Microwave Breast Imaging With DRL-Guided Sparse Radar Scanning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional breast cancer detection methods, such as X-ray mammography and MRI, suffer from limitations like discomfort, health risks, and time consumption, while existing microwave imaging techniques require dense radar measurements, making them inefficient for practical use.

Innovation Solution

A computationally efficient model-based reconstruction algorithm combined with a Deep Reinforcement Learning (DRL) approach is used to optimize radar scanning locations, reducing the number of scans required for accurate tumor detection by leveraging sparsity and intelligent scanning mechanisms.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If dense radar measurements are used in microwave imaging, then measurement precision is improved, but loss of time increases

Engineering Contradiction:
Improvetumor detection accuracyVSAvoidscan duration
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs a preliminary coarse scan at uniform locations to obtain initial microwave data, which is then used by the reinforcement learning agent to intelligently select subsequent scanning locations. This preliminary action enables the system to avoid dense uniform scanning while maintaining detection accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The reinforcement learning agent dynamically adapts the radar scanning locations based on the initial coarse scan data, transitioning from fixed uniform scanning to intelligent adaptive scanning. This dynamic adjustment optimizes the scanning path to focus on regions of interest, reducing overall scan time while maintaining precision.

Inventive Principle:
Principle #15Dynamics

2Productivity

If conventional reconstruction algorithms are used, then ease of operation is maintained, but productivity decreases

Engineering Contradiction:
Improvereconstruction speedVSAvoidalgorithm complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent replaces traditional iterative reconstruction algorithms with a deep learning-based reconstruction network. This substitution transforms the reconstruction process from a computationally intensive iterative optimization problem into a faster neural network inference task, significantly improving reconstruction speed.

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

Solution Approach 2:

The system changes the fundamental approach to reconstruction by using learned parameters from training data rather than solving physical equations iteratively. The deep learning model has been trained to map raw microwave measurements directly to reconstructed images, bypassing the need for complex iterative inversion algorithms.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If uniform scanning locations are used, then ease of operation is improved, but measurement precision decreases

Engineering Contradiction:
Improvetumor localization accuracyVSAvoidscanning complexity
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The system performs a preliminary coarse scan at uniform locations to obtain initial microwave data, which is then used by the reinforcement learning agent to intelligently select subsequent scanning locations. This preliminary action enables the system to avoid dense uniform scanning while maintaining detection accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The reinforcement learning agent uses feedback from the initial coarse scan to adaptively determine subsequent scanning locations. The system continuously refines its scanning strategy based on information gathered from previous measurements, focusing resources on regions most likely to contain tumors.

Inventive Principle:
Principle #23Feedback

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

The method provides improved tumor localization with reduced clutter and up to 2 times Signal to Mean Ratio (SMR) improvement over conventional methods, while significantly reducing scan duration.

Implementation Method 1

a microwave radar is configured to send and receive waves at a set of optimized locations

Methodology Applied
Scientific EffectElectromagnetic radiation: Electromagnetic Induction

Implementation Method 2

estimate an unknown dielectric constant of a region of interest in the specific portion

Methodology Applied
Scientific EffectDielectric property measurement: Dielectric Permittivity

Data Source

PatentEP4480404B1Systems for in body microwave imaging of a subject
Publication Date: 2026.04.15 TATA CONSULTANCY SERVICES LTD
  • EP4480404B1 patent drawingFigure 1
  • EP4480404B1 patent drawingFigure 2(a)~2(b)
  • EP4480404B1 patent drawingFigure 3

AI summary

Detecting cancer early can significantly reduce mortality rate, but this still remains a challenge owing to shortcomings in early screening and detection with existing modalities. Cancer detection is done using known screening methods such as X-ray mammography, Magnetic Resonance Imaging (MRI) and Ultrasound imaging (US). But these conventional methods have their own limitations such as compression discomfort, inherent health risks, expensive, and consume more time and effort. Present disclosure provides system and method for enhanced microwave imaging (MWI) for efficient breast tumor detection by scanning subject's specific body portion to optimize the scan duration. The MWI is framed as an inverse problem by building forward model using a Point Spread Function (PSF) and is solved by imposing sparsity prior since tumor is concentrated to limited regions. The entire scanning duration is optimized by viewing the problem as a sequential decision making process for a Deep Reinforcement Learning (DRL) agent.