and mid-quantum circuit measurement difficulties.
Innovation Solution
A fully quantum U-Net architecture is developed, where backpropagation occurs along quantum layers, incorporating quantum convolution and concatenation operations, with each layer controlled by distinct quantum states in ancillary qubits, enabling sequential application and maintaining output shape.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If quantum operations are added to classical U-Net architecture, then image segmentation performance is improved, but device complexity increases
Solution Approach 1:
The patent divides the quantum circuit into distinct modules: encoding layer, quantum convolution layers (with controlled operations), and output layers. Each quantum convolution layer is controlled by distinct quantum states in ancillary qubits, allowing independent optimization and management of complexity while maintaining overall functionality for image segmentation.
Solution Approach 2:
The patent implements nested quantum convolution layers where each layer is controlled by quantum states from ancillary qubits. The controlled quantum operations are nested within the broader circuit structure, with each layer processing features at different scales. This nesting allows the model to handle complex segmentation tasks while organizing complexity in a hierarchical manner.
2Shape
If quantum convolution layers are applied sequentially, then output mask size is maintained, but computational steps increase
Solution Approach 1:
The patent employs periodic quantum operations where controlled quantum convolution layers are applied in sequence with periodic control from ancillary qubits. Each quantum convolution layer processes the feature map and maintains the spatial dimensions through controlled unitary operations, creating a periodic transformation pattern that preserves output shape while enabling deep feature extraction.
Solution Approach 2:
The patent replaces classical mechanical convolution operations with quantum convolution operations controlled by quantum states. The controlled quantum convolution layers use unitary transformations instead of classical matrix multiplications, providing the same functional output (maintaining mask size) with potentially faster computation through quantum parallelism.
3Measurement precision
If ancillary qubits are added for quantum state control, then layer control precision is improved, but quantum resource requirements increase
Solution Approach 1:
The patent designs the ancillary qubits to serve multiple functions: they control the quantum convolution layers through quantum states, enable the controlled operations, and facilitate the backpropagation process. This multi-functionality reduces the need for separate control mechanisms and optimizes the qubit allocation for achieving precise layer control.
Solution Approach 2:
The patent changes the parameters of the quantum states in ancillary qubits to control different quantum convolution layers. By varying quantum state parameters (such as rotation angles or state vectors), the system achieves precise control over the quantum operations without requiring additional qubits for each control parameter, optimizing the resource usage.
Data Source
AI summary
This disclosure relates generally to method and system for a fully quantum U-Net for image segmentation. Currently in image segmentation methods using quantum classical deep learning hybrid models, quantum operations are either scarce or limited to quantum feature maps or parametrized circuits. The disclosed quantum U-Net contains quantum versions of operations required for segmentation task, namely convolution and concatenation. The quantum U-Net is able to reproduce the predicted output mask having nearly the same size as its input image. In the disclosed architecture, the quantum convolution takes the form of a series of parametrized unitary gates as convolution layers which act locally on the input image data embedded into a quantum circuit to learn its features. The disclosed method is used for medical image segmentation, in food industry and so on.