Ai-powered histological fingerprinting

Patent metadata

Field	Value
Publication number	US-2025044390-A1
Application number	US-202318364498-A
Country	US
Kind code	A1
Filing date	Aug 3, 2023
Priority date	Aug 3, 2023
Publication date	Feb 6, 2025
Grant date	—

Abstract

Systems and methods for AI-powered histological fingerprinting in magnetic resonance imaging. MR signal data of an object is acquired using a high sensitivity scanner. Ground truth tissue microstructure data is acquired for the object. A forward model is learned using machine learning. The forward model is used to generate a dictionary or to train a model to map the signals to the histological parameters including the tissue microstructure of a scanner object. A signal-to-signal translation model is also provided to provide signals with improved sensitivity.

First claim

1 . A system for histological fingerprinting, the system comprising:
an MRI scanner configured to generate signal evolutions for a sequence while scanning a patient; a histological forward model learned using machine learning, the histological forward model configured to input tissue microstructure properties and output signal evolutions; a histological dictionary created using the histological forward model; and an image processor configured to reconstruct an image including at least tissue microstructure properties of the patient, wherein the tissue microstructure properties of the patient are determined using the histological dictionary to map the signal evolutions to the tissue microstructure properties.

2 . The system of claim 1 , wherein the MRI scanner includes a maximum gradient amplitude of 300 mT/m or more.

3 . The system of claim 1 , wherein the histological forward model is learned using an encoder decoder network.

4 . The system of claim 1 , wherein the histological forward model is learned using machine learning with post-mortem high-resolution MRI of human tissue with corresponding microscopy images.

5 . The system of claim 1 , wherein the histological forward model is learned using machine learning with tissue microstructure information acquired using digital pathology and respective signal data.

6 . The system of claim 5 , wherein the respective signal data is simulated.

7 . The system of claim 1 , wherein the tissue microstructure properties includes at least one of cell size, cell shape, or cell wall permeability.

8 . The system of claim 1 , further comprising:
a display configured to display the image.

9 . A method for generating signals with improved sensitivity for histological fingerprinting, the method comprising:
acquiring low sensitivity signal data of an object from a low sensitivity scanner; acquiring high sensitivity signal data of the object from a high sensitivity scanner; and training a model by inputting the low sensitivity signal data and the high sensitivity signal data into the model configured to generate realistic signal data with improved sensitivity when inputting data from the low sensitivity scanner.

10 . The method of claim 9 , wherein the model is trained using a maximum entropy reinforcement learning framework.

11 . The method of claim 9 , wherein the high sensitivity signal data is simulated signal data.

12 . The method of claim 9 , wherein the model comprises a signal-to-signal model generated using both deep learning and reinforcement learning.

13 . The method of claim 9 , wherein the object comprises a specific cell structure.

14 . A method for learning a forward model for histological fingerprinting, the method comprising:
generating electron microscopy of animal tissue; acquiring, by a high sensitivity MRI scanner, signals of the animal tissue; training a forward model using the electron microscopy and the signals of the animal tissue; and outputting a trained forward model for histological fingerprinting.

15 . The method of claim 14 , further comprising:
creating a histological dictionary using the trained forward model.

16 . The method of claim 15 , further comprising:
applying the histological dictionary for image reconstruction during a medical imaging procedure.

17 . The method of claim 14 , wherein the forward model is trained using machine learning with post-mortem high-resolution MRI of human tissue with corresponding microscopy images.

18 . The method of claim 14 , wherein the forward model is trained using machine learning with tissue microstructure information acquired using digital pathology and respective signal data.

19 . The method of claim 18 , wherein the respective signal data is simulated.

20 . The method of claim 14 , wherein the high sensitivity MRI scanner includes a maximum gradient amplitude of 300 mT/m or more.

Assignees

• Siemens Healthineers Ag

Inventors

• Mostapha Mahmoud
• Comaniciu Dorin
• Nadar Mariappan S

Classifications

• A61B5/055

  involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging · CPC title

• G01R33/5608 Primary

  Data processing and visualization specially adapted for MR, e.g. for feature analysis and pattern recognition on the basis of measured MR data, segmentation of measured MR data, edge contour detection on the basis of measured MR data, for enhancing measured MR data in terms of signal-to-noise ratio by means of noise filtering or apodization, for enhancing measured MR data in terms of resolution by means for deblurring, windowing, zero filling, or generation of gray-scaled images, colour-coded images or images displaying vectors instead of pixels (image data processing or generation, in general G06T) · CPC title