Stroke diagnosis apparatus based on artificial intelligence and method

Inventors

Kim, Dohyun • Song, Soo Hwa • Jung, Sumin • Lee, Jin Soo • Lee, Seong Joon

Assignees

Heuron Co Ltd

Abstract

Provided is a stroke diagnosis apparatus based on AI that includes: an image obtainer obtaining a non-contrast CT image related to the brain of at least one patient; a preprocessor pre-processing the non-contrast CT image and determining whether the at least one patient is in a non-hemorrhage state or a hemorrhage state on the basis of the pre-processed image; an image processor normalizing the pre-processed image and dividing and extracting an ROI (Region of Interest) using a preset standard mask template; and a determiner determining whether there is a problem with a cerebral large vessel of the at least one patient using the divided and extracted ROI.

Core Innovation

The invention relates to a stroke diagnosis apparatus based on AI (Artificial Intelligence) that obtains a non-contrast CT image related to a brain of at least one patient. The apparatus pre-processes the non-contrast CT image and determines whether the at least one patient is in a non-hemorrhage state or a hemorrhage state based on the pre-processed image. The apparatus normalizes the pre-processed image and divides and extracts an ROI (Region of Interest) using a preset standard mask template.

Using the divided and extracted ROI, the apparatus includes a determiner that determines whether there is a problem with a cerebral large vessel of the at least one patient. When the determiner estimates an ASPECT score of the at least one patient using the divided and extracted ROI in response to the presence of the cerebral large vessel problem, the apparatus determines that the at least one patient is a patient to whom mechanical thrombectomy is applied only when the estimated ASPECT score is a predetermined value or more.

The disclosure also describes AI-based classification from the ROI to estimate ischemia-related features, including Old infarct (OI), Recent infarct (RI), Frank hypodensity (FH), and Early ischemic sign (EIS), with FH/EIS treated as acute affected parts. The determiner detects emergent large vessel occlusion using a dense MCA sign at an infra-ganglionic level, and when the dense MCA sign is absent it uses HU value frequency and HU-frequency differences across left and right hemispheres with predetermined reference and difference values.

In addition, the disclosure describes optionally calculating a tissue clock and using it with affected-part information such as EIS/FH and/or affected-part volume-based models to support the mechanical thrombectomy determination. The apparatus further describes communicating patient or tissue-clock information, ASPECT, and elapsed-time information to a tertiary hospital, and using the diagnosis to select patient and normal groups to increase clinical trial success probability.

Claims Coverage

The partial content provides one independent claim (clm-00001). The inventive features focus on an AI-based CT image pipeline that gates mechanical thrombectomy on hemorrhage state, cerebral large vessel problem determination, and an ASPECT score threshold using an ROI extracted with a preset standard mask template.

Across the independent claim, the main inventive concept is an AI-based non-contrast CT workflow that pre-processes and classifies hemorrhage state, normalizes and extracts an ROI using a preset standard mask template, determines a cerebral large vessel problem from the ROI, and estimates an ASPECT score and applies a predetermined threshold to decide eligibility for mechanical thrombectomy.

Stated Advantages

Documented Applications