Artificial intelligence (AI) based parkinson's disease diagnosing apparatus and method
Inventors
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-11013452-B1
Publication Date
2021-05-25
Expiration Date
Abstract
The present invention relates to an artificial intelligence-based Parkinson's disease diagnosing apparatus and method. An artificial intelligence-based Parkinson's disease diagnosing apparatus which is an aspect of the present disclosure includes an image acquiring unit which acquires a first image related to a multi-echo magnitude and phase from an MRI obtained by capturing a brain of a patient; an image processing unit which post-processes the acquired first image so as to observe a substantia nigra and a nigrosome-1 region used as imaging biomarkers of the Parkinson's disease; an image analyzing unit which analyzes the post-processed first image to classify a second image including the nigrosome-1 region and detects the nigrosome-1 region from the classified second image; and a diagnosing unit which diagnoses whether the patient has the Parkinson's disease by analyzing whether the detected nigrosome-1 region is normal, in which the image processing unit generates a susceptibility map weighted imaging image by applying a quantitative susceptibility map mask to the first image based on a quantitative susceptibility mapping algorithm to perform the post-processing and the image processing unit further performs at least one operation of angle adjustment, image enlargement, and reslice, on the generated susceptibility map weighted imaging image.
Core Innovation
The invention relates to an apparatus for detecting Parkinson's disease using magnetic resonance imaging, where a plurality of first images of a patient brain are acquired as multi-echo magnitude and phase images. The acquired plurality of first images are post-processed by generating susceptibility map weighted imaging images through application of a quantitative susceptibility mapping algorithm and a quantitative susceptibility map mask. The susceptibility map weighted imaging images are further processed by performing one or more of angle adjustment, image enlargement, or reslice.
The post-processed first images are analyzed to obtain second images classified as including the nigrosome-1 region, and the nigrosome-1 region is detected using machine learning. The classification is performed by detecting a red nucleus and a substantia nigra present in the post-processed first images, and the plurality of post-processed first images are sequentially analyzed to identify at which image the detected red nucleus disappears. In some implementations, a one-stage detector is used, including a convolutional neural network with a feature pyramid network and adjusted classification and regression losses.
A diagnosing unit determines whether the detected nigrosome-1 region is normal, thereby providing information on whether the patient has Parkinson's disease. In some implementations, robustness is improved by combining predictive results from a plurality of learning models using a majority rule. The system is also described as being used to screen patient and normal groups to increase clinical trial success probability.
Claims Coverage
The independent claims are directed to an apparatus and a method for detecting Parkinson's disease and providing diagnostic information based on multi-echo magnitude and phase MRI, QSM-derived susceptibility map weighted imaging, and machine-learning sequential detection of the nigrosome-1 region using red nucleus disappearance. Across the independent claims, the inventive features focus on the image acquisition/post-processing pipeline and the machine-learning sequential nigrosome-1 detection and diagnosis decision.
Multi-echo MRI susceptibility map weighted imaging generation and post-processing
Acquiring a plurality of first images by scanning a brain of a patient using magnetic resonance imaging, the plurality of first images being multi-echo magnitude and phase images; generating a plurality of susceptibility map weighted imaging images by applying a quantitative susceptibility map mask to the plurality of first images based on a quantitative susceptibility mapping algorithm; and performing one or more of angle adjustment, image enlargement, or reslice on the plurality of generated susceptibility map weighted imaging images.
Machine-learning sequential red nucleus disappearance based nigrosome-1 detection
Analyzing the plurality of post-processed first images to obtain a plurality of second images that are classified as including the nigrosome-1 region and detect the nigrosome-1 region of the plurality of second images, the plurality of second images being classified by detecting a red nucleus and a substantia nigra present in the plurality of post-processed first images through machine learning and sequentially analyzing the plurality of post-processed first images to identify at which image the detected red nucleus disappears.
Diagnosing by determining normality of detected nigrosome-1 region
Determining, whether the detected nigrosome-1 region in the plurality of second images is normal.
Providing diagnostic information based on detected nigrosome-1 region
Providing information on the detected nigrosome-1 region to that indicates whether the patient has Parkinson's disease.
Across the independent claims, Parkinson's disease detection information is provided by combining (i) multi-echo magnitude and phase MRI acquisition with (ii) susceptibility map weighted imaging generation using a quantitative susceptibility mapping algorithm and quantitative susceptibility map mask with (iii) angle adjustment, image enlargement, and/or reslice, and (iv) machine-learning sequential detection of the nigrosome-1 region based on detecting a red nucleus and substantia nigra and identifying the image index where the detected red nucleus disappears, followed by (v) normality determination and diagnosis information provision.
Stated Advantages
Increase clinical trial success probability by screening patient and normal groups.
Documented Applications
Screening patient and normal groups to increase clinical trial success probability.
Interested in licensing this patent?