X-ray CT Charge Sharing Correction via Deep Learning Restoration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Direct conversion type X-ray detectors in CT apparatuses face challenges with charge sharing, leading to reduced positional and energy resolution due to charge sharing between adjacent electrodes, which existing correction methods like coincidence counting and single-event-mode cannot fully address, resulting in suboptimal image quality.

Innovation Solution

An X-ray CT apparatus and data processing method that acquire data in two modes – one without coincidence counting correction and another with correction – and use a trained model to restore data with higher spatial resolution, effectively mitigating the effects of charge sharing by combining data from both modes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If charge sharing correction methods (coincidence counting or single-event-mode) are used, then spatial resolution is improved, but signal loss and reduced counting statistics occur

Engineering Contradiction:
Improvespatial resolutionVSAvoidcounting statistics
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

A deep learning model serves as an intermediary that processes raw data without coincidence counting correction and reconstructs images with high spatial resolution. The model learns the charge sharing patterns from training data and applies corrective transformations during inference, achieving spatial resolution improvement without the signal loss associated with traditional correction methods.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces the mechanical/electronic charge sharing correction mechanisms (coincidence counting circuits, single-event-mode logic) with a data-driven deep learning approach. This substitution allows the system to achieve correction benefits through software-based pattern recognition rather than hardware-based event filtering, preserving counting statistics.

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

2Quantity of substance

If data is acquired without coincidence counting correction, then counting statistics are improved, but spatial resolution deteriorates due to charge sharing

Engineering Contradiction:
Improvecounting statisticsVSAvoidspatial resolution
Core Design Contradiction:
Quantity of substanceVSMeasurement precision

Solution Approach 1:

The deep learning model is pre-trained on paired datasets (raw data without correction and reference data with correction or ground truth) to learn the charge sharing correction transformation. This preliminary training phase enables the model to automatically correct spatial resolution issues during actual operation without requiring real-time correction processing that would reduce counting statistics.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system changes the operational parameter from using coincidence counting correction during data acquisition to using a deep learning model that was trained on correction data. This parameter change allows the system to acquire data in the optimal mode (without correction, preserving statistics) while still achieving high spatial resolution through the pre-trained model's corrective transformations.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If traditional charge sharing correction methods are applied, then image quality is improved, but processing complexity and computational load increase

Engineering Contradiction:
Improveimage qualityVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

Complex real-time charge sharing correction processing is replaced with a pre-trained deep learning model that performs corrections during inference. The computationally intensive correction logic is moved from the data acquisition and reconstruction pipeline to the training phase, simplifying the operational processing complexity while maintaining image quality improvements.

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

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 solution enables the collection of images with improved spatial and energy resolution, achieving a high signal-to-noise ratio and enabling better material discrimination without the limitations of existing correction methods.

Implementation Method 1

when X-rays are absorbed by the semiconductor detection element, electron-hole pairs are generated

Methodology Applied
Scientific EffectPhotoelectric effect: Photoelectric Effect

Implementation Method 2

electrodes are provided on a side facing semiconductor detection elements to generate an electric field by the application of a bias voltage

Methodology Applied
Scientific EffectElectric field: Electric Field

Data Source

PatentEP4339894A1X-ray CT apparatus, data processing method, and storage medium
Publication Date: 2024.03.20 CANON MEDICAL SYST CORP
  • EP4339894A1 patent drawingFigure 1
  • EP4339894A1 patent drawingFigure 2A~2B
  • EP4339894A1 patent drawingFigure 3~4

AI summary

An X-ray CT apparatus (1) of an embodiment is provided with an acquisition unit (441) and a restoration unit (445). The acquisition unit (441) acquires first data in a first mode without coincidence counting correction. The restoration unit (445) restores, from the first data, second data having a higher spatial resolution than the first data.