Deep learning based vessel suppression on contrast enhanced MRI images

Abstract

In an aspect of the disclosure, a method, a computer-readable medium, and an apparatus are provided. The method including obtaining a set of first-type medical images of a subject, the first-type medical images containing a first feature and a second feature, and generating one or more second-type medical images based on the set of first-type medical images by selectively reducing signals corresponding to the second feature.

Description

135005-0001W001 1DEEP LEARNING BASED VESSEL SUPPRESSION ON CONTRAST ENHANCED MRI IMAGESCROSS-REFERENCE TO RELATED APPLICATION(S)

[0001] This application claims the benefit of U.S. Provisional Patent Application No.63 / 760,073, entitled “DEEP LEARNING BASED VESSEL SUPPRESSION ON CONTRAST ENHANCED MRI IMAGES” and filed on February 18, 2025, which is expressly incorporated by reference herein in its entirety.TECHNICAL FIELD

[0002] The present disclosure relates generally to medical imaging, and more particularly, to medical imaging with vessel suppression.INTRODUCTION

[0003] Magnetic Resonance Imaging (MRI) is a widely used non-invasiveimagingtechnique that provides high-resolution anatomical and functional imaging of soft tissues. Contrast-enhanced (CE) MRI, using Gadolinium-based Contrast Agent (GBCA) administration, enhances lesion visualization by increasing signal intensity in abnormal tissues. However, enhancing blood vessels can sometimes hinder the delineation of pathology, particularly in gradient echo (GRE) sequences. This issue is especially relevant in cases with small lesions, where differentiating between a blood vessel and a lesion, such as metastasis, can be challenging.

[0004] Spin echo (SE) sequences with vessel suppression can mitigate this issue, but they generally have lower resolution and may require additional scanning, which can be time-consuming and less accessible in normal clinical settings.BRIEF SUMMARY

[0005] The following presents a simplified summary of one or more aspects in order to provide a basic understanding of such aspects. This summary is not an extensive overview of all contemplated aspects. This summary neither identifies key or critical elements of all aspects nor delineates the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to the more detailed description that is presented later.135005-0001W001 2

[0006] In an aspect of the disclosure, a method, a computer-readable medium, and an apparatus are provided. The method including obtaining a set of first-type medical images of a subject, the first-type medical images containing a first feature and a second feature, and generating one or more second-type medical images based on the set of first-type medical images by selectively reducing signals corresponding to the second feature.

[0007] To the accomplishment of the foregoing and related ends, the one or more aspects may include the features hereinafter fully described and particularly pointed out in the claims. The following description and the drawings set forth in detail certain illustrative features of the one or more aspects. These features are indicative, however, of but a few of the various ways in which the principles of various aspects may be employed.BRIEF DESCRIPTION OF THE DRAWINGS

[0008] FIG. 1 is a diagram illustrating a magnetic resonance imaging (MRI) system, in accordance with various aspects of the present disclosure.

[0009] FIG. 2 is a diagram illustrating a contrast enhanced (CE) brain MRI images vessel suppression system, in accordance with various aspects of the present disclosure.

[0010] FIG. 3 is a diagram illustrating an example process fortraining a machine learning model, in accordance with various aspects of the present disclosure.

[0011] FIG. 4 is a diagram illustrating an example of vessel suppression GRE scheme using the CE brain MRI images vessel suppression system, in accordance with various aspects of the present disclosure.

[0012] FIG. 5 is a diagram illustrating qualitative results of vessel suppression in accordance with various aspects of the present disclosure.

[0013] FIG. 6 is a diagram illustrating qualitative and quantitative comparisons of contrast- enhanced brain MRI images across different imaging conditions, in accordance with various aspects of the present disclosure.

[0014] FIGs. 7A and 7B are flowcharts of methods of CE brain MRI images vessel suppression, in accordance with various aspects of the present disclosure.

[0015] FIG. 8 is a block diagram of an embodiment of a computer system 800, which can be utilized in accordance with various aspects of the present disclosure.DETAILED DESCRIPTION135005-0001W001 3

[0016] Various aspects of the present disclosure relate generally to medical imaging techniques, particularly in Magnetic Resonance Imaging (MRI). Some aspects more specifically relate to vessel suppression in contrast-enhanced (CE) MRI images to improve lesion visualization. In some examples, the disclosed techniques use an Artificial Intelligence (Al)-based vessel suppression approach to selectively reduce enhancing blood vessel signals in gradient echo (GRE) sequences, allowing for better differentiation between vascular structures and pathological abnormalities such as lesions or metastases.

[0017] The disclosed techniques offer several advantages over conventional vessel suppression approaches. For example, unlike SE sequences, which suppress vessel signals but suffer from lower resolution, the requirement for specific scanning protocol and / or equipment, and increased scanning time, the technical solution disclosed herein performs vessel suppression as a post-processing step on readily available GRE images, eliminating the need for additional SE acquisitions. By leveraging an Al-based / machine learning-based image-to-image synthesis model, the system can synthesize SE-like images from GRE images, generate a vessel map, and selectively attenuate vessel signals in the GRE images while preserving lesion contrast. This approach enhances diagnostic accuracy, particularly in cases involving small lesions, where distinguishing between blood vessels and metastases is challenging. Furthermore, the adjustable suppression factor allows for customizable vessel attenuation, giving radiologists flexibility in balancing vessel suppression and lesion visibility according to clinical needs.

[0018] Particular aspects of the subject matter described in this disclosure can be implemented to realize one or more of the following potential advantages. In some examples, by applying Al-based / machine learning-based vessel suppression to contrast-enhanced GRE images, the described techniques can be used to improve lesion detectability by reducing vessel signals without compromising lesion contrast, reduce scan time and patient burden by eliminating the need for additional SE sequences, enhance clinical workflow efficiency by providing vessel-suppressed GRE images as a post-processing step, enable dynamic vessel suppression control, allowing radiologists to fine-tune suppression levels based on specific diagnostic needs, and ensure compatibility with existing MRI protocols, utilizing standard GRE imaging sequences rather than requiring specialized acquisitions.135005-0001W001 4

[0019] The detailed description set forth below in connection with the drawings describes various configurations and does not represent the only configurations in which the concepts described herein may be practiced. The detailed description includes specific details for the purpose of providing a thorough understanding of various concepts. However, these concepts may be practiced without these specific details. In some instances, well known structures and components are shown in block diagram form in order to avoid obscuring such concepts.

[0020] While the aspects described in this specification generally refer to medical imaging obtained using magnetic resonance imaging (MRI), the teachings herein are applicable to a broad range of imaging modalities that utilize MRI technology. MRI medical images can include various imaging sequences such as gradient echo (GRE), spin echo (SE), and contrast-enhanced MRI, among others. These imaging sequences provide different contrasts and anatomical details, enabling precise visualization of softtissues, pathological abnormalities, and vascular structures. Accordingly , the term MRI medical images should be understood to encompass all MRI-based imaging techniques and sequences that may be used in conjunction with the disclosed systems and methods.

[0021] While the aspects described herein generally refer to MRI imaging of the brain, the teachings of this specification are applicable to imaging and analysis of any portion of the central nervous system (CNS). The brain comprises various anatomical structures, functional regions, and vascular networks, which may exhibit distinct imaging characteristics depending on the MRI modality used. The disclosed methods for vessel suppression and lesion enhancement are for detecting pathological abnormalities within the brain parenchyma, but similar approaches may also be applied to related neuroimaging tasks. As such, the term brain should be understood to encompass all intracranial regions, including the cerebrum, cerebellum, brainstem, and associated vasculature.

[0022] The teachings of this specification are also applicable to imaging and analysis of any body tissues where vessel suppression and lesion enhancement are beneficial. For example, similar techniques maybe applied in lung imaging to enhance the detection of pulmonary nodules by reducing vascular structures, or in abdominal imaging to improve the visualization of tumors in the liver or kidneys. Additionally, the systems and methods disclosed herein may be applicable in musculoskeletal imaging, where suppression of vascular signals can aid in the assessment of soft tissue abnormalities135005-0001W001 5or inflammatory conditions. Accordingly, the principles disclosed herein should be understood to encompass vessel suppression and lesion enhancement across a variety of MRI applications beyond brain imaging.

[0023] While the aspects described in this specification generally refer to lesions or pathological abnormalities in the brain, the teachings herein are applicable to a variety of neurological conditions thatmay manifest as abnormal tissue structures within MRI images. Such abnormalities can include, but are not limited to, tumors, metastases, ischemic or hemorrhagic strokes, multiple sclerosis plaques, and other neurodegenerative or inflammatory conditions. These abnormalities may exhibit contrast enhancement, signal hyperintensity, or structural distortions, depending on the imaging modality used. Accordingly, the term lesion or pathological abnormality in the brain should be interpreted broadly to encompass any detectable deviation from normal brain tissue thatmay be clinically relevantfordiagnosisortreatmentplanning

[0024] Several aspects of computer systems are presented with reference to various apparatus and methods. These apparatus and methods are described in the following detailed description and illustrated in the accompanying drawings by various blocks, components, circuits, processes, algorithms, etc. (collectively referred to as “elements”). These elements may be implemented using electronic hardware, computer software, or any combination thereof. Whether such elements are implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system.

[0025] By way of example, an element, or any portion of an element, or any combination of elements may be implemented as a “processing system” that includes one or more processors. When multiple processors are implemented, the multiple processors may perform the functions individually or in combination. Examplesof processors include microprocessors, microcontrollers, graphics processing units (GPUs), central processing units (CPUs), application processors, digital signal processors (DSPs), reduced instruction set computing (RISC) processors, systems on a chip (SoC), baseband processors, field programmable gate arrays (FPGAs), programmable logic devices (PLDs), state machines, gated logic, discrete hardware circuits, and other suitable hardware configured to perform the various functionality described throughout this disclosure. One or more processors in the processing system may execute software. Software, whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise, shall be construed broadly135005-0001W001 6to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software components, applications, software applications, software packages, routines, subroutines, objects, executables, threads of execution, procedures, functions, or any combination thereof.

[0026] Accordingly, in one or more example aspects, implementations, and / or use cases, the functions described may be implemented in hardware, software, or any combination thereof. If implemented in software, the functions may be stored on or encoded as one or more instructions or code on a computer-readable medium. Computer-readable media includes computer storage media. Storage media may be any available media that can be accessed by a computer. By way of example, such computer-readable media can include a random-access memory (RAM), a read-only memory (ROM), an electrically erasable programmable ROM (EEPROM), optical disk storage, magnetic disk storage, other magnetic storage devices, combinations of the types of computer- readable media, or any other medium that can be used to store computer executable code in the form of instructions or data structures that can be accessed by a computer.

[0027] While aspects, implementations, and / or use cases are describedin this application by illustration to some examples, additional or different aspects, implementations and / or use cases may come about in many different arrangements and scenarios. Aspects, implementations, and / oruse cases described herein may be implemented across many differingplatform types, devices, systems, shapes, sizes, and packaging arrangements. For example, aspects, implementations, and / or use cases may come about via integrated chip implementations and other non-module-component based devices (e.g., end-user devices, vehicles, communication devices, computing devices, industrial equipment, retail / purchasing devices, medical devices, artificial intelligence (Al)-enabled devices, etc.). While some examples may or may not be specifically directed to use cases or applications, a wide assortment of applicability of described examples may occur. Aspects, implementations, and / oruse cases may range a spectrum from chip-level or modular components to non-modular, non-chip- level implementations and further to aggregate, distributed, or original equipment manufacturer (OEM) devices or systems incorp oratingone or more techniques herein. In some practical settings, devices incorporating described aspects and features may also include additional components and features for implementation and practice of claimed and described aspect. For example, transmission and reception of wireless signals necessarily includes a number of components for analog and digital purposes135005-0001W001 7(e.g., hardware components including antenna, RF-chains, power amplifiers, modulators, buffer, processor(s), interleaver, adders / summers, etc.). Techniques described herein may be practiced in a wide variety of devices, chip-level components, systems, distributed arrangements, aggregated or disaggregated components, end-user devices, etc. of varying sizes, shapes, and constitution.

[0028] FIG. 1 is a diagram illustrating a magnetic resonance imaging (MRI) system 100, in accordance with various aspects of the present disclosure. The MRI system 100 may include a magnet system 103, a patient transport table 105 connected to the magnet system 103, and a controller 101 operably coupled to the magnet system 103. In one example, apatientmay lie on the patient transport table 105 and the magnet system 103 would pass around the patient and scanning for the target site (e.g., the brain of the patient). The controller 101 may control magnetic fields and radio frequency (RF) signals provided by the magnet system 103 and may receive signals from detectors in the magnet system 103. The MRI system 100 may further include a computer system 110 and one or more databases 120 operably coupled to the controller 101 over the network 130. The computer system 110 may be used for implementing the CE brain MRI images vessel suppression system 140. The computer system 110 may be used for generating a synthesized-SE generation model using training datasets. Although the illustrated diagram shows the controller 101 and computer system 110 as separate components, the controller 101 and computer system 110 can be integrated into a single component.

[0029] In some aspects, the controller 101 may be operated to provide the MRI sequence controller information about a pulse sequence and / or to manage the operations of the entire system, according to installed software programs. The controller 101 may also serve as an elementfor instructing a patient to perform tasks, such as, for example, a breath hold by a voice message produced using an automatic voice synthesis technique. The controller 101 may receive commands from an operator which indicate the scan sequence to be performed. The controller 101 may include various components such as a pulse generator module which is configured to operate the system components to carry out the desired scan sequence, producing data that indicate the timing, strength and shape of the RF pulses to be produced, and the timing of and length of the data acquisition window. Pulse generator module may be coupled to a set of gradient amplifiers to control the timing and shape of the gradient pulses to be produced during the scan. Pulse generator module also receives patient data from135005-0001W001 8a physiological acquisition controller that receives signals from sensors attached to the patient, such as electrocardiogram (ECG) signals from electrodes or respiratory signals from a bellows. Pulse generator module may be coupled to a scan room interface circuit which receives signals from various sensors associated with the condition of the patient and the magnet system. The controller 101 may also include or be operatively coupled with a patient positioning system which may receive commands through the scan room interface circuit to move the patient to the desired position for the scan.

[0030] The controller 101 may include a transceiver module which is configured to produce pulses which are amplified by an RF amplifier and coupled to RF coil by a transmit / receive switch. The resulting signals radiated by the excited nuclei in the patient may be sensed by the same RF coil and coupled through transmit / receive switch to a preamplifier. The amplified nuclear magnetic resonance (NMR) signals are demodulated, filtered, and digitized in the receiver section of transceiver. Transmit / receive switch is controlled by a signal from pulse generator module to electrically couple RF amplifier to coil for the transmit mode and to preamplifier for the receive mode. Transmit / receive switch may also enable a separate RF coil (for example, a head coil or surface coil, not shown) to be used in either the transmit mode or receive mode.

[0031] The NMR signals picked up by RF coil may be digitized by the transceiver module and transferred to a memory module coupled to the controller. The receiver in the transceiver module may preserve the phase of the acquired NMR signals in addition to signal magnitude. The down converted NMR signal is applied to an analog-to- digital (AID) converter (not shown) which samples and digitizes the analog NMR signal. The samples may be applied to a digital detector and signal processor which produces in-phase (I) values and quadrature (Q) values corresponding to the received NMR signal. The resulting stream of digitized I and Q values of the received NMR signal may then be employed to reconstruct an image.

[0032] The controller 101 may include one or more processing modules configured to acquire and process contrast-enhanced (CE) images following the administration of a gadolinium -based contrast agent (GBCA). The controller 101 may manage the injection timing and contrast agent distribution by synchronizing scan acquisition with physiological signals, such as cardiac or respiratory cycles, to optimize contrast uptake in the target tissue.135005-0001W001 9

[0033] Specifically, the controller 101 in connection with the magnet system 103 may further be configured to execute different MRI pulse sequences, including gradient echo (GRE) sequences and spin echo (SE) sequences, based on the selected imaging protocol. In GRE sequences, the controller 101 may control the application of variable flip angles and short repetition times (TR) to generate high-resolution images with T1 -weighted contrast and may manage the gradients to compensate for dephasing effects and control. In SE sequences, the controller 101 may manage the refocusing radiofrequency (RF) pulsesto compensate for dephasing effects and generate T1 , 12- weighted, or proton density -weighted images. The acquired image data may be stored in memory and processed using system algorithms for further image reconstruction and analysis. The provided contrast-enhanced brain MRI vessel suppression system may be used to process existing GRE images, selectively suppressing vessel signals while preserving lesion contrast, thereby enhancing visualization for improved diagnosis or treatment planning.

[0034] The controller 101 may include or be coupled to an operator console (not shown) which can include input devices (e.g., keyboard) and control panel and a display. For example, the controller may have input / output (V0) ports connected to an E0 device such as a display, keyboard and printer. In some cases, the operator console may communicate through the network 130 with the computer system 110 that enables an operator to control the production and display of images on a screen of display. The images may be MR images with improved quality generated according to a lesion visualization enhancement scheme. The lesion visualization enhancement scheme may be determined automatically by the CE brain MRI images vessel suppression system 140 and / or by a user as described later herein.

[0035] The MRI system 100 may include a user interface. The user interface may be configured to receive user input and output information to a user. The user input may be related to control of image acquisition. The user input may be related to the operation of the MRI system (e.g., certain threshold settings for controlling program execution, parameters for controllingthe joint estimation of coil sensitivity and image reconstruction, etc.). The user input may be related to various operations or setting; about the lesion visualization enhancement. The user input may include, for example, a selection of a target region for vessel suppression, adjustment of the suppression factor for controlling the degree of vessel attenuation, selection of contrast enhancement settings for lesion visualization, customization of image display135005-0001W001 10parameters for optimized diagnostic review, selection of an MRI sequence type for processing (e.g., GRE or SE), configuration of vessel suppression processing options, other user-defined parameters relatedto image enhancement and analysis, and various others. The user interface may include a screen such as a touch screen and any other user interactive external device such as handheld controller, mouse, joystick, keyboard, trackball, touchpad, button, verbal commands, gesture-recognition, attitude sensor, thermal sensor, touch capacitive sensors, foot switch, or any other device.

[0036] The MRI system 100 may include computer systems 110 and database 120, which may interact with the controller 101. The computer system 110 may include a laptop computer, a desktop computer, a central server, distributed computing system, etc. The processor may be a hardware processor such as a central processing unit (CPU), a graphic processing unit (GPU), a general-purpose processing unit, which can be a single core or multi core processor, a plurality of processors for parallel processing in the form of fine-grained spatial architectures such as a field programmable gate array (FPGA), an application-specific integrated circuit (ASIC), and / or one or more Advanced RISC Machine (ARM) processors. The processor can be any suitable integrated circuits, such as computingplatforms or microprocessors, logic devices and the like. Although the disclosure is described with reference to a processor, other types of integrated circuits and logic devices are also applicable. The processors or machines may not be limited by the data operation capabilities. The processors or machines may perform 512 bit, 256 bit, 128 bit, 64 bit, 32 bit, or 16 bit data operations. Details regarding the functional aspects of computer system 110 are described with respect to FIG. 2 and details regarding the hardware aspects of computer system 110 are described with respect to FIG. 8.

[0037] The MRI system 100 may include one or more databases 120. The one or more databases 120 may utilize any suitable database techniques. For instance, structured query language (SQL) or "NoSQL" database may be utilized for storing MR image data, raw image data, synthesized image data, training datasets, trained model, parameters of an acquisition scheme, etc. Some of the databases may be implemented using various standard data-structures, such as an array, hash, (linked) list, struct, structured text file (e.g., XML), table, ISON, NOSQL and / or the like. Such data structures may be stored in memory and / or in (structured) files. In another alternative, an object-oriented database may be used. Object databases can include a number of object collections that are grouped and / or linked together by common attributes; they135005-0001W001 11may be related to other object collections by some common attributes. Object-oriented databases perform similarly to relational databases with the exception that objects are not just pieces of data but may have other types of functionalities encapsulated within a given object. If the database of the present disclosure is implemented as a data- structure, the use of the database of the present disclosure may be integrated into another component such as the component of the present technical solutions. Also, the database may be implemented as a mix of data structures, objects, and relational structures. Databases may be consolidated and / or distributed in variations through standard data processing techniques. Portions of databases, e.g., tables, may be exported and / or imported and thus decentralized and / or integrated.

[0038] The network 130 may establish connections among the components in the MRI platform and a connection of the magnet system 103 to external systems. The network 130 may include any combination of local area and / or wide area networks using both wireless and / or wired communication systems. For example, the network 130 may include the Internet, mobile telephone networks, as well as other communication technologies. In one aspect, the network 130 may use standard communications technologies and / or protocols. Hence, the network 130 may include links using technologies such as Ethernet, Bluetooth™ (Bluetooth is a trademark of the Bluetooth Special Interest Group (SIG)), Wi-Fi™ (Wi-Fi is a trademark of the Wi-Fi Alliance) based on the Institute of Electrical and Electronics Engineers (IEEE) 802.11 standard, LTE, or NR, worldwide interoperability for microwave access (WiMAX), 2G / 3G / 4G / LTE / 5G, 6G or any future mobile communications protocols, asynchronoustransfermode (ATM), InfiniBand, PCI Express Advanced Switching etc. Other networking protocols used on the network 130 can include multiprotocol label switching (MPLS), the transmission control protocol / Internet protocol (TCP / IP), the User Datagram Protocol (UDP), the hypertext transport protocol (HTTP), the simple mail transfer protocol (SMTP), the file transfer protocol (FTP), and the like. The data exchanged over the network can be represented using technologies and / or formats including image data in binary form (e.g., Portable Networks Graphics (PNG)), the hypertext markup language (HTML), the extensible markup language (XML), etc. In addition, all or some of links can be encrypted using conventional encryption technologies such as secure sockets layers (SSL), transport layer security (TLS), Internet Protocol security (IPsec), etc. In another embodiment, the entities on135005-0001W001 12the network can use custom and / or dedicated data communications technologies instead of, or in addition to, the ones described above.

[0039] FIG. 2 is a diagram illustrating a CE brain MRI images vessel suppression system 200, in accordance with various aspects of the present disclosure. The CE brain MRI images vessel suppression system 200 may include the CE brain MRI images vessel suppression module 240, which may correspond to the CE brain MRI images vessel suppression module 140 as described in FIG. 1. The CE brain MRI images vessel suppression module 240 may include multiple components, including but not limited to, a synthesized-SE generation model training sub-module 202, a synthesized-SE generation sub-module 204, a vessel map generation sub-module 206, a vessel- suppressed GRE generation sub-module 208, and a user interface sub-module 210.

[0040] It is understood that in some aspects, the synthesized-SE generation model training sub-module 202 may be hosted on a separate device from the CE brain MRI vessel suppression module 240. In such cases, the CE brain MRI images vessel suppression module 240 may include a synthesized-SE generation sub-module 204, a vessel map generation sub-module 206, a vessel-suppressed GRE generation sub-module 208, and a user interface sub-module 210, and other applicable modules. The CE brain MRI images vessel suppression system 200 may receive a pre-trained image-to-image synthesis model 203 from the synthesized-SE generation model training sub-module 202, allowing it to generate synthesized SE images from GRE images without performing model training locally.

[0041] In some aspects, the synthesized-SE generation model training sub-module 202 may be configured to obtain and manage training datasets used for model training. For example, the synthesized-SE generation model training sub-module 202 may be configured to train an image-to-image synthesis model 203 based on paired training data 201 received from a database 120. The image-to-image synthesis model 203 may be configured to generate a second-type (e.g., the SE) medical image from a first-type (e.g., the GRE) medical image. Specifically, the image-to-image synthesis model 203 may be trained to synthesize SE images of a patient’s brain based on GRE images of the same patient’s brain. The training data 201 may include a training set of GRE images of patient’s brain and one or more corresponding ground-truth second-type (e.g., the SE) images of the patient’s brain.

[0042] In some aspects, when preparing the training data 201, the training process may involve co-registering the training set of first-type medical images (e.g., GRE135005-0001W001 13sequence) with the corresponding ground-truth third-type medical images (e.g., SE sequence), performing intensity normalization to standardize intensity values across different scans, and applying histogram equalization to enhance contrast consistency between training images (e.g., match the gray matter and white matter intensity between the two image sequences).

[0043] When training the image-to-image synthesis model 203, one or more training synthesized third-type medical images (e.g., SE sequence) may be generated using the image-to-image synthesis model based on the training set of first-type medical images (e.g., GRE sequence). For example, the model parameters of the image-to- image synthesis model 203 may be iteratively adjusted based on the comparison between the one or more training synthesized SE images and the one or more corresponding ground-truth SE images. For example, when performing the comparison at least one of the following loss functions may be considered:1. LI loss, corresponding to an absolute pixel-wise difference between the training synthesized SE images and the corresponding ground-truth SE images,2. structural similarity index measure (SSIM) loss, corresponding to a structural similarity metric between the training SE images and the corresponding ground-truth SE images, or3. perceptual loss, corresponding to a feature-space loss between the training synthesized SE images and the corresponding ground-truth SE images, computed using feature representations extracted from a pre-trained deep neural network.

[0044] Upon completion of training, the image-to-image synthesis model 203 may be refined to include one or more updated model configuration parameters, such as weight parameters and bias values optimized based on the training loss functions (e.g., LI loss, SSIM loss, and perceptual loss), feature extraction parameters that enhance vessel suppression accuracy while preserving lesion contrast, normalization factors adjusted to account for variations in MRI intensity distributions across different datasets, and hyperparameters, including learning rate adjustments, optimization strategies, or regularization factors, to improve model generalization and inference performance. These refined parameters may improve the synthesized SE image quality, leading to more accurate vessel map generation and vessel-suppressed GRE image production in later steps for enhanced lesion visualization.135005-0001W001 14

[0045] Because the image-to-image synthesis model 203 is trained using paired gradient echo (GRE) and spin echo (SE) images / sequences, it enablesthe generation of synthesized SE images from readily available GRE images, without requiring additional SE acquisitions. Since GRE sequences are commonly used in standard MRI protocols and are faster to acquire, this approach eliminates the need for specialized SE sequences, thereby reducing scan time and making vessel suppression more accessible in clinical settings.

[0046] In some aspects, the image-to-image synthesis model 203 may include one or more deep learning modules, such as a Residual U-Net (Res-UNet) model, such as a 2.5D Res-UNet, as well as other neural networks or machine learning models. As a nonlimiting example, the deep learning model maybe trained to generate synthesized SE images from GRE images, enabling SE-like contrast enhancement without requiring additional SE acquisitions. The neural networks used in the model may be based on a large collection of artificial neurons, where each neuron is connected to others in a layered structure to learn hierarchical image representations. These networks may employ nonlinear activation functions, convolutional operations, and residual connections to capture structural and intensity differences between GRE and SE sequences. In some aspects, neural network models may utilize b ackpropagation and gradient-based optimization to refine their internal parameters, improving the accuracy of synthesized SE images over iterative training cycles.

[0047] In some aspects, as stated above, when training the image-to-image synthesis model 203, the synthesized-SE generation model training sub-module 202 may update configurations such as weights, biases, and hyperparameters based on iterative assessments of model predictions. The training data 201 which includes GRE images and corresponding ground-truth SE images may be received from database 120. The training process may involve data preprocessing steps on the training data 201 such as co-registering the SE images with the corresponding GRE images to align spatial features, performing intensity normalization to standardize brightness levels across scans, and applying histogram equalization to enhance contrast consistency between GRE and SE images.

[0048] For example, FIG.3 is a diagram 300 illustrating an example process for training a machine learning model 310, in accordance with various aspects of the present disclosure. The machine learning model 310 may correspond to the image-to-image synthesis model 203 described in FIG.2. In some aspects, the machine learningmodel135005-0001W001 15310 may take inputs 302, which include training sets of GRE images and one or more corresponding ground-truth SE images from the training data 201. The machine learning model 310 may process these inputs 302 and generates outputs 304, which may be synthesized SE images. In some aspects, the image-to-image synthesis model 203 may be trained using the same or a similar approach as illustrated in FIG. 3, implemented within the synth esized-SE generation model training sub-module 202.

[0049] In some aspects, the outputs 304 may be fed back into the machine learning model 310 as active feedback, refining the training process. This feedback loop allows the model to continuously improve its ability to generate synthesized SE images by comparing the synthesized SE images with the corresponding ground-truth SE images. In some aspects, the training process may also incorporate additional feedback, such as user indications of accuracy, reference lab els associated with the inputs 302, or other reference feedback information that guides the model’s optimization. In some aspects, the machine learning model 310 may update its configuration parameters (e.g., weights, biases, and optimization hyperparameters) based on the comparison of its predictions (outputs 304) with the reference feedback. Additionally, if the machine learning model 310 is a neural network, the training process may involve backpropagation, where error signals propagate backward through the network layers to refine connection weights. These updates may be based on the magnitude of the error, adjusting the model parameters to reduce prediction differences over successive iterations.

[0050] In some aspects, to improve training performance and ensure that the synthesized SE images closely resemble ground-truth SE images, as stated above, the model may be optimized using a combination of loss functions, including one or more of LI loss, which minimizes the absolute pixel-wise difference between the synthesized SE images and the ground-truth SE images, SSIM loss, which preserves structural integrity by ensuring similarity between the synthesized SE images and ground-truth SE images, and perceptual loss, which leverages deep feature representations from a pre-trained network to ensure the synthesized SE images maintain realistic textures and intensity distributions. These loss functions collectively guide the training process, ensuringthatthe model accurately synthesizes SEimages from GREimages.

[0051] It is understood that the types of learning models disclosed herein are for illustrative purposes only. Any machine learning model capable of generating one type of medical image from another may be used or trained for application in the described135005-0001W001 16system. For example, while the aspects herein reference a 2.5D Res-UNet, other architectures such as Generative Adversarial Networks (GANs), Transformer-based models, Variational Autoencoders (VAEs), or other deep learning-based image-to- image translation models may also be implemented. The selected model may be optimized based on data availability, computational efficiency, and performance metrics suitable for the given medical imaging task. In some aspects, the synthesized- SE generation model training sub-module 202 may train a model off-line. Alternatively or additionally, the synthesized-SE generation model training submodule 202 may use real-time data as feedback to refine the model.

[0052] It is also understood that the input 302 to the machine learning model 310 may include ID, 2D, 2.5D, or 3D data, depending on the selected model architecture and training approach. For example, ID data may represent extracted signal intensity profiles or feature vectors derived from GRE images. 2D data may consist of individual GRE image slices used for slice-by-slice processing. 2.5D data may include sequences of GRE images, incorporating contextual information from adjacent slices to enhance spatial awareness. 3D datamay comprise volumetric GRE datasets, preserving spatial relationships across multiple slices for full 3D reconstructions. Similarly, the output 304 may be generated as 2D, 2.5D, or 3D data, where 2D outputs correspond to synthesized SE images generated for single slices, 2.5D outputs leverage sequential GRE slices to generate contextually enhanced SE images that improve anatomical consistency, and 3D outputs reconstruct a full volumetric SE dataset, facilitating comprehensive clinical interpretation and analysis. The choice of model architecture may depend on factors such as computational constraints, accuracy requirements, and dataset availability, ensuring optimal performance for vessel suppression and lesion enhancement tasks.

[0053] Referring back to FIG. 2, based on the trained image-to-image synthesis model 203 the synthesized-SE generation sub-module 204 may generate synthesized SE images of a patient’s brain (referred as “synthesized SE image 207” hereinafter) based on GRE images of the same patient’s brain (referred as “GRE images 205” hereinafter) store in database 120.

[0054] For example, FIG. 4 is a diagram 400 illustrating an example of vessel suppression GRE scheme using CE brain MRI images vessel suppression system 240, in accordance with various aspects of the present disclosure. For example, as shown in FIG. 4, at 402, based on the trained image-to-image synthesis model 203, the135005-0001W001 17synthesized SE images 207 maybe generated based on one or more GRE images 205, stored in the database 120.

[0055] Referring back to FIG. 2, in some aspects, the vessel map generation sub-module 206 may generate vessel maps of the patient’s brain (referred to as “vessel map 209” hereinafter) based on the GRE images 205 and the synthesized SE images 207. In some aspects, the GRE images 205 may be received from the database 120. Additionally or alternatively, the GRE images 205 may be received from the synthesized-SE generation sub-module 204.

[0056] For example, as shown in FIG. 4, at 404, the vessel map 209 may be generated by computing a difference between the GRE images 205 and the synthesized SE images 207. In some aspects, the difference computation may involve at least one of:1. Pixel-wise subtraction (GRE - SE), which highlights vessel signals while suppressing other anatomical structures,2. Applying spatial or frequency -based filters, such as high -pass or vessel enhancement filters, to enhance vessel structures while attenuating background signals, or3. Other suitable approaches, such as learning-based segmentation techniques, that isolate vessel signals using deep learning models trained to distinguish vascular features.Since the image-to-image synthesis model 203 (e.g., a GRE2SE network) wastrained on datasets containing various lesions and metastases, it learned to reconstruct pathological abnormalities realistically while simultaneously eliminating vessel signals. As a result, the computed vessel map 209 may isolate vessel structures, effectively distinguishing them from underlying lesions.

[0057] Referring back to FIG. 2, in some aspects, the vessel map 209 may be utilized in the vessel-suppressed GRE generation sub-module 208 to generate vessel-suppressed GRE images 211. For example, the vessel-suppressed GRE generation sub-module 208 may selectively reduce vessel signals in the GRE images 205 while preserving lesion contrast, improving lesion visibility in contrast-enhanced MRI scans. In some aspects, the GRE images 205 may be received from the database 120. Additionally or alternatively, the GRE images 205 may be received from one or more of the synthesized-SE generation sub-module 204 or the vessel map generation sub-module 206.135005-0001W001 18

[0058] For example, as shown in FIG. 4, at step 406, the vessel-suppressed GRE images 211 may be generated by applying a configurable suppression factor to the GRE images 205 based on the intensity values in the vessel map 209. In some aspects, the suppression process may involve:1. linearly scaling down vessel signal intensities using the configurable suppression factor, and / or2. using the vessel map 209 as a mask, where vessel signals on the GRE images 205 are selectively attenuated to obtain the final vessel-suppressed GRE image (e.g., vessel-suppressed GRE image GRE vsup).

[0059] FIG. 5 is a diagram 500 illustrating qualitative results of vessel suppression in accordance with various aspects of the present disclosure. As shown in FIG. 5, the original GRE image 502 (which may correspond to GRE images 205) clearly shows both lesion regions 511 and vessel regions 513. By applying different suppression factors (e.g., 25%, 50%, 75%, 100%), the visibility of lesion regions 511 remains preserved, while vessel regions 513 are progressively suppressed. Specifically, as shown in FIG. 5 :1. At 25% suppression (504), vessel signals are slightly reduced while lesion contrast is unaffected.2. At 50% suppression (506), vessels become less prominent while lesion visibility remains clear.3. At 75% suppression (508), vessel signals are further minimized without significant loss of lesion contrast.4. At 100% suppression (510), vessel structures are about entirely removed, isolating the lesion regions for enhanced diagnostic assessment.The configurable suppression factor allows for adjustable vessel attenuation, enabling radiologists to fine-tune vessel suppression levels based on clinical needs while ensuring that lesion features remain unaffected for accurate diagnosis. Unlike the existing medical images, such as GRE or SE, can either fully display the vessel signals (GRE images: 0% suppression) or eliminate the vessel signals (SE images: 100% suppression), by adjusting the sequence, the technical solutions disclosed herein introduce an intermediate suppression capability, allowing vessels to be partially removed while retaining lesion contrast, offering a level of adjustability not available in standard GRE or SE imaging. This provides a flexible imaging approach where135005-0001W001 19vessel visibility can be dynamically modulated based on the specific diagnostic requirements for each patient.

[0060] It is understood that the suppression factor values discussed herein are for illustrative purposes only. Any value between 0% and 100% may be applied, depending on the desired level of vessel suppression. The optimal suppression level maybe determined based on clinical preferences, pathology type, imaging conditions, or diagnostic workflow requirements. Additionally, suppression may be applied globally or locally, enabling selective vessel attenuation in specific regions of interest while maintaining full vascular visibility in other areas as needed.

[0061] Referring back to FIG. 2, the user interface sub-module 210 may render a graphical user interface (GUI) or user interface (e.g., referred as “GUI 241”) allowing a user to select an acquisition scheme, modify one or more parameters of an acquisition scheme, viewing information related to imaging and acquisition settings and the like. The GUI 241 may show graphical elements that permit a user to view or access information related to image acquisition. In some aspects, the GUI 241 may have various interactive elements such as buttons, textboxes andthe like, which may allow a user to provide input commands or contents by directly typing, clicking or dragging such interactive elements. For example, a user may manually create or modify a scanning pattern, select the suppression factor values, and setotherparametersviathe GUI, as discussed herein.

[0062] In some cases, the GUI 241 may be provided on a display 235. The display 235 may or may not be a touchscreen. The display 235 may be a light-emitting diode (LED) screen, organic light-emitting diode (OLED) screen, liquid crystal display (LCD) screen, plasma screen, or any other type of screen. The display 235 may be configured to show a user interface (UI) or a graphical user interface (GUI) rendered through an application (e.g., via an application programming interface (API) executed on the local computer system or on the cloud).

[0063] The CE brain MRI images vessel suppression system 200 may be implemented in software, hardware, firmware, embedded hardware, standalone hardware, application specific-hardware, or any combination of these. The CE brain MRI images vessel suppression system, modules, components, algorithms and techniques may include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, coupled to receive data and135005-0001W001 20instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device. These computer programs (also known as programs, software, software applications, or code) may include machine instructions for a programmable processor, and may be implemented in a high-level procedural and / or object-oriented programming language, and / or in assembly / machine language. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus, and / or device (such as magnetic discs, optical disks, memory, orProgrammableLogic Devices (PLDs)) used to provide machine instructions and / or data to a programmable processor. The CE brain MRI images vessel suppression system 200 may be a standalone system that is separate from the MR imaging system. Alternatively or in addition to, the CE brain MRI images vessel suppression system 200 may be integral to the MR imaging system such as a component of a controller of the MR imaging system.

[0064] In some cases, the CE brain MRI images vessel suppression system 200 may employ an edge intelligence paradigm that data processing and MR image enhancement is performed at the edge or edge gateway (MRI system). In some instances, machine learning model may be built, developed and trained on a cloud / data center and run on the MRI system. For example, one or more of the synthesized-SE generation model training sub-module 202, the synthesized-SE generation sub-module 204, the vessel map generation sub-module 206, the vessel-suppressed GRE generation sub-module 208, or the user interface sub-module 210 may run on the edge, the cloud, or onpremises environment.

[0065] Performing the technical solutions disclosed herein enhances diagnostic accuracy, particularly in cases involving small lesions, where distinguishing between blood vessels and metastases is challenging. Furthermore, the adjustable suppression factor allows for customizable vessel attenuation, giving radiologists flexibility in balancing vessel suppression and lesion visibility according to clinical needs.

[0066] FIG. 6 is a diagram 600 illustrating qualitative and quantitative comparisons of contrast-enhanced brain MRI images across different imaging conditions, in accordance with various aspects of the present disclosure. As shown in FIG. 6, precontrast 602 represents baseline GRE images before contrast agent administration, standard-dose 604 represents standard post-contrast GRE images, contrast boost 606 represents GRE images with enhanced contrast, emphasizing lesion visibility,135005-0001W001 21standard-vsup 608 represents GRE images processed with standard vessel suppression, reducing vessel signals while preserving lesion contrast, and Cb vsup (contrast boost + vessel suppression) 610 represents GRE images with both contrast enhancement and vessel suppression, aiming to optimize lesion visibility while minimizing vessel artifacts. Each image highlights specific regions of interest (ROIs) using colored boxes for quantitative evaluation, as described in the study.

[0067] For example, a total of 45 patient studies were analyzed to assess whether lesion contrast remains intact while vessel signals are suppressed. To achieve this, three ROIs were selected in each case:1. ROI L (Green box) - Largest enhancing lesion or tumor,2. ROI V (Red box) - Largest enhancing vessel, and3. ROI Brain - Non-enhancing parenchymal region.

[0068] Three key quantitative metrics were computed to evaluate the performance of vessel suppression:1. Contrast-to-Noise Ratio (CNR) which measures the contrast difference between the lesion and the surrounding brain tissue, normalized by noise:2. Lesion-to-Brain Ratio (LBR) which evaluates the relative lesion intensity compared to normal brain tissue:>3. Contrast Enhancement Percentage (CEP) which measures the percentage increase in lesion intensity post-contrast, comparingpre- and post-contrast scans:<where S / Zes(onis the mean intensity inside ROILor ROIVof a post-contrast GRE or GREvsupimage, SIpreiesionis the mean intensity inside ROILor ROIvof the pre-contrast GRE image, SIbrainand SDbrainare the mean and standard deviation inside ROIbrainof the respective images.These metrics (e.g., the CNR, the LBR, and the CEP) were computed forboth / ?0 / L(lesions) and ROIV(vessels) across standard-of-care (SOC) GRE images and GREvsup135005-0001W001 22(Vessel-Suppressed GRE images) to evaluate how vessel suppression impacts lesion visibility.

[0069] As shown in Table 1 below, the quantitative evaluation demonstrates the effects of vessel suppression on lesion and vessel regions:1. Lesion Regions (ROI L):i. The CNR and LBR remain high after vessel suppression, confirming that lesion contrast is preserved.ii. The CEP increases slightly (138.85% — > 140.12%), indicating that vessel suppression has minimal impacton contrast enhancement in lesion regions.2. Vessel Regions (ROI V):i. The CNR drops significantly (8.15 — 2.76), confirmingthatvessel signals are effectively suppressed.ii. The LBR also decreases (2.313 — > 0.811), further demonstrating that vessel intensities are reduced.iii. The CEP decreases from 84.62% to 32.27%, highlighting the suppression of contrast enhancement in vessel regions, ensuring that unwanted vascular signals are attenuated while lesion contrast remains unaffected.Table 1The results validate that the vessel suppression method selectively reduces vascular signals while maintaining lesion contrast, providing a more clinically relevant visualization for enhanced diagnostic accuracy.

[0070] FIG. 7 A is a flowchart of a method 700 of contrast enhancedbrain MRI images vessel suppression, in accordance with various aspects of the present disclosure. The method 700 may be performed by a CE brain MRI images vessel suppression system which may correspond to the computer system 110 in FIG. 1, the CE brain MRI images vessel suppression system 200 in FIG. 2, or the computer system 800 in the hardware implementation of FIG. 8. The method 700 enhances diagnostic accuracy, particularly135005-0001W001 23in cases involving small lesions, where distinguishing between blood vessels and metastases is challenging. Furthermore, the adjustable suppression factor allows for customizable vessel attenuation, giving radiologists flexibility in balancing vessel suppression and lesion visibility according to clinical needs.

[0071] At 710, the CE brain MRI images vessel suppression system may obtain a set of first- type medical images of a subject, the first-type medical images (e.g., one or more of GRE images of a patient’s brain) containing a first feature and a second feature. FIG.2 illustrates an example of the CE brain MRI images vessel suppression system receive the GRE images 205, at 204. In some aspects, 710 may be performed by the processor circuitry 824, the CEbrain MRI images vessel suppression component 840, the transceiver 822, and / or the antenna 880 in FIG. 8.

[0072] At 720, the CE brain MRI images vessel suppression system may generate one or more second-type medical images (e.g., the vessel-suppressed GRE images) based on the set of first-type medical images by selectively reducing signals corresponding to the second feature. FIG. 2 illustrates an example of the CEbrain MRI images vessel suppression system generating the vessel-suppressed GRE images 211, at 210. In some aspects, 720 may be performed by the processor circuitry 824, the CE brain MRI images vessel suppression component 840, the transceiver 822, and / or the antenna 880 in FIG. 8.

[0073] FIG. 7B is a flowchart of method 750, in accordance with various aspects of the present disclosure. The method 700 may be performed by a CE brain MRI images vessel suppression system which may correspond to the computer system 110 in FIG.1 , the CE brain MRI images vessel suppression system 200 in FIG. 2, or the computer system 800 in the hardware implementation of FIG. 8. The method 750 enhances diagnostic accuracy, particularly in cases involving small lesions, where distinguishing between bloodvessels and metastases is challenging. Furthermore, the adjustable suppression factor allows for customizable vessel attenuation, giving radiologists flexibility in balancing vessel suppression and lesion visibility according to clinical needs. Some aspects of FIG. 7B may be similar to the aspects of FIG. 7A and are shown with the same reference number.

[0074] At 710, the CEbrain MRI images vessel suppression system may obtain a set of first- type medical images of a subject, the first-type medical images (e.g., one or more of GRE images of a patient’ s brain) containing a first feature and a second feature. FIG.2 illustrates an example of the CE brain MRI images vessel suppression system135005-0001W001 24receive the GRE images 205, at 204. In some aspects, 710 may be performed by the processor circuitry 824, the CEbrain MRI images vessel suppression component 840, the transceiver 822, and / or the antenna 880 in FIG. 8.

[0075] At 704, the CE brain MRI images vessel suppression system may generate based on the set of first-type medical images, one or more synthesized third-type medical images, using an image-to-image synthesize model, wherein in the one or more synthesized third -type medical images, the second feature is attenuated while the first feature is retained. FIG. 2 illustrates an example of the CE brain MRI images vessel suppression system generating the synthesized SE images 207, at 204. In some aspects, 704 may be performed by the processor circuitry 824, the CE brain MRI images vessel suppression component 840, the transceiver 822, and / or the antenna 880 in FIG. 8.

[0076] At 706, the CE brain MRI images vessel suppression system may generate one or more fourth-type medical images by determining a difference between the set of first- type medical images and the one or more synthesized third-type medical image. FIG.2 illustrates an example of the CE brain MRI images vessel suppression system generating the vessel maps 209 at 210. In some aspects, 706 may be performed by the processor circuitry 824, the CEbrain MRI images vessel suppression component 840, the transceiver 822, and / or the antenna 880 in FIG. 8.

[0077] At 720, the CE brain MRI images vessel suppression system may generate one or more second-type medical images (e.g., the vessel-suppressed GRE images) based on the set of first-type medical images by selectively reducing signals corresponding to the second feature. FIG. 2 illustrates an example of the CEbrain MRI images vessel suppression system generating the vessel-suppressed GRE images 211, at 210. In some aspects, 720 may be performed by the processor circuitry 824, the CE brain MRI images vessel suppression component 840, the transceiver 822, and / or the antenna 880 in FIG. 8.

[0078] At 722, the one or more second-type medical images may be displayed on a display device. FIG. 2 illustrates an example of the CEbrain MRI images vessel suppression system displaying the vessel-suppressed GRE images 211, at display 235 (e.g., facilitated by the GUI 241). In some aspects, 722 may be performed by the processor circuitry 824, the CE brain MRI images vessel suppression component 840, the transceiver 822, and / or the antenna 880 in FIG. 8.135005-0001W001 25

[0079] In some aspects, generating the one or more second-type medical images from the set of first-type medical images includes applying a configurable suppression factor to the set of first-type medical images based on intensity values in the one or more fourth -type medical images, and / or linearly scaling down the signals corresponding to the second feature using the configurable suppression factor.

[0080] In some aspects, the subject corresponds to a brain of a patient; the first feature corresponds to a lesion or pathological abnormality in the brain; and the second feature corresponds to a blood vessel or vascular structure in the brain.

[0081] In some aspects, the set of first-type medical images comprises one or more gradient echo (GRE) medical images; and the second-type of medical images comprises a vessel-suppressed GRE image.

[0082] In some aspects, the first-type medical images and third-type medical images are obtained using a magnetic resonance imaging (MRI) device; and the fourth-type medical image contains signals corresponding to the second feature.

[0083] In some aspects, the synthesized third-type medical image corresponds to a spin echo (SE) medical image; and the fourth-type medical image corresponds to a vessel map.

[0084] In some aspects, the image-to-image synthesize model is trained by obtaining, from a database, a plurality pairs of training data comprising: a training set of first-type medical images; and one or more corresponding ground-truth third-type medical images; generating one or more training synthesized third -type medical images using the image-to-image synthesize model based on the training set of first-type medical images; and adjusting at least one parameter of the image-to-image synthesize model based on the one or more training synthesized third-type medical images and the one or more corresponding ground-truth third-type medical images.

[0085] In some aspects, the training set of first-type medical images comprises one or more gradient echo medical images; and the one or more training synthesized third-type medical images are spin echo (SE) medical image synthesized based on the training set of first-type medical images.

[0086] In some aspects, obtaining, from a database, the plurality pairs of training data comprising: co-registering the training set of first-type medical images with the one or more corresponding ground-truth third-type medical images; performing intensity normalization; and applying histogram equalization.

[0087] In some aspects, adjusting at least one parameter of the image-to-image synthesize model based on the one or more training synthesized third-type medical images and135005-0001W001 26the one or more corresponding ground-truth third-type medical images further comprises: comparingthe one or more training synthesized third -type medical images and the one or more corresponding ground-truth third-type medical images based on at least one of: a LI loss, corresponding to an absolute pixel-wise difference between the one or more training synthesized third-type medical images and the one or more corresponding ground-truth third-type medical images; a structural similarity index measure (SSIM) loss, correspondingto a structural similarity between the one or more training synthesized third-type medical images and the one or more corresponding ground-truth third-type medical images; or a perceptual loss correspondingto featurespace loss between the one or more training synthesized third-type medical images and the one or more corresponding ground-truth third-type medical images. FIG. 2 illustrates an example of the CE brain MRI images vessel suppression system training the image-to-image synthesis model 203, at 202. Additional training details are also illustrated in FIG. 3.

[0088] FIG. 8 is a block diagram of an embodiment of a computer system 800, which can be utilized in accordance with various aspects of the present disclosure. In some aspects, the computer system 800 may correspond to the CE brain MRI images vessel suppression system 200 in FIG. 1 and FIG. 2, or to the CE brain MRI images vessel suppression system in FIG. 7A and / or FIG. 7B. The CE brain MRI images vessel suppression controller 804 may include a CE brain MRI images vessel suppression component 840 (e.g., correspondingto the CE brain MRI images vessel suppression module 240 in FIG. 2, FIG. 3, and FIG. 4) as a component of processor circuitry 824. The CE brain MRI images vessel suppression module 240 may be configured to obtain GRE images for a brain of a patient, synthesize SE images based on the GRE images, generate vessel map(s) based on the GRE images and the synthesized SE images, and vessel-suppressed GRE images based on the GRE images and the vessel maps, as described in connection with the signal flow in FIGs. 2 and 4, and / or including aspects described in connection with any of the flowcharts in FIG. 7A and / or FIG. 7B. In some aspects, the CEbrainMRI images vessel suppression module 240 may also be configured to train the image-to-image synthesis model 203 described in connection with the signal flow in FIGs. 2 and 3.

[0089] As shown, the controller 804 may include memory 826, a power supply 830, and a system bus 812 connecting the various controller components, including the memory 826 (or memory circuitry) associated with at least one processor (e.g., processor135005-0001W001 27circuitry 824). The system bus may include any of a bus memory or bus memory controller, a peripheral bus, and / or a local bus that is able to interact with any other bus architecture. Examples of processors include microprocessors, microcontrollers, graphics processing units (GPUs), central processing units (CPUs), application processors, digital signal processors (DSPs), reduced instruction set computing (RISC) processors, systems on a chip (SoC), baseband processors, field programmable gate arrays (FPGAs), programmable logic devices (PLDs), state machines, gated logic, discrete hardware circuits, and other suitable hardware configured to perform the various functionality described throughout this disclosure. One or more processors in the processing system may execute software. Software, whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise, shall be construed broadly to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software components, applications, software applications, software packages, routines, subroutines, objects, executables, threads of execution, procedures, functions, or any combination thereof.

[0090] In various aspects or examples, the functions described herein (e.g., including the aspects described in connection with FIGS. 2, 3, 4 and / or any of the flowcharts in FIG. 7A and / or FIG. 7B) may be implemented in hardware, software, or any combination of hardware and software. If implemented in software, the functions may be stored on or encoded as one or more instructions, which may be referred to as code, on a computer-readable medium (e.g., a non-transitory computer-readable medium). Computer-readable media includes computer storage media. Storage media may be any type of media thatis able to store data in a form readableby a computer. Examples of computer-readable media may include random-access memory (RAM), read-only memory (ROM), electrically erasable programmable ROM (EEPROM), solid state drives, flash memory cards, digital disks, optical disk storage, magnetic disk storage, other magnetic storage devices, combinations of various types of computer-readable media, or any other medium that can be used to store computer executable code in the form of instructions or data structures that can be accessed by a processing system or computer.

[0091] A basic input / output system (BIOS) may store the basic procedures for transfer of information between elements of the controller. In some aspects, the processor circuitry may include memory. The processor circuitry 824 is configured to provide135005-0001W001 28general processing, including the execution of software stored on the computer- readable medium / memory. The software, when executed by the processor circuitry, causes the CE brain MRI images vessel suppression controller to provide a series of vessel suppression processes, as described in connection with FIGS. 2 and 4, or the model training, as described in connection with FIGS. 2 and 3. In some aspects, the processor may be configured to perform aspects described in connection with any of the flowcharts in FIG. 7A and / or FIG. 7B.

[0092] A user of the controller may enter commandsand information usinginput components such as control buttons, a keyboard, a mouse, a touch panel, or any other input device known to those of ordinary skill in the art, such as, but not limited to, a microphone, joystick, game controller, scanner, etc. Such input devices may plug into the controller 804 through a serial port, which connected to the system bus. The input devices may be incorporated into the controller or may be connected in other ways, such as, without limitation, via a parallel port, a game port, or a universal serial bus (USB). A screen 810 or display device may also be included in, or connected to, the controller 804. In addition, the controller may be equipped with other peripheral output devices (not shown), such as audio speakers, a printer, etc.

[0093] The controller 804 may include a connection component that may include a communication interface (e.g., which may include one or more transceivers 822 and / or one or more antennas 880) that enables the controller 804 communicate with the database 120, the network 130, the GUI 241, and / or the display 235, as described in connection with any of FIG. 2, for example.

[0094] The interface betweenthe controller 804 with other devices, such as the database 120, the network 130, the GUI 241, and / or the display 235 may be a wired connection and / or a wireless connection. In some aspects, the controller 804 may include one or more transceiver 822 and / or one or more antennas 880. In some examples, the controller may be provided as an application at a smart phone, a tablet, a laptop, or other computer device.

[0095] In some aspects, the controller 804 may operate in a network environment, using a network connection to one or more separate, or remote, the rest of the components of computer system 800. Other devices may also be present in the computer network, such as, but not limited to, routers, network stations, peer devices or other network nodes. Network connections can form a local-area computer network (LAN) and / or a wide-area computer network (WAN), which may have access to the Internet. When135005-0001W001 29networks are used, the controller may employ a modem or other modules that enable communications with a wide-area computer network such as the Internet. It will be appreciated by those of ordinary skill in the art that said network connections are nonlimiting examples of numerous well-understood ways of establishing a connection by one computer to another using communication modules.

[0096] It is understood that the specific order or hierarchy of blocks in the processes / flowcharts disclosed is an illustration of example approaches. Based upon design preferences, it is understood that the specific order or hierarchy of blocks in the processes / flowcharts maybe rearranged. Further, some blocks may be combined or omitted. The accompanying method claims present elements of the various blocks in a sample order, and are not limited to the specific order or hierarchy presented.

[0097] The previous description is provided to enable any person skilled in the art to practice the various aspects described herein. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects. Thus, the claims are not limited to the aspects described herein, but are to be accorded the full scope consistent with the language claims. Reference to an element in the singular does not mean “one and only one” unless specifically so stated, but rather “one or more.” Terms such as “if,” “when,” and “while” do not imply an immediate temporal relationship or reaction. That is, these phrases, e.g., “when,” do notimply an immediate action in response to or during the occurrence of an action, but simply imply that if a condition is met then an action will occur, butwithoutrequiringa specific or immediate time constraintforthe action to occur. The word “exemplary” is used herein to mean “serving as an example, instance, or illustration.” Any aspect described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other aspects. Unless specifically stated otherwise, the term “some” refers to one or more. Combinations such as “at least one of A, B, or C,” “one or more of A, B, or C,” “at least one of A, B, and C,” “one or more of A, B, and C,” and “A, B, C, or any combination thereof’ include any combination of A, B, and / or C, and may include multiples of A, multiples of B, or multiples of C. Specifically, combinations such as “at least one of A, B, or C,” “one or more of A, B, or C,” “at least one of A, B, and C,” “one or more of A, B, and C,” and “A, B, C, or any combination thereof’ may be A only, B only, C only, A and B, A and C, B and C, or A and B and C, where any such combinations may contain one or more member or members of A, B, or C. Sets should be interpreted135005-0001W001 30as a set of elements where the elements number one or more. Accordingly, for a set of X, X would include one or more elements. When at least one processor (i.e., a set of one or more processors P) is configured to perform a set of functions F, each processor of P may be configured to perform a subset S of F, where S £ F. Accordingly, each processor of the at least one processor may be configured to perform a particular subset of the set of functions, where the subset is the full set, a proper subset of the set, or an empty subset of the set. A processor may be referred to as processor circuitry. A memory / memory module may be referred to as memory circuitry. If a first apparatus receives datafrom ortransmits data to a second apparatus, the data may be received / transmitted directly between the first and second apparatuses, or indirectly between the first and second apparatuses through a set of apparatuses. A device configured to “output” data or “provide” data, such as a transmission, signal, or message, may transmit the data, for example with a transceiver, or may send the data to a device that transmits the data. A device configured to “obtain” data, such as a transmission, signal, or message, may receive, for example with a transceiver, or may obtain the datafrom a device that receives the data. Information stored in a memory includes instructions and / or data. All structural and functional equivalents to the elements of the various aspects described throughout this disclosure that are known or later come to be known to those of ordinary skill in the art are expressly incorporated herein by reference and are encompassed by the claims. Moreover, nothing disclosed herein is dedicated to the public regardless of whether such disclosure is explicitly recited in the claims. The words “module,” “mechanism,” “element,” “device,” and the like may not be a substitute for the word “means.” As such, no claim element is to be construed as a means plus function unless the element is expressly recited using the phrase “means for.”

[0098] As used herein, the phrase “based on” shall not be construed as a reference to a closed set of information, one or more conditions, one or more factors, or the like. In other words, the phrase “based on A” (where “A” may be information, a condition, a factor, or the like) shall be construed as “based at least on A” unless specifically recited differently.

[0099] The following aspects are illustrative only and may be combined with other aspects or teachings described herein, without limitation.

[0100] Aspect 1 is a method for medical imaging, comprising obtaining a set of first-type medical images of a subject, the first-type medical images containing a first feature135005-0001W001 31and a second feature; and generating one or more second-type medical images based on the set of first-type medical images by selectively reducing signals corresponding to the second feature.

[0101] Aspect2 is the method of aspect 1, wherein the subject corresponds to a brain of a patient; the first feature corresponds to a lesion or pathological abnormality in the brain; and the second feature corresponds to a blood vessel or vascular structure in the brain.

[0102] Aspect 3 is the method of any of aspects 1 and 2, wherein the set of first-type medical images comprises one or more gradient echo (GRE) medical images; and the second- type of medical images comprises a vessel-suppressed GRE image.

[0103] Aspect4 is the method of any of aspects 1 to 3, further comprising: generating based on the set of first-type medical images, one or more synthesized third-type medical images, using an image-to-image synthesize model, wherein in the one or more synthesized third -type medical images, the second feature is attenuated while the first feature is retained; and generating one or more fourth-type medical images by determining a difference between the set of first-type medical images and the one or more synthesized third-type medical image.

[0104] Aspect 5 is the method of any of aspects 1 to 4, wherein: the first-type medical images and third-type medical images are obtained using a magnetic resonance imaging (MRI) device; and the fourth-type medical image contains signals corresponding to the second feature.

[0105] Aspect 6 is the method of any of aspects 1 to 5, wherein generating the one or more second-type medical images from the set of first-type medical images comprises: applying a configurable suppression factor to the set of first-type medical images based on intensity values in the one or more fourth-type medical images.

[0106] Aspect 7 is the method of any of aspects 1 to 6, wherein applying the configurable suppression factor to the set of first-type medical images based on intensity values in the one or more fourth-type medical images comprises: linearly scaling down the signals correspondingtothe second feature using the configurable suppression factor.

[0107] Aspect 8 is the method of any of aspects 1 to 7, wherein: the synthesized third-type medical image corresponds to a spin echo (SE) medical image; and the fourth-type medical image corresponds to a vessel map.

[0108] Aspect9 is the method of any of aspects 1 to 8, wherein the image-to-image synthesize model is trained by: obtaining, from a database, a plurality pairs of training data135005-0001W001 32comprising: a training set of first-type medical images; and one or more corresponding ground-truth third-type medical images; generating one or more training synthesized third-type medical images using the image-to-image synthesize model based on the training set of first-type medical images; and adjusting at least one parameter of the image-to-image synthesize model based on the one or more training synth esizedthird- type medical images and the one or more corresponding ground-truth third-type medical images.

[0109] Aspect 10 is the method of any of aspects 1 to 9, wherein: the training set of first-type medical images comprises one or more gradient echo medical images; and the one or more training synthesized third-type medical images are spin echo (SE) medical image synthesized based on the training set of first-type medical images.

[0110] Aspect 11 is the method of any of aspects 1 to 10, wherein obtaining, from a database, the plurality pairs of training data comprising: co-registering the training set of first- type medical images with the one or more corresponding ground-truth third-type medical images; performing intensity normalization; and applying histogram equalization.

[0111] Aspect 12 is the method of any of aspects 1 to 11, wherein adjusting at least one parameter of the image-to-image synthesize model based on the one or more training synthesized third-type medical images and the one or more corresponding groundtruth third-type medical images further comprises: comparingthe one or more training synthesized third-type medical images and the one or more corresponding groundtruth third-type medical images based on at least one of : a LI loss, corresponding to an absolute pixel-wise difference between the one or more training synthesized third- type medical images and the one or more corresponding ground-truth third-type medical images; a structural similarity index measure (SSIM) loss, corresponding to a structural similarity between the one or more training synthesized third -type medical images and the one or more corresponding ground-truth third-type medical images; or a perceptual loss corresponding to feature-space loss between the one or more training synthesized third-type medical images and the one or more corresponding ground-truth third-type medical images.

[0112] Aspect 13 is the method of any of aspects 1 to 12, further comprising: displaying the one or more second-type medical images on a display device.135005-0001W001 33

[0113] Aspect 14 is an apparatus formedical imagingat a computer device, comprising: at least one memory; and at least one processor coupled to the at least one memory, the at least one processor is configured to perform the method of any of aspects 1 to 13.

[0114] Aspect 15 is an apparatus for medical imaging at a computer device, comprising means for performing each step in the method of any of aspects 1 to 13.

[0115] Aspect 16 is the apparatus of any of aspects 1 to 15, further comprising a transceiver configured to receive or to transmit in association with the method of any of aspects 1 to 13.

[0116] Aspect 17 is a computer-readable medium storing computer executable code at a computer device, the code when executed by at least one processor causes the at least one processor to perform the method of any of aspects 1- 13.

Claims

135005-0001W001 34WHAT IS CLAIMED IS:

1. A computer-implemented method for medical imaging comprising:obtaining a set of first-type medical images of a subject, the first-type medical images containing a first feature and a second feature; andgenerating one or more second-type medical images based on the set of first-type medical images by selectively reducing signals corresponding to the second feature.

2. The computer-implemented method of claim 1, wherein:the subject corresponds to a brain of a patient;the first feature corresponds to a lesion or pathological abnormality in the brain; andthe second feature corresponds to a blood vessel or vascular structure in the brain.

3. The computer-implemented method of claim 2, wherein:the set of first-type medical images comprises one or more gradient echo (GRE) images; andthe second-type of medical images comprises a vessel-suppressed GRE image.

4. The computer-implemented method of claim 1, further comprising:generating based on the set of first-type medical images, one or more synthesized third-type medical images, using an image-to-image synthesize model, wherein in the one or more synthesized third-type medical images, the second feature is attenuated while the first feature is retained; andgenerating one or more fourth -type medical images by determining a difference between the set of first-type medical images and the one or more synthesized third-type medical image.

5. The computer-implemented method of claim 4, wherein:the first-type medical images and third-type medical images are obtained using a magnetic resonance imaging (MRI) device; and135005-0001W001 35the fourth-type medical image contains signals corresponding to the second feature.

6. The computer-implemented method of claim 5, wherein generating the one or more second-type medical images from the set of first-type medical images comprises:applying a configurable suppression factor to the set of first-type medical images based on intensity values in the one or more fourth-type medical images.

7. The computer-implemented method of claim 6, wherein applying the configurable suppression factor to the set of first-type medical images based on intensity values in the one or more fourth-type medical images comprises:linearly scaling down the signals corresponding to the second feature using the configurable suppression factor.

8. The computer-implemented method of claim 6, wherein:the synthesized third-type medical image corresponds to a spin echo (SE) image; andthe fourth-type medical image corresponds to a vessel map.

9. The computer-implemented method of claim 4, wherein the image-to-image synthesize model is trained by:obtaining, from a database, a plurality pairs of training data comprising:a training set of first-type medical images; andone or more corresponding ground-truth third-type medical images; generating one or more training synthesized third-type medical images using the image-to-image synthesize model based on the training set of first-type medical images; andadjusting at least one parameter of the image-to-image synthesize model based on the one or more training synthesized third -type medical images and the one or more corresponding ground-truth third-type medical images.

10. The computer-implemented method of claim 9, wherein:the training set of first-type medical images comprises one or more gradient echo medical images; and135005-0001W001 36the one or more training synthesized third-type medical images are spin echo (SE) image synthesized based on the training set of first-type medical images.

11. The computer-implemented method of claim 9, wherein obtaining, from a database, the plurality pairs of training data comprising:co-registering the training set of first-type medical images with the one or more corresponding ground-truth third-type medical images;performing intensity normalization; andapplying histogram equalization.

12. The computer-implemented method of claim 9, wherein adjusting at least one parameter of the image-to-image synthesize model based on the one or more training synthesized third-type medical images and the one or more corresponding ground-truth third-type medical images further comprises:comparing the one or more training synthesized third -type medical images and the one or more corresponding ground-truth third-type medical images based on at least one of:a LI loss, corresponding to an absolute pixel-wise difference between the one or more training synthesized third-type medical images and the one or more corresponding ground-truth third-type medical images;a structural similarity index measure (SSIM) loss, corresponding to a structural similarity between the one or more training synthesized third-type medical images and the one or more corresponding ground-truth third-type medical images; ora perceptual loss corresponding to feature-space loss between the one or more training synthesized third-type medical images and the one or more corresponding ground-truth third-type medical images.

13. The computer-implemented method of claim 1, further comprising:displaying the one or more second-type medical images on a display device.

14. A system for medical imaging, comprising:memory; and135005-0001W001 37at least one processor coupled to the memory and, based at least in part on information stored in the memory, the at least one processor is configured to:obtain a set of first-type medical images of a subject, the first-type medical images containing a first feature and a second feature; and generate one or more second-type medical images based on the set of first-type medical images by selectively reducing signals corresponding to the second feature, wherein generating the one or more second-type medical images involves processing the set of first-type medical images using an image-to-image synthesis model.

15. The system of claim 14, wherein:the subject corresponds to a brain of a patient;the first feature corresponds to a lesion or pathological abnormality in the brain; andthe second feature corresponds to a blood vessel or vascular structure in the brain.

16. The system of claim 15, wherein:the set of first-type medical images comprises one or more gradient echo (GRE) images; andthe second-type of medical images comprises a vessel-suppressed GRE image.

17. The system of claim 14, wherein the at least one processor is further configured to:generate based on the set of first-type medical images, one or more synthesized third-type medical images, using an image-to-image synthesize model, wherein in the one or more synthesized third-type medical images, the second feature is attenuated while the first feature is retained; andgenerate one or more fourth-type medical images by determining a difference between the set of first-type medical images and the one or more synthesized third-type medical image.

18. The system of claim 17, wherein:the first-type medical images and third-type medical images are obtained using a magnetic resonance imaging (MRI) device; and135005-0001W001 38the fourth-type medical image contains signals corresponding to the second feature.

19. The system of claim 18, wherein to generate the one or more second-type medical images from the set of first-type medical images, the at least one processor is further configured to:applying a configurable suppression factor to the set of first-type medical images based on intensity values in the one or more fourth-type medical images, such that the signals corresponding to the second feature using the configurable suppression factor can be linearly scaled down in the set of first-type medical images.

20. A non-transitory computer-readable storage medium including instructions that, when executed by one or more processors, case the one or more processors to perform operations comprising:obtaining a set of gradient echo (GRE) images of a brain of a patient, the GRE images containing a first feature corresponding to a lesion or pathological abnormality in the brain and a second feature corresponding to a blood vessel or vascular structure in the brain; andgenerating one or more vessel-suppressed GRE images based on the set of GRE images by selectively reducing signals corresponding to the second feature.

Images