Fully Quantum U-Net With Ancillary-Qubit Concatenation for Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing quantum-classical deep learning hybrid models for image segmentation face limitations in quantum operations, and mapping the classical U-Net architecture into the 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 to maintain sequential application and spatial information, and the circuit is measured without ancillary qubits to preserve the image shape.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If quantum operations are limited to quantum feature maps or parametrized circuits in hybrid models, then the model complexity is reduced and implementation is easier, but the computational power and segmentation performance are insufficient

Engineering Contradiction:
Improvequantum operations complexityVSAvoidsegmentation performance
Core Design Contradiction:
Device complexityVSProductivity

Solution Approach 1:

The patent divides the quantum circuit into distinct functional modules: encoding layer, quantum convolution layers (with different kernel sizes), and output layer. Each module performs specific processing functions, allowing the system to achieve complex segmentation capabilities through coordinated simple operations.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces quantum superposition as an additional computational dimension, allowing the circuit to process multiple input patterns simultaneously through quantum states. This enables the model to achieve high segmentation performance without proportionally increasing circuit complexity.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Adaptability or versatility

If the classical U-Net architecture is directly mapped to quantum domain, then the architectural structure is preserved, but the unitary nature and mid-quantum measurement difficulties cause implementation challenges

Engineering Contradiction:
Improvearchitectural adaptabilityVSAvoidquantum implementation ease
Core Design Contradiction:
Adaptability or versatilityVSEase of manufacture

Solution Approach 1:

The patent replaces classical mechanical operations with quantum mechanical operations. Classical convolution operations are substituted with quantum convolution layers that use quantum gates and superposition states. Classical concatenation is replaced with quantum state manipulation that naturally preserves spatial information through quantum mechanics.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent changes the fundamental parameters of the U-Net architecture to be quantum-compatible. Instead of using classical weights and biases, the model uses quantum states and unitary transformations. The architecture parameters are redefined in terms of quantum circuit depth, number of qubits, and gate operations, making the system implementable on quantum hardware.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If quantum convolution layers are applied sequentially with superposition control, then spatial information is preserved and segmentation accuracy is improved, but the circuit depth and computational steps increase

Engineering Contradiction:
Improvesegmentation accuracyVSAvoidcircuit depth
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent implements nested quantum convolution layers where different kernel sizes are applied in a hierarchical manner. Smaller kernel convolutions are nested within larger kernel structures, allowing the system to process spatial information at multiple scales simultaneously without proportionally increasing overall circuit depth.

Inventive Principle:
Principle #7Nested doll (Nesting)

Solution Approach 2:

The patent performs preliminary quantum state preparation and superposition creation before applying the main quantum convolution operations. By pre-preparing the quantum states in ancillary qubits, the system reduces the computational steps required during the main processing phase, effectively managing circuit depth while maintaining high accuracy.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4693118A1Method and system for a fully quantum u-net for image segmentation
Publication Date: 2026.02.11 TATA CONSULTANCY SERVICES LTD
  • EP4693118A1 patent drawingFigure 1
  • EP4693118A1 patent drawingFigure 2A
  • EP4693118A1 patent drawingFigure 2B

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.