Method and system for automatic detection and analysis of rectal gas effects on diffusion weighted imaging

Abstract

A method includes acquiring a localizer magnetic resonance imaging (MRI) image of a subject, segmenting a prostate in the localizer MRI image to create a prostate segmentation mask, determining a region of interest (ROI) around the prostate, applying contrast normalization to normalize the localizer MRI image, segmenting a rectal gas region within the ROI around the prostate, calculating a morphological parameter based on the prostate segmentation mask and the rectal gas region, determining acquisition parameters for diffusion weighted imaging (DWI), predicting a DWI distortion probability based on the morphological parameter and the acquisition parameters, and generating an alert to a technician about potential DWI artifacts when the DWI distortion probability exceeds a threshold. The automatic detection and analysis of rectal gas enables the technician to take corrective action prior to DWI acquisition, thereby reducing the need for repeat examinations and improving the diagnostic quality of prostate MRI studies.

Description

Technical Field

[0001] The embodiments of the subject matter disclosed herein relate to medical imaging, and more specifically to magnetic resonance imaging (MRI). Background Technology

[0002] In a magnetic resonance imaging (MRI) system, a subject is positioned within a strong magnetic field and exposed to radio frequency (RF) pulses, which cause hydrogen atoms in the subject's tissues to emit signals. These signals are detected and processed to generate images of the subject's internal structures. Diffusion-weighted imaging (DWI) is a dedicated MRI technique that measures the random Brownian motion of water molecules within tissues. This dedicated MRI technique provides valuable diagnostic information, particularly for oncology applications such as the detection and characterization of prostate cancer.

[0003] Prostate MRI examinations typically follow the Prostate Imaging Reporting and Data System (PI-RADS) guidelines, which recommend the use of DWI, along with T2-weighted (T2W) and dynamic contrast-enhanced (DCE) imaging, to detect prostate cancer and perform risk scoring. This requires that the DWI data be of sufficiently good quality for radiologist interpretation and subsequent risk assessment. However, various factors can degrade DWI image quality, with rectal gas being a significant issue.

[0004] Rectal gas can cause susceptible artifacts in echo-planar imaging (EPI) DWI images of the prostate, leading to image distortion and signal loss. These artifacts occur at the air-tissue interface, making accurate assessment and study of prostate tumors challenging. Multiple studies have highlighted this issue, noting that the presence of rectal gas can render DWI sequences diagnostically worthless or significantly impair their interpretability. When DWI sequences are affected by rectal gas artifacts, radiologists may be unable to properly assess the peripheral areas of the prostate, potentially leading to missed lesions or inaccurate characterization of identified abnormalities.

[0005] Current methods for addressing this issue are limited. Some facilities employ patient preparation protocols, such as administering laxatives or enemas before scanning, but these measures are not always effective and can be uncomfortable for patients. During scanning, technicians must manually review images to identify problematic rectal gas, which requires experience and can disrupt workflows. Some commercial systems prompt technicians to manually indicate whether rectal gas is suspected, but this approach still relies on human judgment and does not provide automated analysis of the potential impact on image quality. Furthermore, rectal gas can move during the examination due to natural peristalsis. Gas that was not present during the initial locator scan may move into the rectum during subsequent sequences, potentially affecting DWI quality. This temporal variability adds another layer of complexity to managing artifacts caused by rectal gas.

[0006] When DWI sequences are affected by rectal gas artifacts, repeated examinations may be necessary, leading to increased scan time, reduced patient turnover, increased costs, and patient inconvenience. In some cases, patients may need to return for repeat examinations on a different day, further delaying diagnosis and treatment. Therefore, there remains an unmet need for automated systems that can detect rectal gas, analyze its potential impact on DWI image quality, and alert technicians before DWI acquisition, allowing for timely intervention and optimization of scan parameters. Such systems would improve workflow efficiency, reduce the need for repeat examinations, and ultimately improve the diagnostic quality of prostate MRI studies. Summary of the Invention

[0007] In one example, a method includes: acquiring a locator MRI image of a subject; segmenting the prostate in the locator MRI image to create a prostate segmentation mask; identifying a region of interest (ROI) surrounding the prostate; applying contrast normalization to standardize the locator MRI image; segmenting a rectal gas region within the ROI surrounding the prostate; calculating morphological parameters based on the prostate segmentation mask and the rectal gas region; determining acquisition parameters for DWI; predicting the probability of DWI distortion based on the morphological parameters and acquisition parameters; generating an alert to a technician regarding potential DWI artifacts when the probability of DWI distortion exceeds a threshold; automatically adjusting scan parameters when the probability of DWI distortion exceeds the threshold; and acquiring medical images based on the adjusted scan parameters.

[0008] In another example, a system includes one or more processors and a non-transitory memory storing instructions that can be executed by the one or more processors to: acquire an initial locator MRI image of a subject; segment the prostate in the initial locator MRI image; determine the region of interest (ROI) around the segmented prostate; apply contrast normalization to normalize the initial locator MRI image; segment rectal gas regions within the ROI around the prostate; selectively filter the segmented rectal gas regions; calculate morphological parameters based on the segmented prostate and the segmented rectal gas regions; determine acquisition parameters for DWI; predict the probability of DWI distortion based on the morphological parameters and the acquisition parameters; and generate an alert to a technician when the predicted probability of DWI distortion exceeds a threshold.

[0009] In another example, a method for predicting DWI distortion in prostate MRI includes: acquiring a locator MRI image of a subject; using a deep learning-based model to segment the prostate in the locator MRI image to create a prostate segmentation mask; determining ROIs extending a predetermined distance from the prostate segmentation mask along multiple directions; applying contrast normalization to remove radio frequency bias in the locator MRI image; calculating the average signal intensity in the ROI surrounding the prostate; setting a gas threshold as a certain proportion of the average signal intensity; identifying pixels within the ROI with normalized signal intensity below the gas threshold to create a binary gas mask; applying connected component analysis to remove gas regions smaller than a predetermined size; calculating the minimum surface distance between the remaining gas regions and the prostate; calculating the total gas volume within the ROI; retrieving acquisition parameters for planning the DWI sequence; encoding the minimum surface distance, total gas volume, and acquisition parameters as input features for a prediction model; applying the prediction model to generate the probability of DWI distortion; and displaying an alarm on a display device when the probability of DWI distortion exceeds a threshold.

[0010] The method described above offers several advantages over current technologies. By automatically detecting and analyzing rectal gas before DWI acquisition, the system enables technicians to take corrective actions based on patient preparation or DWI acquisition parameter adjustments. This reduces the need for re-examinations, which can lead to increased scan time, reduced patient turnover, increased costs, and patient inconvenience. The automated nature of the detection system eliminates reliance on human judgment and experience, thus providing a consistent assessment of potential DWI distortions across different operators and facilities. Furthermore, by analyzing both morphological parameters (such as gas volume and distance from the prostate) and acquisition parameters, the system provides more accurate predictions of potential artifacts than visual examination alone. Prostate segmentation and connected component analysis using a deep learning-based model to filter gas regions further improve the accuracy and reliability of the predictions. This ultimately improves workflow efficiency, enhances the diagnostic quality of prostate MRI studies, and increases the likelihood that DWI sequences will be of sufficient quality to facilitate proper reading and risk scoring according to PI-RADS guidelines.

[0011] The above advantages, as well as other advantages and features of this specification, will become apparent from the following detailed description when considered alone or in conjunction with the accompanying drawings. It should be understood that the above summary is provided to present a series of concepts further described in the detailed description in a simplified form. This is not intended to identify key or essential features of the claimed subject matter, the scope of which is uniquely defined by the claims following the detailed description. Furthermore, the claimed subject matter is not limited to specific implementations that address any shortcomings pointed out above or in any part of this disclosure. Attached Figure Description

[0012] A better understanding of the various aspects of this disclosure can be achieved by reading the following detailed description and referring to the accompanying drawings, in which:

[0013] Figure 1 This is a schematic diagram illustrating an imaging system for acquiring MRI data according to an embodiment of the present disclosure;

[0014] Figure 2 This is a block diagram illustrating an image processing apparatus according to an embodiment of the present disclosure;

[0015] Figure 3 A flowchart illustrating a method for automatically detecting and analyzing the effect of rectal gas on diffusion-weighted imaging according to an embodiment of the present disclosure is shown;

[0016] Figure 4 A flowchart illustrating a method for segmenting a rectal gas region within a region of interest surrounding the prostate, according to an embodiment of the present disclosure, is shown.

[0017] Figure 5A flowchart illustrating a method for calculating morphological parameters based on a prostate segmentation mask and segmented gas regions according to an embodiment of the present disclosure is shown.

[0018] Figure 6 A flowchart illustrating a method for managing MRI scan protocols based on gas analysis and distortion probability according to an embodiment of the present disclosure is shown;

[0019] Figure 7 A visual representation of a contrast normalization process for enhancing medical images according to an embodiment of the present disclosure is shown;

[0020] Figure 8 A visualization of a prostate segmentation process according to an embodiment of the present disclosure is shown, illustrating the transformation from an input medical image to a segmented output image;

[0021] Figure 9 A comparison of anatomical scan images before and after gas fractionation according to an embodiment of the present disclosure is shown; and

[0022] Figure 10 A medical imaging visualization illustrating ROI selection is shown according to an embodiment of this disclosure. Detailed Implementation

[0023] The following description relates to the automated detection and analysis of the effects of rectal gas on diffusion-weighted imaging (DWI). In particular, the following description relates to methods and systems for automatically detecting rectal gas in positioner magnetic resonance imaging (MRI) images, analyzing the potential impact of rectal gas on DWI quality, and providing timely alerts to technicians before DWI acquisition to enable corrective actions.

[0024] Rectal gas can cause significant susceptible artifacts in echo-planar imaging (EPI) DWI sequences of the prostate, leading to image distortion and signal loss at the air-tissue interface. These artifacts can render DWI sequences diagnostically worthless or significantly impair their interpretability, potentially resulting in missed lesions or inaccurate characterization of identified abnormalities. Current methods for addressing this problem are limited and typically rely on manual review of images by technicians to identify problematic rectal gas, which requires experience and can disrupt workflows. When DWI sequences are affected by rectal gas artifacts, repeated examinations may be necessary, leading to increased scan time, reduced patient turnover, increased costs, and patient inconvenience.

[0025] In one implementation, MRI images can be obtained using an MRI imaging system such as Figure 1The MRI system 10 shown is used for acquisition. The MRI system 10 includes a static magnetic field magnet unit 12, a gradient coil unit 13, RF coil units 14 and 15, and other components for acquiring MRI data of a subject positioned within the imaging space 18. The acquired MRI data can be transmitted via... Figure 2 The image processing system 200 shown is used for processing, and the image processing system includes an image processing device 202 having a processor 204 and a non-transitory memory 206 storing various processing modules for analyzing prostate images and predicting distortion.

[0026] Image processing system 200 executable Figure 3 The illustrated method 300 includes one or more operations that illustrate a workflow for automatically detecting and analyzing the effects of rectal gas on DWI. Method 300 includes acquiring locator MRI images, segmenting the prostate, determining the ROI around the prostate, applying contrast normalization, segmenting the rectal gas region, calculating morphological parameters, determining acquisition parameters, predicting the probability of DWI distortion, and generating an alarm if necessary. The gas segmentation portion of method 300 is described in... Figure 4 The figure describes in detail the process for segmenting rectal gas regions within a region of interest (ROI) surrounding the prostate, including applying contrast normalization, determining the ROI, calculating the average signal intensity, setting a gas threshold, identifying pixels below the threshold, and applying post-processing to remove small regions of detected gas.

[0027] Once the gas region is divided, the system is as follows: Figure 5 The diagram illustrates the process of calculating morphological parameters based on a prostate segmentation mask and the segmented gas regions. This includes calculating the minimum surface distance between the gas and the prostate, calculating the total gas volume, retrieving and acquiring parameters, and encoding these parameters as input features for a prediction model. Based on the distortion probability predicted only, the system can achieve... Figure 6 The protocol management workflow shown in the diagram illustrates a method for managing MRI scan protocols based on gas analysis and distortion probability, including receiving DWI distortion probability results, displaying alerts with highlighted gas areas, presenting intervention options to technicians, and updating the scan protocol based on the selected intervention.

[0028] The contrast normalization process applied in methods 300 and 400 is as follows: Figure 7 Visually, the figure illustrates the transformation from an initial medical image to a contrast-normalized medical image with standardized intensity values, which facilitate more reliable gas segmentation. Figure 8 A visualization of the prostate segmentation process referenced in Method 300 is shown, illustrating how a deep learning-based model transforms an input medical image into a segmented output image with the prostate highlighted. Figure 9A comparison of anatomical scan images before and after gas segmentation is shown, demonstrating how the gas segmentation process identifies rectal gas regions that may potentially cause susceptible artifacts in DWI sequences. Finally, Figure 10 The medical imaging visualization with ROI selection is shown, including a full anatomical view image with the highlighted ROI and a magnified view of the region, showing detailed segmentation of the prostate and any detected rectal gas areas.

[0029] This automated approach addresses the limitations of existing methods by providing a system that detects rectal gas, analyzes its potential impact on DWI image quality, and alerts technicians before DWI acquisition, allowing for timely intervention and adjustment of scanning parameters. By analyzing both morphological parameters (such as gas volume and distance from the prostate) and acquisition parameters, the system provides more accurate predictions of potential artifacts than visual examination alone. This ultimately improves workflow efficiency, reduces the need for repeat examinations, and enhances the diagnostic quality of prostate MRI studies, increasing the likelihood that DWI sequences will have sufficient quality to facilitate proper reading and risk scoring according to PI-RADS guidelines.

[0030] First refer to Figure 1 The diagram illustrates an MRI apparatus 10, which includes a static magnetic field magnet unit 12, a gradient coil unit 13, an RF coil unit 14, an RF body coil unit 15 (e.g., an image coil unit), a transmit / receive (T / R) switch 20, an RF driver unit 22, a gradient coil driver unit 23, a data acquisition unit 24, a controller unit 25, a patient bed or examination table 26, a data processing unit 31, a scan control device 32, and a display unit 33. In some embodiments, the RF coil unit 14 is a surface coil, which is a local coil typically placed near the anatomical structures of interest of the subject 16. Herein, the RF body coil unit 15 is a transmitting coil that transmits RF signals, and the local surface of the RF coil unit 14 receives MR signals. Therefore, the transmitting body coil (e.g., the RF body coil unit 15) and the surface receiving coil (e.g., the RF coil unit 14) are independent but electromagnetically coupled components. The MRI apparatus 10 transmits electromagnetic pulse signals to the subject 16, which is placed in an imaging space 18 that forms a static magnetic field, to perform a scan to obtain magnetic resonance signals from the subject 16. One or more images of subject 16 can be reconstructed based on the magnetic resonance signals obtained by scanning.

[0031] The static magnetic field magnet unit 12 includes, for example, a toroidal superconducting magnet mounted within a toroidal vacuum container. The magnet defines a cylindrical space surrounding the subject 16 and generates a constant main static magnetic field B0.

[0032] The MRI apparatus 10 also includes a gradient coil unit 13 that forms a gradient magnetic field in the imaging space 18 to provide three-dimensional positional information for magnetic resonance signals received by the RF coil array. The gradient coil unit 13 includes three gradient coil systems, each generating a gradient magnetic field along one of three spatial axes perpendicular to each other, and generating gradient fields in each of the frequency encoding direction, phase encoding direction, and slice selection direction, depending on the imaging conditions. More specifically, the gradient coil unit 13 applies a gradient field in the slice selection direction (or scan direction) of the subject 16 to select slices; and the RF body coil unit 15 or the local RF coil array can send RF pulses to the selected slices of the subject 16. The gradient coil unit 13 also applies a gradient field in the phase encoding direction of the subject 16 to perform phase encoding of the magnetic resonance signals from the RF pulse-excited slices. Then, the gradient coil unit 13 applies a gradient field in the frequency encoding direction of the subject 16 to perform frequency encoding of the magnetic resonance signals from the RF pulse-excited slices.

[0033] RF coil unit 14 is configured, for example, to surround the imaging region of subject 16. In some examples, RF coil unit 14 may be referred to as a surface coil or receiving coil. In the static magnetic field space or imaging space 18 in which the static magnetic field B0 is formed by static magnetic field magnet unit 12, RF coil unit 15 sends RF pulses as electromagnetic waves to subject 16 based on control signals from controller unit 25, thereby generating a high-frequency magnetic field B1. This excites proton spins in the slice to be imaged in subject 16. RF coil unit 14 receives electromagnetic waves generated as magnetic resonance signals when the proton spins thus excited in the slice to be imaged in subject 16 return to alignment with the initial magnetization vector. In some embodiments, RF coil unit 14 may transmit RF pulses and receive MR signals. In other embodiments, RF coil unit 14 may be used only to receive MR signals without transmitting RF pulses.

[0034] The RF body coil unit 15 is configured, for example, to surround the imaging space 18 and generate RF magnetic field pulses within the imaging space 18 that are orthogonal to the main magnetic field B0 generated by the static magnetic field magnet unit 12 to excite the nucleus. The RF body coil unit 15 is fixedly attached to and connected to the MRI apparatus 10, unlike the RF coil unit 14, which can be disconnected from the MRI apparatus 10 and replaced by another RF coil unit. Furthermore, while local coils (such as the RF coil unit 14) can only send or receive signals to or from a local area of ​​the subject 16, the RF body coil unit 15 typically has a larger coverage area. For example, the RF body coil unit 15 can be used to send or receive signals to or from the whole body of the subject 16. Using only receiving local coils and transmitting body coils provides uniform RF excitation and good image homogeneity, at the cost of higher RF power deposited in the subject. For transmit-receive local coils, the local coil provides RF excitation to the region of interest and receives the MR signal, thereby reducing the RF power deposited in the subject. It should be understood that the specific use of the RF coil unit 14 and / or the RF body coil unit 15 depends on the imaging application.

[0035] When operating in receive mode, T / R switch 20 selectively connects RF body coil unit 15 to data acquisition unit 24, and when operating in transmit mode, it selectively connects the RF body coil unit to RF driver unit 22. Similarly, when RF coil unit 14 operates in receive mode, T / R switch 20 selectively connects RF coil unit 14 to data acquisition unit 24, and when the RF coil unit operates in transmit mode, it selectively connects the RF coil unit to RF driver unit 22. When both RF coil unit 14 and RF body coil unit 15 are used for a single scan, for example, if RF coil unit 14 is configured to receive MR signals and RF body coil unit 15 is configured to transmit RF signals, T / R switch 20 can direct control signals from RF driver unit 22 to RF body coil unit 15 while simultaneously directing the received MR signals from RF coil unit 14 to data acquisition unit 24. The coil of RF coil unit 15 can be configured to operate in transmit-only mode or transmit-receive mode. The coil of RF coil unit 14 can be configured to operate in transmit-receive mode or receive-only mode.

[0036] The RF driver unit 22 includes a gate modulator, an RF power amplifier (not shown), and an RF oscillator (not shown), which drive an RF coil (e.g., RF body coil unit 15) and generate a high-frequency magnetic field in the imaging space 18. Based on a control signal from the controller unit 25 and using the gate modulator, the RF driver unit 22 modulates the RF signal received from the RF oscillator into a signal with a predetermined timing and a predetermined envelope. The RF signal modulated by the gate modulator is amplified by the RF power amplifier and then output to the RF body coil unit 15.

[0037] The gradient coil driver unit 23 drives the gradient coil unit 13 based on control signals from the controller unit 25, thereby generating a gradient magnetic field in the imaging space 18. The gradient coil driver unit 23 includes three systems (not shown) of driver circuits corresponding to the three gradient coil systems included in the gradient coil unit 13.

[0038] The data acquisition unit 24 includes a preamplifier, a phase detector, and an analog-to-digital converter for acquiring the magnetic resonance signal received by the RF coil unit 14. In the data acquisition unit 24, the phase detector uses the output of the RF oscillator from the RF driver unit 22 as a reference signal to perform phase detection on the magnetic resonance signal received from the RF coil unit 14 and amplified by the preamplifier. The phase-detected analog magnetic resonance signal is then output to the analog-to-digital converter for conversion into a digital signal. The resulting digital signal is then output to the data processing unit 31.

[0039] The MRI apparatus 10 includes an examination table 26 for placing a subject 16 thereon. The subject 16 can be moved inside and outside the imaging space 18 by moving the examination table 26 based on control signals from the controller unit 25.

[0040] The controller unit 25 includes a computer and a non-transitory computer-readable storage medium thereon storing computer-executable instructions. When executed by the computer, these instructions cause various components of the device 10 to perform operations corresponding to a predetermined scanning protocol. The non-transitory computer-readable storage medium may include one or more of a solid-state drive (SSD), hard disk drive (HDD), hybrid drive, optical disc (e.g., CD, DVD, Blu-ray), flash memory device, random access memory (RAM) device, read-only memory (ROM) device, or any other suitable non-transitory storage medium. In some embodiments, the non-transitory computer-readable storage medium may be a cloud-based or network-attached storage system accessible by the controller unit 25 via a wired or wireless network connection.

[0041] The controller unit 25 is connected to the scanning control device 32 and processes the operation signals input to the scanning control device 32. It also controls the inspection table 26, the RF driver unit 22, the gradient coil driver unit 23, and the data acquisition unit 24 by outputting control signals to them. The controller unit 25 also controls the data processing unit 31 and the display unit 33 based on the operation signals received from the scanning control device 32 to obtain the desired image.

[0042] The scanning control device 32 includes user input devices such as a touchscreen, keyboard, and mouse. The operator uses the scanning control device 32 to, for example, input data as an imaging scheme and set the area to be processed in the imaging sequence. Data regarding the imaging scheme and the area to be processed in the imaging sequence is output to the controller unit 25.

[0043] The data processing unit 31 includes a computer and a recording medium on which a program to be executed by the computer to perform predetermined data processing is recorded. The data processing unit 31 is connected to the controller unit 25 and performs data processing based on control signals received from the controller unit 25. The data processing unit 31 is also connected to the data acquisition unit 24 and generates spectral data by applying various image processing operations to the magnetic resonance signals output from the data acquisition unit 24. In some embodiments, the data processing unit 31 includes a rectal gas detection and analysis module configured to automatically detect and analyze rectal gas in prostate MRI scans. The rectal gas detection and analysis module may include a prostate segmentation component, a rectal gas segmentation component, and a distortion prediction component that uses a machine learning model to predict the probability of distortion in diffusion-weighted imaging (DWI) based on parameters such as the minimum distance between the gas pocket and the prostate, gas volume, and acquisition parameters.

[0044] Display unit 33 includes a display device and displays images on the display screen of the display device based on control signals received from controller unit 25. Display unit 33 displays images, for example, of input items regarding operational data entered by the operator from scan control device 32. Display unit 33 also displays two-dimensional (2D) slice images or three-dimensional (3D) images of the subject 16 generated by data processing unit 31. In some embodiments, display unit 33 is configured to display a warning message to the operator regarding the presence of rectal gas and its potential impact on DWI image quality, thereby allowing the operator to take corrective actions before continuing DWI acquisition.

[0045] During an MRI scan performed using the MRI apparatus 10, a subject can be positioned within the imaging space 18, and an acquisition protocol can be executed to acquire the subject's MR signal. The acquisition protocol may include multiple pulse sequences, wherein in each pulse sequence, one or more RF pulses applied via the RF body coil unit 15 are used to prepare contrast, and the gradient coil unit 13 is controlled to spatially encode the resulting MR signal. The spatially encoded MR signal received by the RF coil unit 14 is digitized and stored in k-space. Therefore, k-space data or k-space datasets can refer to the raw MR signal before processing into an image. In some examples, the raw MR signal may be filled with a line in k-space for each pulse sequence (also referred to as a repetition time). In other examples, the raw MR signal may be filled with a line in k-space for each echo, wherein more than one echo is generated for each pulse sequence / repetition time. k-space data may also be referred to herein as imaging data or MR data.

[0046] For prostate imaging applications, the MRI unit 10 can be configured to perform diffusion-weighted imaging (DWI) sequences that are particularly sensitive to susceptible artifacts caused by rectal gas. The controller unit 25 and data processing unit 31 can be configured to automate a workflow to predict and mitigate the impact of rectal gas on DWI image quality. This workflow utilizes triplane locator images acquired during scan planning to detect rectal gas, determine its proximity to the prostate, and assess its potential impact on DWI image quality. The workflow may include segmenting the prostate, locating areas around the prostate where rectal gas might significantly affect its function, segmenting rectal gas within these areas, and analyzing various parameters that affect distortion in echo-planar imaging (EPI) DWI sequences. Based on this analysis, the system can provide warnings to the operator and suggest appropriate corrective actions or alternative acquisition schemes to minimize artifacts and ensure diagnostic image quality.

[0047] refer to Figure 2 An image processing system 200 for analyzing prostate images and predicting distortion is shown. The image processing system 200 is configured to automatically detect rectal gas in locator MRI images, analyze its potential impact on DWI quality, and provide timely alerts to technicians before DWI acquisition to enable corrective actions. The image processing system 200 can be integrated into an MRI system (such as...). Figure 1 The image processing system 200 may be provided as a separate system communicatively coupled to the MRI system via wired and / or wireless connections, as shown in the MRI apparatus. In some embodiments, at least a portion of the image processing system 200 is located at a separate device (e.g., a workstation) that can receive MRI data from the MRI system or from a storage device storing MRI data generated by the MRI system.

[0048] Image processing system 200 includes image processing device 202 with processor 204 configured to execute machine-readable instructions stored in non-transitory memory 206. Processor 204 may be single-core or multi-core, and programs executing thereon may be configured for parallel or distributed processing. In some embodiments, processor 204 may optionally include separate components distributed across two or more devices, which may be remotely located and / or configured for collaborative processing. In some embodiments, one or more aspects of processor 204 may be virtualized and executed by a remotely accessible networked computing device configured in a cloud computing configuration. Processor 204 may be communicatively coupled to non-transitory memory 206, display device 224, and user input device 222, and is configured to execute various modules stored in non-transitory memory 206 for automated detection and analysis of the effects of rectal gas on DWI.

[0049] Non-transitory memory 206 stores machine-readable instructions that, when executed by processor 204, cause image processing device 202 to perform various operations related to the automatic detection and analysis of the effects of rectal gas on DWI. In one embodiment, non-transitory memory 206 stores a prostate segmentation module 208, a ROI determination module 210, a gas segmentation module 212, a gas filtering module 214, a morphological parameter determination module 216, a distortion prediction model 218, and an alarm generation module 220. In another embodiment, non-transitory memory 206 may also store MRI data, including locator MRI images, segmentation masks, and other data generated during MRI image processing. Non-transitory memory 206 may further store training data for one or more machine learning models used in various modules, such as deep learning-based models for prostate segmentation and gas detection.

[0050] Display device 224 may include one or more display devices utilizing virtually any type of technology. In one embodiment, display device 224 may include a computer monitor and may display MRI images, including locator images, segmentation results, and alarms generated by alarm generation module 220. In another embodiment, display device 224 may be integrated into an MRI system, such as in… Figure 1 The display unit of the MRI system shown is located therein. The display device 224 may be integrated into the same package as the processor 204, the non-transitory memory 206, and / or the user input device 222, or it may be a peripheral display device and may include a monitor, a touch screen, a projector, or other display devices known in the art, which enable the user to view MRI images generated by the MRI system and / or interact with various data stored in the non-transitory memory 206.

[0051] User input device 222 may include one or more of a touchscreen, keyboard, mouse, touchpad, or other devices configured to enable a user to interact with and manipulate data within image processing system 200. In one embodiment, user input device 222 may enable a user to select MRI data for analysis of the effects of rectal gas on DWI, or for further processing using various modules stored in non-transitory memory 206. In another embodiment, user input device 222 may enable a user to acknowledge alarms generated by alarm generation module 220 and select appropriate corrective actions, such as administering a microenema to reduce rectal gas or switching to a more robust DWI acquisition scheme.

[0052] The prostate segmentation module 208 is configured to segment the prostate in a locator MRI image to create a prostate segmentation mask. In one embodiment, the prostate segmentation module 208 may include a deep learning-based model specifically developed for prostate segmentation on sagittal and coronal locators. The segmentation model can output a binary mask identifying prostate regions within the locator image. In another embodiment, the prostate segmentation module 208 may utilize existing prostate segmentation models, such as those used in automated scan planning systems, to identify the boundaries of the prostate. The prostate segmentation module 208 provides the prostate segmentation mask to the ROI determination module 210 and the morphological parameter determination module 216 for further processing.

[0053] The ROI determination module 210 is configured to determine a region of interest (ROI) around the prostate based on a prostate segmentation mask generated by the prostate segmentation module 208. In one embodiment, the ROI may be defined as extending a predetermined distance from the prostate segmentation mask along multiple directions (such as 50 mm upward, 20 mm downward, 10 mm forward, and 80 mm backward). This ROI focuses subsequent analysis on the area most likely to contain rectal gas that may affect DWI quality. In another embodiment, the ROI may be dynamically determined based on the size and location of the prostate, thereby scaling the extension distance to ensure adequate coverage of the rectal region adjacent to the prostate. The ROI determination module 210 provides the determined ROI to the gas segmentation module 212 for segmenting the rectal gas region.

[0054] Gas segmentation module 212 is configured to segment rectal gas regions within a region of interest (ROI) surrounding the prostate. In one embodiment, gas segmentation module 212 may apply a normalized contrast normalization (SCNe) to the locator MRI image to remove radio frequency (RF) bias and normalize a threshold for subsequent segmentation steps. Gas regions can then be identified by applying the threshold to normalized signal intensity, where pixels with normalized intensities below a certain threshold (e.g., 0.3 times the average prostate signal intensity) are classified as gas. In another embodiment, gas segmentation module 212 may employ a convolutional neural network (CNN) for gas segmentation, potentially providing improved resilience to segmentation errors and better adaptability to various image characteristics. Gas segmentation module 212 provides the segmented gas regions to gas filtering module 214 for further processing.

[0055] Gas filtering module 214 is configured to selectively filter segmented rectal gas regions identified by gas segmentation module 212. In one embodiment, gas filtering module 214 may apply post-processing to remove regions smaller than a predetermined number of pixels or voxels (e.g., 100 voxels), as these small regions are unlikely to cause significant DWI distortion and may represent noise or segmentation errors. This filtering step helps focus the analysis on gas sacs large enough to potentially affect image quality. In another embodiment, gas filtering module 214 may apply morphological operations such as erosion and dilation to refine gas region boundaries and connect nearby gas sacs that may represent single, continuous gas regions. Gas filtering module 214 provides the filtered gas regions to morphological parameter determination module 216 for further analysis.

[0056] The morphological parameter determination module 216 is configured to calculate morphological parameters based on the prostate segmentation mask and the filtered gas regions. In one embodiment, the morphological parameters may include the minimum surface distance between the gas regions and the prostate, calculated using distance transformation techniques such as Hausdorff distance. Additionally, the total gas volume within the ROI can be calculated by summing the volumes of all identified gas regions. In another embodiment, the morphological parameter determination module 216 may calculate additional morphological parameters, such as the shape and orientation of the gas pockets relative to the prostate, or the percentage of the prostate's posterior surface adjacent to the gas. The morphological parameter determination module 216 provides the calculated morphological parameters to the distortion prediction model 218 for predicting DWI distortion probability.

[0057] Distortion prediction model 218 is configured to predict the DWI distortion probability based on morphological parameters calculated by morphological parameter determination module 216 and acquisition parameters used for the DWI sequence. In one embodiment, distortion prediction model 218 may include a multivariate logistic regression model trained on a dataset of cases with known DWI distortion results. The model may use inputs such as the minimum surface distance between the gas and the prostate and the total gas volume. In another embodiment, distortion prediction model 218 may employ a machine learning model, such as a random forest or neural network, to capture the nonlinear relationship between the input parameters and the distortion probability. Distortion prediction model 218 provides the predicted DWI distortion probability to alarm generation module 220 for determining whether to generate an alarm.

[0058] Alarm generation module 220 is configured to generate an alert to technicians regarding potential DWI artifacts when the DWI distortion probability predicted by distortion prediction model 218 exceeds a threshold. In one embodiment, the threshold can be determined by performing a receiver operating characteristic (ROC) analysis of the performance of the prediction model on a validation dataset, where a cutoff value (e.g., >0.52) is selected to balance sensitivity and specificity. If the predicted probability exceeds the threshold, alarm generation module 220 can prominently display the alert on display device 224, highlighting the rectal gas area on the locator image and indicating a high probability of DWI distortion. In another embodiment, alarm generation module 220 may also suggest optimizing scan parameters to mitigate potential artifacts, such as applying a microenema prior to the scan to reduce rectal gas, or switching to a more robust DWI acquisition scheme less susceptible to susceptible artifacts, such as PROPELLER diffusion-weighted imaging.

[0059] refer to Figure 3 A flowchart of a method 300 for predicting and managing diffusion-weighted imaging (DWI) distortion in prostate MRI is shown. Method 300 provides an automated approach for detecting rectal gas, analyzing its potential impact on DWI image quality, and alerting technicians prior to DWI acquisition to enable timely intervention and optimization of scan parameters. Method 300 begins with acquiring initial locator MRI images and proceeds through various processing steps to predict the probability of DWI distortion, ultimately providing appropriate alerts and recommendations to technicians as necessary.

[0060] At operation 302, the method includes acquiring initial locator MRI images of the subject. In one embodiment, the initial locator MRI images may include a three-plane locator for scan planning, such as sagittal, coronal, and axial views. These locator images are typically acquired at the start of a prostate MRI examination and serve as the basis for subsequent scan planning. In another embodiment, the initial locator MRI images may be acquired immediately prior to DWI acquisition to minimize the impact of gas movement due to natural peristalsis, as gas not present during the initial locator scan may move into the rectum during subsequent sequences.

[0061] At operation 304, the method includes segmenting the prostate in the locator MRI image to create a prostate segmentation mask. In one embodiment, prostate segmentation can be performed using a deep learning-based model specifically developed for prostate segmentation on sagittal and coronal locators. The segmentation model outputs a binary mask identifying prostate regions within the locator image. In another embodiment, the method can leverage existing prostate segmentation models, such as those used in automated scan planning systems, to identify the boundaries of the prostate.

[0062] At operation 306, the method includes determining a region of interest (ROI) surrounding the prostate. In one embodiment, the ROI may be defined as extending a predetermined distance from the prostate segmentation mask along multiple directions (such as 50 mm upward, 20 mm downward, 10 mm forward, and 80 mm backward). This ROI focuses subsequent analysis on the area most likely to contain rectal gas that may affect DWI quality. In another embodiment, the ROI may be dynamically determined based on the size and location of the prostate, thereby scaling the extension distance to ensure adequate coverage of the rectal region adjacent to the prostate.

[0063] At operation 308, the method includes applying contrast normalization to standardize the locator MRI image. In one embodiment, contrast normalization may be performed using a standardized contrast normalization (SCNe) technique to remove radio frequency (RF) bias in the locator data and normalize thresholds used for subsequent segmentation steps. This normalization helps ensure consistent performance across different scanning protocols and devices. In another embodiment, contrast normalization may include histogram equalization or other intensity normalization techniques to account for variations in image acquisition parameters.

[0064] At operation 310, the method includes segmenting rectal gas regions within a region of interest (ROI) surrounding the prostate. In one embodiment, gas regions can be identified by applying a threshold to a normalized signal intensity, where pixels with a normalized intensity below a certain threshold (e.g., 0.3 times the average prostate signal intensity) are classified as gas. This approach leverages the characteristic appearance of gas as a low-signal (dark) region in MRI. In another embodiment, a convolutional neural network (CNN) can be employed for gas segmentation, potentially providing improved resilience to segmentation errors and better adaptability to various image characteristics.

[0065] At operation 312, the method includes selectively filtering the segmented rectal gas regions. In one embodiment, post-processing may be applied to remove regions smaller than a predetermined number of pixels or voxels (e.g., 100 voxels), as these small regions are unlikely to cause significant DWI distortion and may represent noise or segmentation errors. This filtering step helps focus the analysis on gas sacs large enough to potentially affect image quality. In another embodiment, morphological operations such as erosion and dilation may be applied to refine the gas region boundaries and connect nearby gas sacs that may represent single, continuous gas regions.

[0066] At operation 314, the method includes calculating morphological parameters based on a prostate segmentation mask and a rectal gas region segmentation mask. In one embodiment, the morphological parameters may include the minimum surface distance between the gas region and the prostate, calculated using a distance transformation technique such as Hausdorff distance. Additionally, the total gas volume within the ROI can be calculated by summing the volumes of all identified gas regions. In another embodiment, additional morphological parameters may be calculated, such as the shape and orientation of the gas sacs relative to the prostate, or the percentage of the prostate's posterior surface adjacent to the gas.

[0067] At operation 316, the method includes determining acquisition parameters for the DWI sequence. In one embodiment, acquisition parameters may include echo interval, phase encoding line, B0 inhomogeneity, field strength, acquisition scheme (e.g., single-shot EPI, FOCUS, multiple-shot), phase encoding orientation, and parallel imaging factor. These parameters significantly affect the sensitivity of the DWI sequence to distortion from rectal gas. In another embodiment, acquisition parameters can be automatically retrieved from the protocol settings of the MRI scanner, thereby ensuring that the predictive model uses the actual parameters to be used for the upcoming DWI acquisition.

[0068] At operation 318, the method includes predicting the probability of DWI distortion based on morphological parameters and acquisition parameters. In one implementation, a multivariate logistic regression model can be used to predict the probability of DWI distortion, wherein the model is trained on a dataset of cases with known DWI distortion results. The model can use inputs such as the minimum surface distance between the gas and the prostate (BDmin) and the total gas volume (GasVol_cc), where formulas include: In another implementation, machine learning models, such as random forests or neural networks, can be used to capture the non-linear relationship between input parameters and the probability of distortion.

[0069] At operation 320, the method includes determining whether the predicted probability exceeds a threshold. In one embodiment, the threshold can be determined by a receiver operating characteristic (ROC) analysis of the performance of the predictive model on a validation dataset, where a cutoff value (e.g., >0.52) is selected to balance sensitivity and specificity. This threshold represents the point at which intervention is required to prevent poor image quality. In another embodiment, the threshold may be able to be adjusted based on a specific clinical context or institutional preference, thereby allowing for customization of the system's sensitivity.

[0070] At operation 322, if the predicted probability does not exceed a threshold (from the no path in decision 320), the method includes continuing standard DWI acquisition. In one implementation, the system can continue planned DWI acquisition using default protocol parameters, as the risk of significant artifacts is considered acceptablely low. This allows the workflow to continue without interruption when analysis indicates that rectal gas is unlikely to cause problematic distortion in DWI images. In another implementation, the system can still display information about the detected gas to the technician, but without an emergency alarm, thus allowing the technician to make an informed decision regarding whether to continue standard acquisition or make optional adjustments.

[0071] At operation 324, if the predicted probability exceeds a threshold (the path is determined by decision 320), the method includes generating an alert to the technician regarding potential DWI artifacts. In one embodiment, the alert may be prominently displayed on the scanner console, highlighting areas of rectal gas on the locator image and indicating a high probability of DWI distortion. The alert may include specific information about the gas volume and its proximity to the prostate to help the technician understand the severity of the potential problem. In another embodiment, the alert may be integrated into the scanner workflow, requiring confirmation from the technician before continuing DWI acquisition.

[0072] At operation 326, following the generation of an alarm, the method includes recommending optimization of scan parameters to mitigate potential artifacts. In one implementation, the system may advise the patient to take preparatory measures, such as administering a microenema to reduce rectal gas before continuing the scan. Alternatively, the system may recommend switching to a more robust DWI acquisition protocol, such as PROPELLER diffusion-weighted imaging, which is less susceptible to susceptible artifacts. In another implementation, the system may provide multiple optimization options and their estimated validity, allowing technicians to select the most appropriate intervention based on the specific clinical situation and available resources.

[0073] Method 300 offers several technical benefits over conventional methods. By automatically detecting and analyzing rectal gas prior to DWI acquisition, the system enables technicians to take corrective actions based on patient preparation or DWI acquisition parameter adjustments, reducing the need for repeat examinations. This automated approach eliminates reliance on human judgment and experience, providing a consistent assessment of potential DWI distortions across different operators and facilities. Furthermore, by analyzing both morphological parameters (such as gas volume and distance from the prostate) and acquisition parameters, the system provides more accurate predictions of potential artifacts than visual examination alone. Method 300 also addresses the issue of gas movement during examination by enabling DWI acquisition immediately after locator scanning when used with an automated scan planning system. This ultimately improves workflow efficiency, enhances the diagnostic quality of prostate MRI studies, and increases the likelihood that DWI sequences will have sufficient quality to facilitate proper reading and risk scoring according to PI-RADS guidelines.

[0074] refer to Figure 4 The diagram illustrates a flowchart of a method 400 for segmenting rectal gas regions in MRI images. Method 400 provides an automated approach for detecting and segmenting rectal gas regions within a region of interest surrounding the prostate in magnetic resonance imaging (MRI) images. Method 400 begins by receiving a locator MRI image and proceeds through various processing steps to identify and segment the rectal gas regions, ultimately providing a mask of the segmented rectal gas regions that can be used for subsequent analysis of potential diffusion-weighted imaging (DWI) distortion.

[0075] At operation 402, the method includes receiving locator MRI images of the subject. In one embodiment, the locator MRI images may include a three-plane locator for scan planning, such as sagittal, coronal, and axial views. These locator images are typically acquired at the start of a prostate MRI examination and serve as the basis for subsequent scan planning and analysis. In another embodiment, locator MRI images may be acquired immediately prior to DWI acquisition to minimize the impact of gas movement due to natural peristalsis, as gas not present during the initial locator scan may move into the rectum during subsequent sequences, potentially affecting DWI quality.

[0076] At operation 404, the method includes applying Standardized Contrast Normalization (SCNe) to normalize the locator MRI image. In one embodiment, SCNe can be used for contrast normalization to remove radio frequency (RF) bias in the locator data and normalize thresholds used in subsequent segmentation steps. This normalization helps ensure consistent performance across different scanning protocols and devices by standardizing image intensity values. In another embodiment, contrast normalization may include histogram equalization or other intensity normalization techniques to account for variations in image acquisition parameters, thereby ensuring consistent subsequent thresholding operations regardless of the original image contrast characteristics.

[0077] At operation 406, the method includes determining a region of interest (ROI) surrounding the prostate. In one embodiment, the ROI may be defined as extending a predetermined distance from the prostate segmentation mask along multiple directions (such as 50 mm upward, 20 mm downward, 10 mm forward, and 80 mm backward). This ROI focuses subsequent analysis on the area most likely to contain rectal gas that could affect DWI quality. In another embodiment, the ROI may be dynamically determined based on the size and location of the prostate, thereby scaling the extension distance to ensure adequate coverage of the rectal region adjacent to the prostate while minimizing computational requirements by excluding irrelevant areas.

[0078] At operation 408, the method includes calculating the average signal intensity in the region of interest (ROI) surrounding the prostate. In one embodiment, the average signal intensity can be calculated by summing the normalized signal intensities of all pixels within the ROI (which includes both the prostate and surrounding tissue) and dividing by the total number of pixels in the ROI. This average is used as a reference for determining a threshold for gas segmentation in subsequent steps. In another embodiment, a weighted method can be used to calculate the average signal intensity, giving more importance to regions within the ROI that are more likely to be affected by artifacts caused by rectal gas in the DWI sequence, such as posterior regions adjacent to the rectum.

[0079] At operation 410, the method includes setting a gas threshold based on the average signal intensity determined at operation 408. In one embodiment, the gas threshold may be set as a fixed percentage of the average prostate ROI signal intensity, such as 0.3 times the average intensity. This method utilizes the characteristic appearance of gas as a low-signal (dark) region relative to prostate tissue in MRI. In another embodiment, the gas threshold may be adaptively determined based on both the average prostate signal intensity and the overall image histogram, thereby allowing for more robust segmentation across different scanning parameters and patient anatomy.

[0080] At operation 412, the method includes identifying all pixels within the ROI that have a normalized signal intensity below a gas threshold. In one implementation, a binary mask can be created where pixels with a normalized intensity below the threshold are assigned a value of 1 (indicating gas), and all other pixels are assigned a value of 0. This binary segmentation method provides a clear depiction between the gas region and the surrounding tissue. In another implementation, a probabilistic method can be used where each pixel is assigned a probability of being gas based on its distance from the threshold, thereby allowing for finer segmentation in boundary regions where partial volumetric effects may occur.

[0081] At operation 414, the method includes applying post-processing to remove small regions of segmented gas smaller than a predetermined number of pixels / voxels. In one embodiment, a connected component analysis can be performed on the binary gas mask, and segmented gas regions smaller than a predetermined size (e.g., 100 voxels) can be removed because these small gas regions are unlikely to cause significant DWI distortion and may represent noise or segmentation errors. This filtering step helps to focus the analysis on gas sacs that are large enough to potentially affect image quality. In another embodiment, morphological operations such as erosion and dilation can be applied to refine gas region boundaries and connect nearby gas sacs that may represent single, continuous gas regions, thereby improving the anatomical accuracy of segmentation.

[0082] At operation 416, the method includes outputting the segmented rectal gas regions. In one embodiment, the output may be in the form of a binary mask identifying the location of rectal gas within the ROI surrounding the prostate. This mask can then be used for subsequent analyses, such as calculating morphological parameters, like the minimum surface distance between the gas and the prostate or the total gas volume. In another embodiment, the output may include both the binary mask and quantitative measurements of the segmented gas regions, such as volume, distance from the prostate, and shape characteristics, which can be directly used to predict the probability of DWI distortion.

[0083] Method 400 offers several technical benefits superior to conventional methods. By automatically detecting and segmenting rectal gas regions in locator MRI images, the method enables subsequent analysis of the potential impact on DWI quality without manual intervention. The use of contrast normalization ensures consistent performance across different scanning protocols and devices, while the ROI-based approach optimizes processing efficiency by focusing computational resources on relevant anatomical regions. Adaptive thresholding based on prostate ROI signal intensity accounts for variations in image acquisition parameters with minimal computational complexity, and post-processing steps help eliminate false positives and improve segmentation accuracy through efficient connected component analysis. This automated approach eliminates reliance on human judgment and experience, providing a consistent assessment of rectal gas presence across different operators and facilities, while reducing computational requirements by targeting only the most relevant image regions and using computationally efficient threshold-based gas region segmentation. Selective filtering of small gas regions further simplifies the computational pipeline by removing areas unlikely to cause significant DWI distortion, ultimately contributing to improved workflow efficiency, reduced processing time, and enhanced diagnostic quality in prostate MRI studies.

[0084] refer to Figure 5 The diagram illustrates a flowchart of a method 500 for generating input features for a predictive model using prostate and rectal gas region data. Method 500 provides an automated approach for calculating morphological parameters based on prostate segmentation masks and rectal gas region segmentation masks, and encoding these parameters, along with acquisition parameters, into input features for a predictive model. The predictive model can be used to predict the probability of DWI distortion due to rectal gas, enabling technicians to take corrective actions prior to DWI acquisition. Method 500 begins by receiving segmented data and proceeds through various processing steps to generate input features for the predictive model.

[0085] At operation 502, the method includes receiving a segmented prostate mask and segmented gas regions. In one embodiment, the segmented prostate mask can be generated using a deep learning-based model specifically developed for prostate segmentation on sagittal and coronal locators. The segmentation model can output a binary mask identifying prostate regions within the locator image. In another embodiment, the method can utilize existing prostate segmentation models, such as those used in automated scan planning systems, to identify the boundaries of the prostate. The segmented gas regions can be... Figure 4 The process described in the document is used to generate the signal, which includes applying contrast normalization, determining the ROI around the prostate, calculating the average signal intensity, setting a gas threshold, identifying pixels below the threshold, and applying post-processing to remove small regions.

[0086] At operation 504, the method includes calculating the minimum surface distance between the segmented rectal gas region and the prostate. In one embodiment, a distance transformation technique (such as Hausdorff distance) can be used to calculate the minimum surface distance, which measures the maximum distance required to travel from a point in one set to the nearest point in another set. This distance metric provides a quantitative measure of how close the gas pocket is to the prostate, which is a factor in determining the likelihood of susceptible artifacts in DWI.

[0087] At operation 506, the method includes calculating the total gas volume within the region of interest (ROI) surrounding the prostate. In one embodiment, the total gas volume can be calculated by summing the volumes of all identified gas regions within the ROI, where the volume is expressed in cubic centimeters. This calculation provides a quantitative measurement of the amount of gas present near the prostate, which is another relevant factor in determining the likelihood of susceptible artifacts in DWI. In another embodiment, the gas volume calculation may include additional processing to account for partial volume effects at the boundaries of gas regions, thereby providing a more accurate estimate of the total gas volume.

[0088] At operation 508, the method includes retrieving acquisition parameters for the DWI sequence. In one embodiment, acquisition parameters may include echo interval, phase encoding line, B0 inhomogeneity, field strength, acquisition scheme (e.g., single-shot EPI, FOCUS, multiple-shot), phase encoding orientation, and parallel imaging factor. These parameters significantly affect the sensitivity of the DWI sequence to distortion from rectal gas. In another embodiment, acquisition parameters can be automatically retrieved from the protocol settings of the MRI scanner, thereby ensuring that the predictive model uses the actual parameters to be used for the upcoming DWI acquisition.

[0089] At operation 510, the method includes encoding morphological parameters and acquisition parameters as input features of a prediction model. In one implementation, the morphological parameters (minimum surface distance and total gas volume) and acquisition parameters may be encoded as input features of a multivariate logistic regression model used to predict the probability of DWI distortion. The logistic regression model can be trained on a dataset of cases with known DWI distortion results, where the formulas are such as: In another implementation, the encoded parameters can be used as input features for a machine learning model, such as a random forest or a neural network, which may be able to capture the non-linear relationship between the input parameters and the probability of distortion.

[0090] Method 500 offers several technical benefits superior to conventional methods. By using distance transformation techniques such as Hausdorff distance to calculate the minimum surface distance between the gas and the prostate, the method enables precise quantification of the spatial proximity between these structures, a major determinant of susceptibility artifacts. Similarly, the calculation of the total gas volume within a specifically defined ROI around the prostate provides a targeted measurement of the gas presence in the most relevant anatomical region, rather than considering all gas in the field of view. These specific morphological parameters, along with the encoding of acquisition parameters such as echo interval, phase encoding line, B0 inhomogeneity, field strength, and acquisition scheme, create a comprehensive feature set that captures the most significant parameters influencing susceptibility artifact formation. This selection of parameters considers both the physical proximity of the gas to the prostate and the total gas load, which, together with the acquisition parameters, determine the severity of susceptibility artifacts in the DWI sequence.

[0091] refer to Figure 6 A flowchart of a method 600 for managing MRI scan protocols based on gas analysis and distortion probability is shown. Method 600 provides an automated approach for alerting technicians to potential diffuse DWI artifacts caused by rectal gas and facilitating appropriate interventions to optimize scan quality. Method 600 begins by receiving DWI distortion probability and gas analysis results and proceeds through various steps to update the scan protocol based on technician-selected interventions.

[0092] At operation 602, the method includes receiving the DWI distortion probability and gas analysis results. In one embodiment, the DWI distortion probability can be generated by a distortion prediction model of analytical morphological parameters (such as the minimum surface distance between the gas and the prostate and the total gas volume) and acquisition parameters (such as echo interval, phase encoding lines, and field strength). The gas analysis results can be included using a combination of... Figure 4 The described method calculates the segmented rectal gas region within the region of interest (ROI) surrounding the prostate. In another implementation, the DWI distortion probability can be received from a multivariate logistic regression model trained on a dataset of cases with known DWI distortion results.

[0093] At operation 604, the method includes displaying an alarm on the scanner console showing a gas region highlighted on the locator image, the gas volume measurement result, the distance from the prostate, and the predicted probability of distortion. In one embodiment, an alarm may be prominently displayed on the scanner console when the predicted DWI distortion probability exceeds a threshold (e.g., >0.52), wherein the gas region is highlighted in contrasting colors on the locator image to clearly indicate its position relative to the prostate. The alarm may include specific quantitative information about the gas volume and its minimum distance from the prostate to help technicians understand the severity of the potential problem. In another embodiment, the alarm may include a visual representation of the predicted distortion probability, such as a gauge or meter indicating the likelihood of significant artifacts in an upcoming DWI acquisition, which has color coding (e.g., green for low probability, yellow for medium probability, and red for high probability) to facilitate rapid interpretation by technicians.

[0094] At operation 606, the method includes presenting intervention options to the technician. In one embodiment, intervention options may include patient preparation measures, such as administering a microenema to reduce rectal gas before continuing the scan, repositioning the patient to potentially remove the gas from the prostate, or switching to a more robust DWI acquisition protocol less susceptible to susceptible artifacts, such as PROPELLER diffusion-weighted imaging. Each option may be presented with estimated effectiveness and time requirements to assist the technician in making an informed decision. In another embodiment, intervention options may be dynamically generated based on the specific characteristics of the detected gas, wherein the system recommends the most appropriate intervention based on factors such as gas volume, proximity to the prostate, and available acquisition protocols at the specific facility.

[0095] At operation 608, the method includes waiting for technician confirmation and selection. In one implementation, the system may require explicit confirmation from the technician before continuing DWI acquisition to ensure that potential problems are addressed rather than ignored. The technician can use the scanner console interface to select one of the presented intervention options, or choose to continue standard acquisition regardless of warnings if clinical judgment indicates that artifacts may not significantly affect diagnostic quality. In another implementation, the system may implement a timeout feature that automatically selects a default intervention option if no selection is made within a predetermined time period, thus balancing the need for technician input with workflow efficiency considerations.

[0096] At operation 610, the method includes recording the selected intervention and timestamps. In one implementation, the system can record the technician's selection along with the timestamps, gas analysis results, and predicted distortion probabilities in a structured log, which can be reviewed later for quality assurance and process improvement purposes. This recording capability allows the facility to track the frequency of rectal gas-related problems and the effectiveness of various interventions over time. In another implementation, the recorded information can be automatically incorporated as technical notes into the radiology report, informing the radiologist interpreting the image about potential image quality problems and the steps taken to resolve them, which can be valuable background information for image interpretation.

[0097] At operation 612, the method includes updating the scanning protocol based on the selected intervention. In one implementation, if the technician selects an alternative acquisition protocol (such as PROPELLER DWI), the system can automatically update the scanning protocol parameters accordingly, ensuring the use of an appropriate sequence without requiring manual parameter adjustments by the technician. If a patient preparation measure is selected, the system can pause the scanning queue to allow time for intervention before resuming the standard protocol. In another implementation, the system can perform adaptive protocol adjustments based on the specific characteristics of the detected gas, such as modifying the phase encoding direction to minimize the effects of susceptible artifacts, or adjusting the parallel imaging factor to reduce distortion while maintaining a sufficient signal-to-noise ratio.

[0098] refer to Figure 7 The figure illustrates a visual representation of the contrast normalization process used to enhance medical images. It shows the transformation from an initial medical image 702 to a contrast-normalized medical image 704 by applying contrast normalization process 706. The initial medical image 702 and the contrast-normalized medical image 704 are sagittal views of the pelvic region, which can be used to detect and analyze the impact of rectal gas on the diffusion-weighted imaging (DWI) quality in prostate MRI examinations.

[0099] The initial medical image 702 is a locator MRI image acquired during the subject's scan planning. In one embodiment, the initial medical image 702 may be a sagittal locator image obtained at the start of a prostate MRI examination, showing the anatomical structures including the prostate, rectum, and surrounding tissues, where the contrast between different tissue types is relatively low. In another embodiment, the initial medical image 702 may be acquired immediately prior to DWI acquisition to minimize the impact of gas movement due to natural peristalsis, as gas not present during the initial locator scan may move into the rectum during subsequent sequences. The initial medical image 702 typically exhibits radio frequency (RF) bias and non-normalized intensity values, which may be used to complicate subsequent segmentation and analysis steps for detecting rectal gas regions that may potentially affect the quality of DWI images.

[0100] The contrast normalization process 706 applies a standardized contrast normalization (SCNe) to the initial medical image 702 to remove radio frequency bias and normalize image intensity values. In one embodiment, the contrast normalization process 706 may be performed as part of an automated workflow that includes prostate segmentation, region of interest (ROI) determination, and rectal gas detection. In another embodiment, the contrast normalization process 706 may include histogram equalization or other intensity normalization techniques to account for variations in image acquisition parameters. The contrast normalization process 706 helps ensure consistent performance across different scanning protocols and devices by normalizing image intensity values, which is beneficial for subsequent thresholding operations used to identify rectal gas regions. This process can be performed using an image processing device (such as...) Figure 2 The instructions in the non-transitory memory of the image processing device shown are implemented by one or more processors.

[0101] The contrast-normalized medical image 704 is the output of the contrast normalization process 706, showing enhanced contrast and normalized intensity values. In one embodiment, the contrast-normalized medical image 704 can be used directly for subsequent processing steps, including identifying the region of interest (ROI) around the prostate, calculating the average signal intensity in the prostate region, setting a gas threshold based on the average prostate signal intensity, and identifying pixels within the ROI with normalized signal intensities below the threshold to segment rectal gas regions. In another embodiment, the contrast-normalized medical image 704 can be displayed to a technician along with the segmentation results to provide visual confirmation of detected rectal gas regions. The enhanced contrast in the contrast-normalized medical image 704 improves the visibility of anatomical structures and tissue boundaries, thereby facilitating more accurate segmentation of both the prostate and rectal gas regions.

[0102] By standardizing image intensity values, the contrast normalization process 706 enables more reliable thresholding of gas segmentation, where pixels with normalized intensities below a certain threshold are classified as gas. This method leverages the characteristic appearance of gas as a low-signal (dark) region relative to prostate tissue in MRI. The contrast normalization process 706 is connected to subsequent processing steps, including gas segmentation, morphological parameter calculation, and DWI distortion probability prediction, forming an integrated workflow for the automated detection and analysis of rectal gas prior to DWI acquisition, enabling timely intervention and optimization of scan parameters.

[0103] refer to Figure 8 The figure illustrates a visualization of the prostate segmentation process, showing the transformation from an input medical image to a segmented output image. The figure demonstrates the transformation from an initial medical image 802 to a segmented medical image 804 by applying the prostate segmentation process 806. The initial medical image 802 and the segmented medical image 804 are locator MRI images of the pelvic region, which can be used to detect and analyze the impact of rectal gas on the DWI quality in prostate MRI examinations.

[0104] The initial medical image 802 is a locator MRI image acquired during the subject's scan planning. In one embodiment, the initial medical image 802 may be a sagittal locator image obtained at the start of a prostate MRI examination, showing the anatomical structures including the prostate, rectum, and surrounding tissues. In another embodiment, the initial medical image 802 may be acquired immediately prior to DWI acquisition to minimize the impact of gas movement due to natural peristalsis, as gas not present during the initial locator scan may move into the rectum during subsequent sequences.

[0105] The prostate segmentation process 806 applies a deep learning-based model to the initial medical image 802 to identify and segment the prostate. In one embodiment, the prostate segmentation process 806 may be performed by the prostate segmentation module 208 as part of an automated workflow that includes region of interest (ROI) determination, gas segmentation, and rectal gas detection. In another embodiment, the prostate segmentation process 806 may utilize existing prostate segmentation models, such as those used in automated scan planning systems, to identify the boundaries of the prostate. The prostate segmentation process 806 helps ensure accurate depiction of the prostate boundaries, which is beneficial for subsequent determination of ROIs around the prostate and for calculating morphological parameters, such as the minimum surface distance between rectal gas and the prostate.

[0106] The segmented medical image 804 is the output of the prostate segmentation process 806, showing the prostate highlighted on the original image using color overlay. In one embodiment, the segmented medical image 804 may include a binary mask identifying regions of the prostate within the locator image, which is then used in subsequent processing steps, including determining the region of interest (ROI) around the prostate, segmenting rectal gas regions within the ROI, and calculating morphological parameters based on the spatial relationship between the prostate and any detected rectal gas. In another embodiment, the segmented medical image 804 may be displayed to a technician along with the gas segmentation results to provide visual confirmation of the anatomical relationship between the prostate and the detected rectal gas regions. By accurately segmenting the prostate, the prostate segmentation process 806 enables more precise determination of the ROI around the prostate, which focuses subsequent gas segmentation and analysis on the areas most likely to contain rectal gas that may affect DWI quality.

[0107] refer to Figure 9 The figure shows a comparison of anatomical scan images before and after gas segmentation processing. It illustrates a visual representation of the gas segmentation process used to detect rectal gas that can cause susceptible artifacts in DWI of the prostate. The figure includes the original anatomical scan image 902, the processed anatomical scan image 904 with gas segmentation, and the gas segmentation process 906 connecting images 902 and 904.

[0108] The raw anatomical scan image 902 is a sagittal locator MRI image acquired during the subject's scan planning. In one embodiment, the raw anatomical scan image 902 may be a sagittal locator image obtained at the start of a prostate MRI examination, showing the anatomical structures including the prostate, rectum, and surrounding tissues. In another embodiment, the raw anatomical scan image 902 may be acquired immediately prior to DWI acquisition to minimize the impact of gas movement due to natural peristalsis, as gas not present during the initial locator scan may move into the rectum during subsequent sequences. The raw anatomical scan image 902 typically exhibits RF bias and non-normalized intensity values, which may be used to complicate subsequent segmentation and analysis steps for detecting rectal gas regions that may potentially affect DWI image quality.

[0109] The gas segmentation process 906 applies a normalized contrast normalization (SCNe) to the raw anatomical scan image 902 to remove radio frequency bias and normalize image intensity values. In one embodiment, the gas segmentation process 906 may include pixels identifying a region of interest (ROI) surrounding the prostate with a normalized signal intensity below a predetermined threshold, where the threshold is set as a proportion of the average signal intensity in the prostate region. This approach leverages the characteristic appearance of gas as a low-signal (dark) region relative to prostate tissue in MRI. In another embodiment, the gas segmentation process 906 may employ a CNN for gas segmentation, potentially providing improved resilience to segmentation errors and better adaptability to various image characteristics. The gas segmentation process 906 may also include a post-processing step to remove regions smaller than a predetermined number of pixels or voxels (e.g., 100 voxels), as these small regions are unlikely to cause significant DWI distortion and may represent noise or segmentation errors.

[0110] The processed anatomical scan image 904 is the output of the gas segmentation process 906, showing enhanced contrast and segmented gas regions. In one embodiment, the processed anatomical scan image 904 may include a highlighted region representing segmented rectal gas pockets that may potentially cause susceptible artifacts in DWI sequences. In another embodiment, the processed anatomical scan image 904 can be used for subsequent processing steps, including calculating morphological parameters, such as the minimum surface distance between the gas and the prostate or the total gas volume, which can be used to predict the probability of DWI distortion. The enhanced contrast in the processed anatomical scan image 904 improves the visibility of anatomical structures and tissue boundaries, thereby facilitating more accurate segmentation of both the prostate and rectal gas regions.

[0111] refer to Figure 10 The image shows a medical imaging visualization displaying the ROI selection. The visualization includes a full anatomical view image 1002 containing the highlighted ROI selection 1006, which is shown in more detail in a magnified ROI view 1004.

[0112] The complete anatomical view image 1002 may be a locator MRI image acquired during a prostate MRI examination. In one embodiment, the complete anatomical view image 1002 may be a sagittal locator image showing a pelvic region including the prostate and surrounding anatomical structures. In another embodiment, the complete anatomical view image 1002 may be an axial T2-weighted image obtained prior to DWI acquisition. Contrast normalization techniques (such as SCNe) may be used to process the complete anatomical view image 1002 to normalize image intensity across different acquisitions and remove RF bias from the locator data, thereby enabling more consistent threshold-based segmentation of anatomical structures.

[0113] ROI selection 1006 identifies a specific region around the prostate for detailed analysis. In one embodiment, ROI selection 1006 may be automatically determined by ROI determination module 210 based on prostate segmentation performed by prostate segmentation module 208. ROI selection 1006 may extend a predetermined distance from the prostate mask. In another embodiment, ROI selection 1006 may be manually defined by a user via user input device 222. ROI selection 1006 serves as a boundary within which rectal gas regions are identified and analyzed to determine their potential impact on DWI image quality.

[0114] The magnified ROI view 1004 provides a magnified visualization of the region defined by ROI selection 1006. In one embodiment, the magnified ROI view 1004 may display the segmented prostate in one color and any detected rectal gas areas in another color. In another embodiment, the magnified ROI view 1004 may include overlay information showing the minimum surface distance between the prostate and the nearest rectal gas pocket. The magnified ROI view 1004 may also display calculated morphological parameters, such as the minimum boundary distance (BDmin) between the gas pocket and the prostate, and the total gas volume within the ROI. Distortion prediction model 218 uses these parameters to calculate the probability of DWI distortion due to the presence of rectal gas.

[0115] This disclosure also provides support for a method comprising: acquiring a locator magnetic resonance imaging (MRI) image of a subject; segmenting organs in the locator MRI image to create an organ segmentation mask; identifying a region of interest (ROI) surrounding the organ; applying contrast normalization to normalize the locator MRI image; segmenting a gas region within the ROI surrounding the organ; calculating morphological parameters based on the organ segmentation mask and the gas region; determining acquisition parameters for diffusion-weighted imaging (DWI); predicting a DWI distortion probability based on the morphological parameters and the acquisition parameters; generating an alert to a technician regarding potential DWI artifacts when the DWI distortion probability exceeds a threshold; automatically adjusting scan parameters when the DWI distortion probability exceeds the threshold; and acquiring a medical image based on the adjusted scan parameters. In a first example of the method, segmenting the gas region includes: calculating the average signal intensity in the ROI surrounding the organ; setting a gas threshold based on the average signal intensity; identifying pixels within the ROI with a normalized signal intensity below the gas threshold; and applying post-processing to remove regions with fewer than a predetermined number of pixels. In a second example, which optionally includes the first example, calculating the morphological parameters includes: calculating the minimum surface distance between the segmented gas region and the organ, and calculating the total gas volume within the ROI surrounding the organ. In a third example, which optionally includes one or both of the first and second examples, predicting the DWI distortion probability includes inputting the morphological parameters and acquisition parameters into a logistic regression model. In a fourth example, which optionally includes one or more of the first to third examples, the logistic regression model is trained using multiple training cases with known DWI distortion results. In a fifth example, which optionally includes one or more of the first to fourth examples, determining the ROI surrounding the organ includes defining a region extending upwards, downwards, forwards, and backwards by predetermined distances from the organ segmentation mask. In a sixth example, which optionally includes one or more of the first to fifth examples, the acquisition parameters include at least one of the following: echo interval, phase encoding line, B0 inhomogeneity, field strength, acquisition scheme, phase encoding direction, and parallel imaging factor. In a seventh example, which optionally includes one or more of the first to sixth examples, the method further includes: suggesting optimization of scan parameters when the DWI distortion probability exceeds a threshold.

[0116] This disclosure also provides support for a system comprising: one or more processors, and a non-transitory memory storing instructions executable by the one or more processors to: acquire an initial locator magnetic resonance imaging (MRI) image of a subject; segment the prostate in the initial locator MRI image; determine a region of interest (ROI) surrounding the segmented prostate; apply contrast normalization to normalize the initial locator MRI image; segment rectal gas regions within the ROI surrounding the prostate; selectively filter the segmented rectal gas regions; calculate morphological parameters based on the segmented prostate and the segmented rectal gas regions; determine acquisition parameters for diffusion-weighted imaging (DWI); predict a DWI distortion probability based on the morphological parameters and the acquisition parameters; and generate an alert to a technician when the predicted DWI distortion probability exceeds a threshold. In a first example of the system, the instructions for segmenting the rectal gas regions include instructions for: calculating an average signal intensity in the segmented prostate; setting a gas threshold based on the average signal intensity in the segmented prostate; and identifying pixels within the ROI having a normalized signal intensity below the gas threshold. In a second example, which optionally includes the first example, the instructions for selectively filtering the segmented rectal gas regions include instructions for removing regions smaller than a predetermined number of pixels. In a third example, which optionally includes one or both of the first and second examples, the morphological parameters include the minimum surface distance between the segmented rectal gas regions and the segmented prostate, and the total gas volume within the ROI. In a fourth example, which optionally includes one or more of the first to third examples, the instructions for calculating the minimum surface distance include instructions for using a Hausdorff distance filter. In a fifth example, which optionally includes one or more of the first to fourth examples, the instructions for predicting the DWI distortion probability include instructions for encoding the morphological parameters and acquisition parameters as input features for a prediction model, and applying a logistic regression model to the input features to generate the DWI distortion probability. In a sixth example, which optionally includes one or more of the first to fifth examples, a logistic regression model is trained using multiple training cases, each training case including morphological parameters, acquisition parameters, and a set of corresponding labels indicating whether DWI distortion has occurred. In a seventh example, which optionally includes one or more of the first to sixth examples, the instructions also include instructions for suggesting adjustments to scan parameters in response to a predicted DWI distortion probability exceeding a threshold.

[0117] This disclosure also provides support for a method for predicting diffusion-weighted imaging (DWI) distortion in prostate magnetic resonance imaging (MRI), the method comprising: acquiring a locator MRI image of a subject; segmenting the prostate in the locator MRI image using a deep learning-based model to create a prostate segmentation mask; identifying regions of interest (ROIs) extending a predetermined distance from the prostate segmentation mask along multiple directions; applying contrast normalization to remove radio frequency bias in the locator MRI image; calculating the average signal intensity in the ROI surrounding the prostate; setting a gas threshold as a percentage of the average signal intensity; identifying pixels within the ROI with normalized signal intensity below the gas threshold to create a binary gas mask; applying connected component analysis to remove gas regions smaller than a predetermined size; calculating the minimum surface distance between the remaining gas regions and the prostate; calculating the total gas volume within the ROI; retrieving acquisition parameters for a planned DWI sequence; encoding the minimum surface distance, total gas volume, and acquisition parameters as input features for a prediction model; applying the prediction model to generate a probability of DWI distortion; and displaying an alarm on a display device when the probability of DWI distortion exceeds a threshold. In a first example of the method, the method further includes: in response to the probability of DWI distortion exceeding a threshold: displaying the remaining gas area on a display device and suggesting an alternative DWI acquisition scheme that is less susceptible to susceptible artifacts. In a second example of the method that optionally includes the first example, predicting the probability of DWI distortion includes applying a statistical model that correlates the minimum surface distance, total gas volume, and acquisition parameters with known DWI distortion results from a training dataset. In a third example of the method that optionally includes one or both of the first and second examples, the predictive model includes a multivariate logistic regression model trained on a dataset of cases with known DWI distortion results.

[0118] When describing elements of various embodiments of this disclosure, the articles “a,” “an,” and “the” are intended to indicate the presence of one or more such elements. The terms “first,” “second,” etc., do not indicate any order, quantity, or importance, but are used to distinguish one element from another. The terms “comprising,” “including,” and “having” are intended to be inclusive and indicate that additional elements may exist in addition to the listed elements. As used herein, the terms “connected to,” “coupled to,” etc., indicate that an object (e.g., a material, element, structure, component, etc.) may be connected to or coupled to another object, regardless of whether the object is directly connected to or coupled to the other object, or whether one or more intermediary objects exist between the two objects. Furthermore, it should be understood that references to “an embodiment” or “an embodiment” of this disclosure are not intended to be construed as excluding the existence of additional embodiments also incorporating the referenced features.

[0119] In addition to any modifications previously indicated, those skilled in the art can devise many other variations and alternative arrangements without departing from the spirit and scope of this specification, and the appended claims are intended to cover such modifications and arrangements. Therefore, although the information has been described in particular and detail above in conjunction with what is now considered to be the most practical and preferred aspects, it will be apparent to those skilled in the art that many modifications, including but not limited to those in form, function, mode of operation, and purpose, can be made without departing from the principles and concepts set forth herein. Likewise, as used herein, embodiments and implementations are intended to be illustrative in all respects and should not be construed as limiting in any way.

Claims

1. A method (300), the method comprising: Positioner magnetic resonance imaging (MRI) images (302) of subject (16) were acquired; Segmenting organs in the locator MRI image to create an organ segmentation mask (304); Determine the region of interest (ROI) surrounding the organ (306); Contrast normalization (706) is applied to normalize the MRI images of the locator (308). Segment the gas region (906) (310) within the ROI surrounding the organ; Morphological parameters (314) are calculated based on the organ segmentation mask and the gas region (906). Determine the acquisition parameters for diffusion-weighted imaging (DWI) (316); The distortion prediction model (218) is used to predict the DWI distortion probability based on the morphological parameters and the acquisition parameters (318). When the DWI distortion probability exceeds the threshold, an alert (324) is generated for the technician regarding potential DWI artifacts. When the DWI distortion probability exceeds the threshold, the scanning parameters (326) are automatically adjusted; and Medical images are acquired based on the adjusted scanning parameters.

2. The method (300) according to claim 1, wherein segmenting the gas region (906) comprises: Calculate the average signal intensity in the ROI surrounding the organ; The gas threshold is set based on the average signal strength. Identify pixels within the ROI that have a normalized signal strength below the gas threshold; as well as Post-processing is applied to remove regions smaller than a predetermined number of pixels.

3. The method (300) according to claim 1, wherein calculating the morphological parameters includes: Calculate the minimum surface distance between the segmented gas region (906) and the organ; as well as Calculate the total gas volume within the ROI surrounding the organ.

4. The method (300) according to claim 1, wherein predicting the DWI distortion probability includes inputting the morphological parameters and the acquisition parameters into the distortion prediction model (218).

5. The method (300) of claim 4, wherein the distortion prediction model (218) is trained using a plurality of training cases with known DWI distortion results.

6. The method (300) of claim 1, wherein determining the ROI around the organ includes defining a region extending upward, downward, forward, and backward a predetermined distance from the organ segmentation mask.

7. The method (300) according to claim 1, wherein the acquisition parameters include at least one of the following: echo interval, phase coding line, BO inhomogeneity, field strength, acquisition scheme, phase coding direction and parallel imaging factor.

8. The method (300) of claim 1, wherein the alarm includes a suggested adjustment of the scan parameters based on the DWI distortion probability.

9. A system (200) comprising: One or more processors (204); and Non-transitory memory (206) stores instructions that can be executed by the one or more processors (204) to: Initial locator magnetic resonance imaging (MRI) images (702) of subject (16) were obtained; The prostate in the initial locator MRI image (702) is segmented using the prostate segmentation module (208); The Region of Interest (ROI) determination module (210) is used to determine the ROI around the segmented prostate. Contrast normalization (706) is applied to normalize the initial locator MRI image (702); The gas segmentation module (212) is used to segment the rectal gas region (906) within the ROI surrounding the prostate. The gas filtration module (214) is used to selectively filter the segmented rectal gas region (906). The morphological parameter determination module (216) calculates morphological parameters based on the segmented prostate and the segmented rectal gas region (906); Determine the acquisition parameters used for diffusion-weighted imaging (DWI); The distortion prediction model (218) is used to predict the DWI distortion probability based on the morphological parameters and the acquisition parameters; and When the predicted DWI distortion probability exceeds the threshold, an alarm is generated for the technician using the alarm generation module (220).

10. The system (200) of claim 9, wherein the instructions for segmenting the rectal gas region (906) include instructions for: Calculate the average signal intensity in the segmented prostate; The gas threshold is set based on the average signal intensity in the segmented prostate; and Identify pixels within the ROI that have a normalized signal strength below the gas threshold.

11. The system (200) of claim 9, wherein the instructions for selectively filtering the segmented rectal gas region (906) include instructions for removing regions less than a predetermined number of pixels.

12. The system (200) of claim 9, wherein the morphological parameters include the minimum surface distance between the segmented rectal gas region (906) and the segmented prostate and the total gas volume within the ROI.

13. The system (200) of claim 12, wherein the instructions for calculating the minimum surface distance include instructions for using a Hausdorf distance filter.

14. The system (200) of claim 9, wherein the instructions for predicting the DWI distortion probability include instructions for performing the following: The morphological parameters and the acquisition parameters are encoded as input features for the distortion prediction model (218); and The logistic regression model is applied to the input features to generate the DWI distortion probability.

15. The system (200) of claim 14, wherein the distortion prediction model (218) is trained using a plurality of training cases, each training case including morphological parameters, acquisition parameters and a set of corresponding labels indicating whether DWI distortion has occurred.

Images