A system for high precision neurosurgery with advanced neuroimaging analytics

Abstract

The disclosed system provides an integrated neuroimaging analytics platform for comprehensive preoperative neurosurgical planning, intraoperative neurosurgical guidance and post-operative assessment by processing multimodal MRI (structural, diffusion, functional, and angiography) and CT data. The platform performs detailed anatomical characterization by delineating tumor subregions (peritumoral edema, enhancing tumor, and necrotic core), segmenting brain tissues (gray matter, white matter, and CSF), and executing lobe, cortical / subcortical parcellation. Advanced 3D rendering visualizes tumors alongside critical white matter fibre tracts derived from diffusion MRI, while MRA data is used to segment cerebrovascular structures, and functional MRI analysis identifies eloquent cortices associated with motor, speech, and visual functions. All results are integrated within a user-friendly GUI featuring advanced multiplanar slicing and a smart brush for interactive mask editing, complemented by a speech-to-text engine for streamlined analytical reporting. This comprehensive approach facilitates precise and efficient surgical planning, thereby enhancing patient safety and improving clinical outcomes.

Description

[0001] A SYSTEM FOR HIGH PRECISION NEUROSURGERY WITH ADVANCED NEUROIMAGING ANAEYTICS

[0002] PRIORITY CEAIM

[0003]

[0001] The instant patent application is related to and claims priority from the India Provisional Patent Application no: 202421092791, entitled, “A SYSTEM FOR HIGH PRECISION NEUROSURGERY WITH ADVANCED NEUROIMAGING ANALYTICS , Filed on: 27th Nov 2024, which is incorporated in its entirety herewith.

[0004] TECHNICAL FIELD

[0005]

[0002] The present disclosure is in the technical field of a system for high precision, image guided neurosurgery planning with advanced neuroimaging analytics. More particularly, the system comprises multiple software modules for processing MRI and CT images of a human subject, integrated with hardware components that perform various computational tasks and assist in visualizing precise information about subject brain’s intricate structural and functional connectivity. The system provides advanced visualization tools with rendered 2D and 3D views across all planes for preoperative neurosurgical planning, representing a significant advancement toward minimizing the risk of postsurgical neurological deficits.

[0006] BACKGROUND OF THE DISCLOSED EMBODIMENT

[0007]

[0003] In the current state of art, visualization is limited to 2D slices. Advanced visualization techniques that integrate 2D slices with 3D structures such as tumor, tissue, fibre tracts, cerebrovasculature, parcellated cortical and subcortical regions and functionally active regions- are currently lacking.

[0008]

[0004] Assessment of size, location, and volumetric quantification of intracranial volume (ICV), gray matter (GM), white matter (WM), cerebrospinal fluid (CSF), and parcellated cortical and subcortical regions are currently inadequate.

[0009]

[0005] The currently available neuroimaging analytical technologies identifies location, size and volume of the tumor or specific anatomical structure by performing manual segmentation of MRI images. Manual segmentation is labour intensive and prone to inter reader variability.

[0010]

[0006] Generation of white matter fibre tracts suffers several limitations due to poor signal quality, motion artifact, premature termination of fibres, crossing and fanning of the fibres. Further, complex brain structures such as corpus callosum and brainstem where fibres cross, bend or diverge, it is difficult to construct fibre tracts which is essential for clinical point of view where accuracy and reproducibility are important.

[0011]

[0007] Reconstruction of cerebrovascular structure with higher sensitivity and segmented tumor with distinct colour coding for tumor subregions such as enhancing tumor, non-enhancing tumor necrotic core, edema are not available.

[0012]

[0008] Generation of resting state networks to map critical functional regions and task specific active regions with 3D rendering along with segmented lesion or tumor or region of interest and fibre tracts are indispensable for surgical planning.

[0013]

[0009] 3D rendering and visualization in 3D space with superposition of the above components with high-definition image quality can add precision in surgical planning which in current clinical practices are lacking.

[0014]

[0010] Therefore, there is an urgent need for developing a system for high precision neurosurgery planning with advanced neuroimaging analytics for image guided neurosurgery planning to visualize brain intricate structural and functional connectivity in 2D and 3D rendered view in multiplanar 3D slicer tool.

[0015] SUMMARY OF THE DISCLOSED EMBODIMENT

[0016] [OH] The present disclosure relates to a system for high precision image guided neurosurgery planning platform with advanced neuroimaging analytics. More particularly, the system comprises multiple software modules used on MRI and CT images of a human subject and connected to a hardware for processing tasks, rendering 2D and 3D views and visualization.

[0017]

[0012] According to the aspects of the present disclosure, the system includes a software solution visualized on the computer to support the preoperative neurosurgical planning.

[0018]

[0013] According to one more aspect of the present disclosure, the features of the neuroimaging analytical system include pre-processing of MRI from different MR sequences, delineation of the brain tumor into distinct subregions, parcellation of brain in cortical and subcortical regions, generation of white matter fibre tract, and specific region of interests, segmentation of cerebrovascular system, analysis of resting state fMRI and task-based fMRI, advanced 3D rendering, and graphical user interface (GUI) for visualization.

[0019]

[0014] According to one more aspect of the present disclosure, the disclosed system allows the user to submit Magnetic Resonance Imaging (MRI) images with or without CT image of the subject as input. These images are pre-processed and are subjected to different tasks. The task includes tissue segmentation, autosegmentation of the tumor into distinct subregions such as peritumoral edema, enhancing tumor and necrotic core, blood vessel segmentation, parcellation of cortical and subcortical regions, generation of fibre tract using both tensors based deterministic method and Constrained Spherical Deconvolution (CSD) probabilistic method and resting state functional network. The software tool virtually positioned the lesion or tumor subregions and any other structural or functional region of interest in the brain can be visualized in 2D and 3D space.

[0020]

[0015] According to another aspect of the present disclosure, the system guide dissection or removal of tumors, biopsy of the critical area of lesions or removal of any structure within the cranium or any other associated components the system comprises of multiple tasks or modules for preoperative planning and post-operative assessment.

[0021]

[0016] According to more aspects of the present disclosure, the proposed system determines the major cerebrovascular system of the brain in 2D and 3D space and locates the functional regions of the brain surrounding the lesion or tumor or epileptic region to guide neurosuigeon to perform safer maximal resection that minimizes risk of damaging eloquent brain regions.

[0022]

[0017] Several aspects of the disclosed embodiment are described below with reference to examples for illustration. However, one skilled in the relevant art will recognize that the disclosed embodiment can be practiced without one or more of the specific details or with other methods, components, materials and so forth. In other instances, well-known structures, materials, or operations are not shown in detail to avoid obscuring the features of the disclosed embodiment. Furthermore, the features / aspects described can be practiced in various combinations, though only some of the combinations are described herein for conciseness.

[0023] BRIEF DESCRIPTION OF THE DRAWINGS

[0024]

[0018] The disclosed embodiment describes the accompanying drawings briefly as follows.

[0025]

[0019] FIG. 1 illustrates a control panel of a neurosurgery planning platform (NEURO VIZ) for both admin and user interfaces, which enables monitoring of different functions, according to the aspects of the disclosed embodiment.

[0026]

[0020] FIG. 2 illustrates the steps involved in tumor segmentation into peritumoral edema (ED), enhancing tumor (ET) and necrotic core (NR) according to the aspects of the disclosed embodiment.

[0027]

[0021] FIG. 3 illustrates the location of the tumor subregions on the Tl-weighted (Tlw) image in axial, sagittal, coronal planes, and in 3D sagittal orientation showing whole tumor, tumor core, and necrotic core according to the aspects of the disclosed embodiment.

[0028]

[0022] FIG. 4 illustrates the steps involved in the tissue segmentation from Tlw image into GM, WM, and CSF. For the Tlw image possessing tumor, first tumor core is delineated and the corresponding region is subtracted from WM, GM, and CSF according to the aspects of the disclosed embodiment.

[0029]

[0023] FIG. 5 illustrates the combined mask with the subtracted tumor core region overlaid on the Tlw image in axial, sagittal, coronal and 3D rendered image. Tumor core consisting of enhancing tumor and necrotic core is superposed on combined image, WM, GM, CSF separately, according to the aspects of the disclosed embodiment.

[0030]

[0024] FIG. 6 illustrates the steps involved in atlas based parcellation, lobe based parcellation and model based parcellation, according to the aspects of the disclosed embodiment.

[0031]

[0025] FIG. 7 illustrates the parcellated subcortical, cortical, lobes, model based parcellated regions, according to the aspects of the disclosed embodiment.

[0032]

[0026] FIG. 8 illustrates the flowchart of segmentation of cerebrovascular system from MR angiography (MRA), according to the aspects of the disclosed embodiment.

[0033]

[0027] FIG. 9 illustrates the delineation of blood vessels in axial, sagittal, coronal and 3D rendered views, maximum intensity projection (MIP), delineated vessels in 3D according to the aspects of the disclosed embodiment.

[0034]

[0028] FIG.10 illustrates the steps involved for the construction of white matter fibre tracts from DTI image as per DTI tensor deterministic method and CSD probabilistic method with edema correction, according to the aspects of the disclosed embodiment.

[0035]

[0029] FIG.ll illustrates superposition of Arcuate fasciculus determined using tensor based deterministic method and CSD based probabilistic method on Tlw image. Further, it shows all tumor proximal tracts within 5 mm of the tumor core, according to the aspects of the disclosed embodiment.

[0036]

[0030] FIG. 12 illustrates steps involved in determining resting state networks from rs-fMRI, according to the aspects of the disclosed embodiment.

[0037]

[0031] FIG. 13 illustrates visual, occipital, and auditory resting state networks derived from rs- fMRI, according to the aspects of the disclosed embodiment.

[0038]

[0032] FIG. 14 illustrates steps involved in preprocessing of CT image, according to the aspects of the disclosed embodiment.

[0039]

[0033] FIG. 15A illustrates a visualization page of the NEUROVIZ application that displays loaded images in the left panel, a top panel containing tools for customizing the visualization, and a central workspace divided into four sections for axial, coronal, sagittal, and 3D views. The right panel provides options to display results generated by different processing tasks, while the narrow panel on the far right offers task-specific controls.

[0034] FIG. 15B illustrates the fused CT and MRI visualization, where the skull bone extracted from CT and the white matter derived from MRI are displayed in the same spatial coordinate system. In the axial, coronal, and sagittal planes, the white matter is overlaid on the Tl- weighted image, and the corresponding 3D rendering is shown in the 3D view section, which is controlled by the multiplanar 3D slicing tool.

[0040]

[0035] FIG. 16 illustrates about 2D and 3D edit tools integrated in neurosurgery planning platform (NEURO VIZ). There is smart brush to edit GM, WM, CSF masks or tumor masks and create any region of interest. A 3D edit tool is there to directly edit 3D objects.

[0041]

[0036] In the drawings, like reference numbers indicate specific element which requires explanation. The drawing in which an element first appears is indicated by the leftmost digit(s) in the corresponding reference number.

[0042] DETAILED DESCRIPTION OF THE DISCLOSED EMBODIMENT

[0043]

[0037] It is to be understood that the present disclosure is not limited in its application to the details of construction and the arrangement of components set forth in the following description or illustrated in the drawings. The present disclosure is capable of other embodiments and of being practiced or of being carried out in various ways. Also, it is to be understood that the phraseology and terminology used herein is for the purpose of description and should not be regarded as limiting.

[0044]

[0038] The use of “including”, “comprising” or “having” and variations thereof herein is meant to encompass the items listed thereafter and equivalents thereof as well as additional items. The terms “a” and “an” herein do not denote a limitation of quantity but rather denote the presence of at least one of the referenced items. Further, the use of terms “first”, “second”, and “third”, and the like, herein do not denote any order, quantity, or importance, but rather are used to distinguish one element from another. Further, the terms “comprising,” “including,” “having,” and the like, as used with respect to embodiments of the present disclosure, are synonymous. As used herein, the terms “agent,” “component,” or “ingredient” are used interchangeably and refer to a particular item that includes one or more chemical compounds (e.g., a food item from one or more plants that comprise one or more naturally occurring chemical compounds).

[0045]

[0039] The term “component” or “ingredient” as used herein, refers to a particular item that includes one or more chemical compounds, i.e., edible compounds sourced from plants, fungi or algae that comprise one or more naturally occurring chemical compounds.

[0046]

[0040] Alternative embodiments of the present disclosure and their equivalents may be devised without parting from the spirit or scope of the present disclosure. It should be noted that any discussion herein regarding “one embodiment”, “an embodiment”, “an exemplary embodiment”, and the like indicate that the embodiment described may include a particular feature, structure, or characteristic and that such particular feature, structure, or characteristic may not necessarily be included in every embodiment. In addition, references to the foregoing do not necessarily comprise a reference to the same embodiment. Finally, irrespective of whether, it is explicitly described, one of ordinary skill in the art would readily appreciate that each of the particular features, structures, or characteristics of the given embodiments may be utilized in connection or combination with those of any other embodiment discussed herein.

[0047]

[0041] As used herein, the singular forms “a”, “an”, and “the” include both singular and plural referents unless the context clearly dictates otherwise. By way of example, “a dosage” refers to one or more than one dosage.

[0048]

[0042] The terms “comprising”, “comprises” and “comprised of’ as used herein are synonymous with “including”, “includes” or “containing”, “contains”, and are inclusive or open-ended and do not exclude additional, non-recited members, elements or method steps.

[0049]

[0043] All documents cited in the present specification are hereby incorporated by reference in their totality. In particular, the teachings of all documents herein specifically referred to are incorporated by reference.

[0050]

[0044] Example embodiments of the disclosed embodiments are described with reference to the accompanying figures.

[0051]

[0045] In the drawings, like reference numbers generally indicate identical, functionally similar, and / or structurally similar elements. The drawing in which an element first appears is indicated by the leftmost digit(s) in the corresponding reference number.

[0052]

[0046] 1. DEFINITIONS

[0053]

[0047] The term “brain tumor” refers to abnormal growth of cells within the brain or central nervous system that forms a mass. Tumor in the brain can be either benign (non-cancerous) or malignant (cancerous) and vary widely in their appearance in imaging, growth rate, impact, and treatment options.

[0054]

[0048] The term “glioma” refers to a growth of glial cells that starts in the brain or spinal cord. As glioma grows, it forms mass of cells called tumor.

[0055]

[0049] The term “computer” refers to a computing device which takes input from user and process computation and provides output. It includes memory for processing and storing information. Computer includes monitor, keyboard, mouse and controls network interface to accomplish various network communication request. The network interface may be a part of computer or connected remotely through internet. Network connection communicates from computer to computer or computer to cloud through switches, routers, firewall. Information across the network passes back and forth. Cloud computer represents a virtual computing platform may be located at different geographical location and spread across various location with multiple processers which can perform as single computer.

[0056]

[0050] The term “data” is referred to the quantitative, qualitative information, processed or unprocessed data, multidimensional array of data, characters. Data can be facts or opinions. The term data is used for singular or as well as plural form.

[0057]

[0051] The term “database” refers to data store or data repository which are organized collection of data. Data access to and frow can be provided via computer program from one or more databases located on the internet either in source which could be a hospital’s Picture Archiving and Communication System (PACS) or destination that could be a cloud server.

[0058]

[0052] The term “NoSQL” refers to a non-relational database to store and manage large sets of data in JSON-like documents. This allows to handle structured or unstructured data of varied data types for faster query performance for the web application. They are widely used in applications for scalability, flexibility, and high performance in handling varied data types.

[0059]

[0053] The term “Graphics Processing Unit (GPU)” refers to a specialized electronic circuit design to accelerate graphics processing and handle complex calculations efficiently. GPU consists of thousands of cores that enables parallel processing which has different architecture than Central Processing Unit (CPU). CPUs generally consist of smaller number of cores specialized for executing sequential instructions.

[0060]

[0054] The term “input device” generally referred to any device such as a keyboard, mouse, or touch screen which is associated with computer and configured to receive input to a processor or memory.

[0061]

[0055] The term “Random Access Memory (RAM) refers volatile memory which is quickly accessible in a computer that stores data and program instructions to be processed in CPU while performing tasks.

[0062]

[0056] The term “Module” or “Engine” refers to collections of computer instructions or series of logics or source code that is executed to perform certain tasks.

[0063]

[0057] The term “output device” generally refers to hardware that received data from computer and convert them as text, image, audio or physical output which are displayed on monitor, printer, speaker, projector, headphone, specialized mixed reality device.

[0064]

[0058] The term “tractography” provides insights of the structural connectivity pattern about how different brain regions are anatomically connected and communicate with each other through white matter pathways. Tractography is used in neurosurgery to visualize and analyse white matter tracts which are linked to eloquent brain regions and associated with motor, sensory, and cognitive functions. During the surgery damages to the fibre tracts are minimized which would be one of the major causes of neurological deficit post-surgery. Tractography also assists in pre-procedural decision-making in deep brain stimulation and helps in the diagnosis and treatment planning for various neurological disorders, that includes tumor, epilepsy, traumatic brain injury, multiple sclerosis, and neurodegenerative diseases.

[0065]

[0059] The term “parcellation” is a process in which brain image is divided into distinct, nonoverlapping regions based on anatomical, functional features. It helps to identify and organize different brain areas based on structural features, connectivity, or patterns of activity. Parcellation is useful in neurosurgical planning, medical teaching and learning to identify structural and functional areas of brain.

[0066]

[0060] The term “Resting State Network (RSN)” is a set of brain regions that show synchronized activity when a person is at rest. RSNs reveal networks such as motor, language, visual, or default mode networks which are to be protected during surgery to avoid post-operative deficits. Particularly, RSNs are useful when patients cannot perform tasks such as children, unconscious, or neurologically impaired patients.

[0067]

[0061] The term “connectome” represents comprehensive map of the brain’s neural connections, showing how different regions are structurally or functionally connected to function coherently. These regions represent critical pathways to be protected during neurosurgery such as tumor, epilepsy, traumatic brain injury, vascular anomalies.

[0068]

[0062] The term “image guided neurosurgery planning” refers to the use of medical imaging data to plan and guide neurosurgical procedures with high precision. It helps to map brain structures and functions, allowing surgeons to plan and perform operations safely and precisely.

[0063] The term “neuroimaging” refers to a medical imaging techniques and medical images refers to structural MRI, fMRI, DTI, or CT scans that shows brain's structure and function.

[0069]

[0064] The term “preoperative planning” refers to the surgical strategies devised by neurosurgeons prior to actual surgery to take decisions on positioning the subject, performing the craniotomy, and identifying anatomical reference points to access the site of intervention which could be a tumor or cerebrovascular anomalies or traumatic brain injury or epilepsy.

[0070]

[0065] The term “MRI” refers to magnetic resonance imaging technique that uses radio waves and magnets to provide exquisite details of brain internal structure, spinal cord, and other body parts, especially soft tissues with great details without any ionizing radiation.

[0071]

[0066] The term “MR angiography” refers to a non-invasive imaging test that uses specialized MR sequence to examine blood vessels anatomy.

[0067] The term “tumor segmentation” refers to segmentation of tumor tissues of brain and labelling automatically the tumor subregions.

[0072]

[0068] The term “cerebrovascular” refers to the brain and its blood vessels, which includes the arteries and veins.

[0073]

[0069] The term “cortical” refers to the cerebral cortex region of the brain that contains functional areas such as: motor, visual, sensory, auditory, prefrontal cortex, Broca’s and Wernicke’s areas. Motor areas control motor activities, while sensory areas receive sensory information through afferent fibres via thalamic nuclei.

[0074]

[0070] The term “fMRI” refers to a functional magnetic resonance imaging (fMRI) scan that shows activity in specific areas of the brain. This imaging technique uses the BOLD (Blood Oxygen Level-Dependent) signal to measure brain activity. It maps brain areas which are more active during a task or at rest.

[0075]

[0071] The term “CT” refers to cross-sectional scans obtained using Computed Tomography (CT), which uses X-rays and computer processing to produce detailed internal structures of any body part.

[0076]

[0072] The term “GUI” refers to a graphical user interface, a visual interface that allows users to interact with electronic devices using graphical elements like icons, buttons and menus.

[0077]

[0073] The term “PACS” refers to picture archiving and communications system, a standard platform to manage medical images.

[0078]

[0074] The term “HIPPA” refers to Health Insurance Portability and Accountability Act, an US based law that protects the privacy and security of health information.

[0079]

[0075] The term “GDPR” refers to the General Data Protection Regulation (GDPR) is a European Union (EU) law that protects individuals' personal data and fundamental rights.

[0080]

[0076] The term “Talairach Atlas” or “Harvard-Oxford Atlas” or “Schaefer Atlas or Smith Atlas”, is a 3-dimensional coordinate system (known as an 'atlas') of the human brain, which is used to map structural and functional regions of the brain.

[0081]

[0077] The term “NEURO VIZ” refers to an advanced platform for neurosurgical plarming of the present disclosure.

[0082]

[0078] 2. OVERVIEW

[0083]

[0079] The disclosed embodiment illustrates the details of the image guided neurosurgery planning platform designed for preoperative, intraoperative planning and postoperative evaluation. NEURO VIZ is a neurosurgery planning platform providing intricate details about the structural and functional connectivity of brain with customized tools to access insights of the brain by advanced visualization in 2D and 3D along all planes.

[0080] The features of the neuroimaging analytical system include pre-processing of MRI from different MR sequences, delineation of the brain tumor into distinct subregions, segmentation of tissue into GM, WM and CSF, parcellation of brain into cortical subcortical regions, and brain lobes, cerebrovascular segmentation, generation of white matter fibre tracts, construction of resting state networks and CT image processing, 3D rendering of the region of interests, advanced visualization with multiplanar 3D slicer tool, and smart brush and 3D edit option to create or edit ROI. Further, advanced visualization offers visualization of all tumor proximal fibre tracts within 10 mm of the tumor and overlap percentage of white matter fibre tract with tumor if any tract is overlapping with tumor core. The platform provides advanced, interactive 2D and 3D visualization capabilities that enable accurate and efficient neurosurgical planning.

[0081] NEURO VIZ application is an advanced platform for neurosurgical planning. The application has a user side where users can upload medical images such as MRI and CT scans for processing and generating a personalized brain map to visualize structural and functional connections before and during the neurosurgery (FIG. 1). The platform assists neurosurgeons in achieving high precision, thereby helping to minimize post-surgery neurological deficits.

[0084]

[0082] Customized tools and advanced visualization options are available on the user side to access accurate information about the subject’s brain internal structures. On the user side, a dashboard is provided that contains records of the processed data of each subject (102). There is a panel for adding new patient data that requires processing (104), which can be monitored in the process monitoring dashboard (106). There is also a settings section for each user, which contains information such as the license key, validity, software version, and credits used by the user (108).

[0085]

[0083] The admin side controls all user access restrictions. The admin side possesses a list of free trial users (110), paid users (112), and inactive users (114). The admin can add auser (116), activate a user (118), address user complaints (120), and manage user feedback (122). Furthermore, there is a dashboard to monitor and address application errors (124), process monitoring (126), app usage (128), and resource monitoring (130).

[0086]

[0084] 3. DELINEATION OF THE BRAIN TUMOR INTO DISTINCT SUBREGIONS

[0087]

[0085] Delineation of brain tumor into distinct subregions requires several processing steps before desired results are obtained as shown in FIG. 2. Medical images are either accessed from PACS or uploaded directly to the application from a local system, DVD, or USB drive. Images are accessed in the platform as Digital Imaging and Communications in Medicine (DICOM) format. As a first step, DICOM images are deidentified as per HIPPA and GDPR compliance as soon as images are uploaded into the application even before any processing is performed (202). The structural MRI images such as Tlw, T2 -weighted (T2w), post-contrast- T1 weighted (Tic) and fluid attenuated inversion recovery (FLAIR) are converted to NI1TI from DICOM (204). Subsequently, images undergo field of view (FOV) correction (206) followed by removal of inner noises (208). N4-bias correction is another important preprocessing step for MRI images which removes magnetic field inhomogeneity (210) followed by noise removal from background (212). Intensity normalization of the images is then performed (214). A brain mask is created using function from the available software library to be used for volume calculation (216). The images are then registered to the same anatomical space (218) and then registered to the standard space (220). Subsequently, another brain mask is created in standard space (222) to be used for skull stripping (224). The brain volume used in visualization is calculated just after normalization in native space (226). Subsequently, tumor segmentation was performed using our pretrained model to obtained whole tumor, tumor core and enhancing tumor (228). For tumor segmentation, a large dataset comprising approximately 2,900 patients MRI scans (T1 weighted, T2 -weighted, FLAIR and T1 weighted post contract) is pre-processed and used to train a 3D U-Net model employing a 5 -fold cross-validation strategy. During training, a hybrid loss function combining Dice loss and weighted binary cross-entropy is employed to effectively address class imbalance and improve segmentation accuracy.

[0088] Where, LDiceis Dice loss, LWBCEis weighted cross entropy loss and Vindicates weighting factor which typically varies between 0 and 1.

[0089]

[0086] The model is trained for 75 epochs and final loss achieved is 0.1. For tumor segmentation, the performance metrics obtained for the whole tumor region include a Dice score of 95%, sensitivity of 94%, and specificity of 100%. For the tumor core, the Dice score, sensitivity, and specificity are 93%, 92%, and 100%, respectively. For the enhancing tumor region, the corresponding values are 92%, 91%, and 100%. The tumor subregions corresponding to edema, enhancing tumor, and necrotic core are subsequently derived from these predicted segmentation masks. The tumor masks are resampled back to native space (230) subsequently volumes are determined (232) and saved (234). The tumor is delineated into distinct subregions such as peritumoral edema (ED) in green, enhancing tumor (ET) in yellow and necrotic core (NR) in red as shown in FIG.3. The corresponding volumes are estimated for these regions as shown in the Table 1.

[0087] Table 1. Estimated volume of edema, enhancing tumor and necrotic core in native space for the same patient whose subregions are shown in FIG. 3.

[0090]

[0088] FIG. 3 illustrates the location of the tumor subregions on the Tlw image in axial (302), sagittal (304) and coronal planes (306). The tumor is also shown in 3D with head positioned in sagittal direction showing all three regions (308), tumor core (310) and necrotic core (312).

[0091]

[0089] 4. TISSUE SEGMENTATION INTO GRAY MATTER, WHITE MATTER AND CORTICOSPINAL FLUID

[0092]

[0090] In another aspect of the disclosure, the system is configured to perform tissue classification into white matter (WM), grey matter (GM), corticospinal fluid (CSF) on skull stripped brain that helps in quantifying volumetric information of these regions. The steps involved in tissue classification are shown in the flowchart in FIG. 4. Tissue segmentation is performed on the Tlw image, where the DICOM image is anonymized (402) and converted to NlfTI format (404), followed by FOV correction (406). Subsequently, inner noise correction (408), N4 bias correction (410), and background noise removal (412) are performed on the Tlw image. Thereafter, normalization is applied (414), and a brain mask is generated using a pretrained model (416). The generated brain mask is used for skull stripping (418) and brain volume calculation (420).

[0093]

[0091] Following this, automatic segmentation of GM, WM, and CSF is carried out using a pretrained model (422). The delineation of GM, WM and CSF is taken from the trained model which is trained using 2100 patients Tl-weighted image and GM, WM and CSF annotation. The model was trained where dice and cross entropy loss is minimized to achieve good accuracy. The segmented masks are then resampled to the native space (424) and combined into a single mask (426).

[0094]

[0092] For tumor subjects, GM, WM, and CSF estimation is performed slightly differently. The tumor core is subtracted (428) from the combined mask before volumetric estimation of GM, WM, and CSF (430). Subsequently, all files are saved (432).

[0095]

[0093] The combined mask with the subtracted tumor region overlaid on the Tlw image is shown in FIG. 5 in axial (502), sagittal (504), and coronal (506) planes, as well as in 3D view (508). A similar representation, including the tumor core mask that comprises the enhancing tumor and necrotic core regions, is shown in axial (510), sagittal (512), coronal (514), and 3D (516) views. Separately, white matter along with the tumor core is shown in axial (518), sagittal (520), coronal (522), and 3D (524) views; gray matter is shown in axial (526), sagittal (528), coronal (530), and 3D (532) views; and CSF is shown in axial (534), sagittal (536), coronal (538), and 3D (540) views. Volumetry assessment of brain, WM, GM and CSF of a patient is given in Table 2.

[0096]

[0094] Table 2. Calculated intracranial, WM, GM and CSF volumes of a patient.

[0097]

[0095] 5. BRAIN PARCELLATION INTO DISTINCT REGIONS

[0098]

[0096] The spatial organization of brain elements pose a key challenge in translating this information into clinical care. Parcellation of brain into structural regions have immense implication. It divides the brain into multiple, spatially discontinuous regions. The steps involved in parcellation are shown in the flowchart in FIG. 6. For brain parcellation into distinct structural regions, the Tlw image is used. The image is first anonymized (602), similar to other MRI images, and then converted from DICOM to NIfTI format (604). This is followed by field-of-view (FOV) correction (606), inner brain noise removal (608), N4 bias correction (610), and background noise removal (612). Subsequently, normalization is performed (614), and a brain mask is generated (616) for skull stripping (618). The brain mask is then used to estimate brain volume (620). These pre-processing steps are common across all three types of parcellation.

[0099]

[0097] (i) ATLAS BASED PARCELLATION (ABP)

[0100]

[0098] Brain atlas (622) of choice can be used to obtain desired parcellation (automated anatomical labelling (AAL), Harvard-Oxford MNI structural atlas, Desikan-Killiany atlas). The selected brain atlas and corresponding Tlw image in Montreal Neurological Institute 152 (MNI152) is non-linearly registered to the processed Tlw image (624) and that matrix is used to generate multiple spatial regions in native space (628). Distinct brain regions can be extracted from this Tlw image (630). These files are saved for later use in visualization (632). ABP segments brain into 48 cortical regions and 21 subcortical regions.

[0101]

[0099] FIG. 7 shows some of the subcortical parcellation in the axial (702), sagittal (704), coronal (706), and 3D (708) views, as well as cortical parcellation in the axial (710), sagittal (712), coronal (714), and 3D (716) views. Volumetry information is given for each parcellated region.

[0100] (ii) BRAIN LOBE PARCELLATION

[0102]

[0101] In another aspect, brain lobe parcellation is carried out in the similar way as described in cortical and subcortical parcellation by taking brain lobe atlas (634) registering it to Tlw image (636) and obtaining the lobe segments in native space (638) and likewise each lobe is extracted (640) and saved (642).

[0103]

[0102] The parcellated brain lobes include the left and right frontal, temporal, parietal, occipital, and insular lobes, atotal of 10 regions as shown in FIG. 7 in axial (718), sagittal (720), coronal (722) and 3D (724). In FIG. 7, the left frontal and parietal lobes are not displayed to provide a clearer view of the internal structures of the lobes, and the 3D structure is without outer face cover. For each parcellated region, volume information is provided.

[0104]

[0103] (iii) MODEL BASED PARCELLATION (MBP)

[0105]

[0104] In another aspect, model based parcellation is performed to segment cortical and subcortical regions based on a pretrained model (644) where processed Tlw image is used for parcellation. The pretrained model is taken from an open-source library. The segmented regions are resampled to native space (646) and region of interest can be extracted (648) and corresponding files are saved (650). Model based parcellation segments 32 regions of the brain such as brainstem, left- and right- putamen, pallidum, caudate, thalamus, hippocampus, amygdala, accumbens, lateral ventricles, inferior lateral ventricles, cerebellum etc. FIG.7 shows some of these regions for one hemisphere in axial (726), sagittal (728), coronal (730) and 3D views (732). MBP deduces volumes of each region alongside of the region name in the application.

[0106]

[0105] 6. SEGMENTATION OF CEREBROVASCULAR SYSTEM

[0107]

[0106] The detailed processing steps are given in FIG. 8 as a flowchart. MRA DICOM file is anonymized (802), converted to NlfTI (804), subjected to FOV correction (806), inner noise removal (808), intensity normalization (810) and creation of brain mask (812). Subsequently, MRA image is registered to Tlw image (814) and following which skull stripping is performed (816) which provides brain volume information (818). Voxel corresponding to vessels are extracted based on vessel processing algorithm (820) wherein image is analysed at multiple scales by convolving it with Gaussian filters of varying standard deviations. Following which second-order partial derivatives are taken at each voxel and eigenvalues are calculated from this matrix which in turns provides local curvature at that point and that is how it traced out the complete cerebrovascular systems. Subsequently, threshold is used to remove noise (822) and disconnected vessels (824) and leaving only connected vessels information. This file is saved (826) for 3D vessel rendering and visualization. Similar method of extraction of vessels can also be carried out from CTA image.

[0108]

[0107] FIG. 9 shows the delineation of blood vessels in axial (902), sagittal (904), coronal (906) and 3D rendered (908) views. There is also another option to visualize maximum intensity projection (MIP) in axial (910), sagittal (912) and coronal (914) planes. Rendered blood vessels can be viewed with brain volume turn on (908) or off (916). NEUROVIZ also performs volume-based 3D MIP vessel projection to delineate blood vessels from sTIW 3D FFE FS MRI sequences in addition to MRA and CTA images.

[0109]

[0108] 7. CONSTRUCTION WHITE MATTER FIBRE TRACT

[0110]

[0109] Tractography is extremely important in neurosurgery, especially for preoperative planning and intraoperative plarming. It provides a virtual map of the brain’s white matter pathways, allowing neurosurgeons to plan safer and more precise neurosurgical intervention. Tractography reconstructs white matter fibre tracts and construct the communication highways of brain that connect different functional regions. These include motor pathways formed by the corticospinal tract, language pathways constituted by the arcuate fasciculus, visual pathways represented by the optic radiations, and memory and emotion pathways mediated by the cingulum and fornix. By visualizing these tracts surgeons understand which regions must be preserved to maintain essential functions such as movement, speech, vision or memory. Deducing such fibre tracts is important for preoperative and intraoperative neurosurgical panning to identify the relationship between tumor and adjacent fibre tracts. It defines safest surgical corridor to remove tumor without disrupting critical tracts. It assists in carrying out maximal safe tumor resection. It has applications not only in tumor surgery but also in traumatic brain injury, epilepsy surgery, and other neurosurgical interventions.

[0111] [HO] In another aspect of this disclosure, diffusion MRI is used for generating white matter fibre tracts which is crucial for preoperative and intraoperative plarming. This technique assists in mapping the functionally critical pathways in and around the surgical site and thus reduces permanent neurological deficit post neurosurgery. Further, tractography is indispensable in functional neurosurgery such as deep brain stimulation for optimizing the stimulation sites for maximizing therapeutic benefit. Most often the regions where crucial fibre tracts are located used as a direct site for deep brain stimulation.

[0112] [Hl] Diffusion Tensor Imaging (DTI) is used for the construction of white matter fibre tracts by acquiring diffusion-weighted images in multiple gradient directions with one or more b- values, that characterize the orientation of white matter pathways.

[0112] In another aspect of this disclosure, steps used for the construction of fibre tracts from DTI image are given in FIG. 10. The first step involves DICOM anonymization (1002), followed by the conversion of DICOM files into NIfTI file format (1004), and the creation of a brain mask from the DTI image (1006). Subsequently, a gradient check is conducted to ensure that the diffusion gradient directions are correctly aligned with the image data (1008). This is followed by denoising (1010), Gibbs unringing correction (1012), distortion correction (1014), and N4 bias correction (1016), after which a mask is generated from the corrected image (1018). If posterior to anterior (PA) and anterior to posterior (AP) image pairs are acquired, then phase-encoding direction correction is performed by creating an off-resonance field map from the difference of the PA and AP images and applying a wrap field to the entire DTI series to correct the susceptibility-induced distortions. The number of non-zero b-values associated with the DICOM image, which constitute diffusion shells, are then detected (1020). The image is subsequently up sampled by a factor of 1.25 to achieve better anatomical representation (1022). Thereafter, diffusion metrics such as fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) are computed (1024). Next, a response function is generated to ensure accurate fibre orientation estimation, leading to improved streamline reconstruction (1026). For the CSD method, Fiber Orientation Distribution (FOD) estimation is performed (1028). However, for DTI tensor-based modelling, this step is not required since there is only one principal diffusion direction per voxel, whereas CSD accounts for multiple fibre orientations within a voxel. Subsequently, FOD normalization is performed (1030). Both FOD estimation and normalization steps are applicable only to the CSD probabilistic method (1032).

[0113]

[0115] Subsequently, the Tlw image is registered to the B0 image (1034), followed by the generation of the fractional anisotropy (FA) map (1036). The outer boundary of the FA map is refined by applying the brain mask (1038). Tissue segmentation is then performed on the Tlw image, and seed region is defined to ensure biologically valid fibre propagation (1040). Presence of free water in the edema region supress the anisotropy diffusion of water and results in premature termination of white matter fibre. Therefore, edema correction is applied to suppress the isotropic signal by modelled it as CSF-like tissue for the estimation of fibre orientation distributions (FODs). Streamlines once generated for whole brain, it is registered to the standard space (1042). The whole-brain streamlines are further aligned to the same standard space (1044). Fragmented streamlines that do not satisfy the continuity criteria are filtered out (1046). Specific fibre tracts are subsequently extracted through clustering, resulting in the refinement of 87 distinct fibre tracts (1048). Finally, all streamlines are transformed back to the native space (1050), and the resulting files are saved for visualization (1052).

[0114]

[0113] In another aspect, fibre tracts obtained can be visualized in our application where a specific tract can be switched on or off. The fibre bundles are grouped in the platform as association, brainstem, cerebellum, commissural, cranial nerve and projection fibre bundles. Arcuate fasciculus left and right fibre tracts generate using DTI tensor model which regulate speech from association group is overlaid on Tlw image in axial (1102), sagittal (1104) and coronal (1106) views and the same tract obtained from CSD probabilistic method are overlaid on axial (1108), sagittal (1110) and coronal (1112) plane as shown in FIG.ll. Compared to the conventional DTI tensor model, CSD probabilistic tractography generates well-resolved streamlines with higher anatomical accuracy by effectively capturing crossing, branching, and fanning fibre configurations. While DTI assumes a single principal diffusion direction per voxel, CSD method models multiple fibre orientations within a voxel, enabling a more precise representation of the brain’s complex white matter architecture.

[0115]

[0114] In another aspect, fibre tracts are viewed in 3D along with 3D of tumor subregions such as peritumoral edema, enhancing region and necrotic core regions in the application. 3D visualization of tumor, fibre tracts together inside the brain provides spatial organization of tumor and its proximity to the crucial fibre tracts as a preoperative guidance to the neurosurgeon for planning surgery. White matter fibre tracts within 5 mm from tumor core which include enhancing tumor and necrotic core are shown in axial (1114), sagittal (1116) and coronal (1118) views in FIG. 11. In case any fibre tract is overlapping with the tumor core, it shows percentage of overlap between the tumor and the respective fibre tract. There is also an option for selecting multiple regions of interests (ROIs) such as cortical, subcortical parcellated regions, organ at risk or blood vessels etc.

[0116]

[0115] 8. MAPPING OF RESTING STATE NETWORKS

[0117]

[0116] The construction of resting-state fMRI (rsfMRI) and task-based fMRI (tbfMRI) Blood Oxygen Level-Dependent (BOLD) signals provide complementary insights into the brain’s functional architecture. It reflects hemodynamic changes in the brain cortical region which is primarily related to neuronal activity associated with visual, auditory, motor function etc. While rsfMRI captures spontaneous neural activity and intrinsic connectivity patterns, tbfMRI reveals task-evoked activations, allowing precise mapping of eloquent regions. Integrating these signals with fibre tracts enables the generation of comprehensive functional connectomes, which depict the intricate network of interactions among cortical and subcortical regions. Such connectomes are invaluable in preoperative neurosurgical plarming, as they help to identify critical functional pathways, minimize disruption to essential networks, and enhance surgical precision. By leveraging both spontaneous and task-driven BOLD dynamics, clinicians can achieve a more accurate, patient-specific representation of brain connectivity, preserve neurological function and ultimately improve surgical outcomes. Resting state network maps essential functional networks with shorter acquisition and which is consistent across all subjects compared to task-based activation. Resting state networks are indispensable for patients who are young, elderly or cognitively impaired.

[0118]

[0117] Derivation of resting state network is challenging due to low-signal to noise ratio, complex temporal dynamics, overlapping networks, physiological noises. However, resting state network can be derived following the steps shown in FIG. 12. The DICOM image is first anonymized (1202) and converted to NIfTI format (1204), followed by orientation correction (1206). The initial five volumes are removed (1208) to ensure signal stability, after which slice timing correction is performed to align the temporal acquisition of slices, allowing accurate representation of BOLD signal fluctuations across the whole brain (1210). Spikes are then removed from the BOLD signal (1212), and motion correction is applied to minimize head movement artifacts (1214). A brain mask is subsequently created (1216), and tissue segmentation is carried out (1218). The resulting image is registered to standard space (1220), and nuisance regression is performed to remove non-neuronal fluctuations (1222). Low bandpass filtering is then applied within the 0.01-0.1 Hz range to retain slow oscillations associated with resting-state networks (1224).

[0119]

[0118] Spatial smoothing is performed next to enhance signal quality and improve the signal-to- noise ratio (1226). Independent Component Analysis - Automatic Removal Of Motion Artifacts (ICA -AROMA) is then conducted to decompose the data into spatially independent components (1228) which automatically identify and removes motion-related noise components. Among these, some components represent resting state networks, while others correspond to motion, physiological noise, or scanner drift; the latter are regressed out, retaining only meaningful neural signals. These retained components are further filtered through template matching with standard resting state network maps (1230) with a spatial correlation coefficient threshold greater than 0.3-0.5 and DICE similarity threshold of greater than 0.3. The major resting state network deduced by our application are visual network, occipital network, default mode network, sensorimotor network, auditory network, left executive control network, right executive control network and dorsal attention network. Finally, the data are resampled back to native space (1232), and the resulting files are saved for visualization (1234).

[0119] In another aspect of this disclosure, distinct resting state networks derived from rsfMRI are shown in FIG 13. Visual network is shown in axial (1302), sagittal (1304), coronal (1306) and 3D views (1308). Similarly, occipital network is shown in axial (1310), sagittal (1312), coronal (1314) and 3D view (1316) in FIG. 13. Axial (1318), sagittal (1320) and coronal (1322), 3D views (1324) of auditory network is also shown in FIG 13.

[0120]

[0120] 9. IMPLICATION OF CT IMAGE PROCESSING IN NEUROSURGERY

[0121]

[0121] The most significant contribution of CT image processing is the ability to create detailed three-dimensional models of a patient's brain and skull. CT scans provide details of the bony structure while MRI offers complementary information of the soft tissues. Our application fuses these two types of scans into single model which offers combine strengths of both CT and MRI. Fused image often acts like a computer assisted navigation system during neurosurgery. This technology shortens the operation times and reduces unnecessary neurovascular injuries. Operating with intraoperative CT scan facility assists surgeon to adjust brain shift. Inclusion of CT along with MRI is indispensable neurosurgical procedure which offers greater accuracy and safety.

[0122]

[0122] In another aspect of this disclosure, NEUROVIZ fuses all processed results from preoperative MRI images with intraoperative CT scans to perform global correction of brain shift due to CSF leakage occurring during neurosurgery. In another aspect of this disclosure, CT image processing is carried out as illustrated in FIG. 14 flowchart, beginning with anonymization of the DICOM image (1402), followed by conversion of the DICOM file to NIfTI format (1404). The NIfTI image is then converted to aNumPy array (1406) and subjected to preprocessing steps such as removal of extraneous elements including the patient bed, table, and surrounding air (1408).

[0123]

[0123] The image is subsequently converted to the Hounsfield Unit (HU) range to visualize bone tissue for surgical planning (1412). However, the HU window can be adjusted to display bone (1510), internal brain structures, such as GM, WM (1512), and CSF, thereby facilitating comparison and alignment with MRI scans. The processed image is then resampled to native space (1414), and the final file is saved for visualization (1416).

[0124]

[0124] 10. GRAPHICAL USER INTERFACE (GUI) FOR ADVANCED VISUALIZATION

[0125]

[0125] A GUI platform application was designed to enhance precision, subject safety, efficient neurosurgical planning for neurosurgeons, medical staff teaching and to explain the condition of brain to the subject’s family. The platform was also designed as HIPPA and GDPR compliant. It provides a simple interactive interface with plug and plays buttons so that even a new user with limited technical knowledge can efficiently use it. However, interpreting clinical conditions and performing surgical planning require the expertise of a qualified clinical specialist. It offers high-quality advanced 2D and 3D visualization with 3D rendering option to view tumor, GM, WM, CSF, fibre tracts, lobe, cortical and subcortical parcellated regions, brain vascular system and functionally active regions, bony part of brain in multiplanar 3D slicing tool to view personalized brain intricate details in any plane of interest.

[0126]

[0126] The present disclosure provides a customized advanced graphical user interface (GUI) for patient-specific neuroimaging visualization, as illustrated in FIG. 15A and 15B. The left panel (1502) displays the available medical images, which can be selectively rendered in the main visualization canvas. The main canvas is divided into four sections for viewing axial (1504), sagittal (1508), coronal (1506), and 3D perspectives (1510). The top control panel (1512) includes a comprehensive set of visualization tools such as reset, full-screen mode, zoom, pan, rotation, crosshair activation, inversion, brightness adjustment, colormap selection, lighting control, measurement utilities, cine playback, and 3D conversion tools. There is also a tool to convert overlaid mask as outline (1514).

[0127]

[0127] The right panel (1516) provides task-specific visualization controls including options for displaying anatomical structures (1518), tumor subregion volumes (1522), and opacity adjustments (1524). The narrow rightmost panel (1520) offers additional visualization features tailored to specific analytical tasks, such as detailed brain structures, blood vessels, white matter fibre tracts, parcellation maps, resting-state networks, 2D and 3D editing tools, multiplanar 3D slicer functionality, and automated report generation.

[0128]

[0128] FIG. 15B illustrates white matter (WM) overlaid on a T1 -weighted (T1W) image in axial (1526), coronal (1528), sagittal (1530), and 3D views. The skull (1534), derived from CT imaging, and the 3D WM structures (1532) derived from MRI, are spatially fused within a unified coordinate space. Estimated WM, gray matter (GM), and cerebrospinal fluid (CSF) volumes (1538) are also displayed. The system additionally provides Hounsfield Unit (HU) adjustment controls (1536) to enable visualization of specific CT tissue and bone intensity ranges.

[0129]

[0129] NEUROVIZ integrates multiple imaging modalities within a common spatial framework to deliver comprehensive insights into intricate brain structural and functional networks. The application is not limited to MRI and CT, but also supports visualization of digital subtraction angiography (DSA), MR spectroscopy, perfusion MRI, CT angiography, and other imaging datasets.

[0130] Processed results generated in NEURO VIZ are fully compatible with neuronavigation systems used for real-time monitoring during neurosurgery. The present disclosure also includes smart brush 2D edit tool (FIG. 16) to edit tumor subregions such as edema, enhancing tumor, necrotic core (1604), GM, WM, CSF masks (1606), or any region of interest either generated manually or predicted by models (1608). It also has 3D edit tool to remove wrongly predicted tumor mask which also reflects on 2D (1610).

[0130]

[0131] The said disclosure allows user to interactively customize views for better understanding of anatomical and functional details. GUI integrates multiple types of MR sequences such as structural MRI, diffusion MRI, functional MRI, MR angiography, CT image into a single coherent platform for critical surgical planning.

[0131]

[0132] The application allows the user to tailor the outputs as per specific requirements. It facilitates instant report generation with quantitative volumetric information and information required for surgical planning.

[0132]

[0133] Users can generate report through a speech to text converter engine which enhances efficiency by eliminating manual typing. A customized complaint and feedback form with integrated speech to text converter application is also be a part of the application to keep adding needs and requirements of the user time to time. The complain and feedback are incorporated and implemented with every new release of the software in this platform. Provision of these automated workflows reduce error and provide consistent and accurate results.

[0133]

[0134] The application is provided to the healthcare facilities with either on premise or on secure cloud server. There is API and token-based authentication to access the application. At every stage of transit high data security and encryption is employed to make the complete dataflow highly secure. The data is also encrypted at the storage. NEURO VIZ integration with PACS is seamless and the results obtained by NEUROVIZ are neuronavigation system compatible making it useful for intraoperative guidance along with preoperative planning.

[0134]

[0135] The disclosed platform provides an end-to-end solution, encompassing data acquisition, preprocessing, advanced analysis, visualization, and reporting, while ensuring high data security and encryption to protect sensitive information at every stage of the workflow.

[0135]

[0136] 11. IMPLEMENTATION OF THE NEUROSURGERY PLATFORM WITH ADVANCED NEUROIMAGING ANALYTICS

[0136]

[0137] Pre-processing of structural MRI, diffusion MRI, magnetic resonance angiography and functional MRI data, de-lineating of brain tumor into distinct subregions (FIG. 3), segmentation of brain into GM, WM and CSF (FIG. 5), parcellation of brain into distinct lobe, cortical and subcortical regions (FIG. 7), rendering 3D images of tumor, white matter fibre tracts viewed in 3D along with 3D of tumor subregions such as peritumoral edema, enhancing tumor and necrotic core region (FIG. 11), segmentation of cerebrovascular systems from MRA and 2D and 3D visualization (FIG. 9), generating and analysing resting state networks (FIG. 13) and task-based fMRI images of the cortical brain including regions responsible for motor, speech and visual, CT processing tool, smart brush edit option for 2D and 3D edit, advanced visualization with multiplanar 3D slicing tool, designing a GUI application for integrated visualization of different MRI sequences processed results into a single coherent platform for critical surgical planning.

[0137]

[0138] Further, it integrates MRI- based neuroimaging analytical reporting system through a speech to text conversion engine to enhance accuracy, efficiency and precision which results in patient safety, faster recovery and positive outcomes.

[0138]

[0139] 12. USES, APPLICATIONS, BENEFITS AND BEST MODE TO PRACTICE

[0139]

[0140] The neuroimaging analytics platform is designed for preoperative neurosurgery planning and postoperative assessment. The advanced features offered by the product would help neurosurgeons to enhance surgical precision, allowing them to achieve safer resection. The product is designed by keeping neurosurgeons' requirements in mind, ensuring it meets their specific needs. This holistic approach enables neurosurgeons to improve surgical outcomes, reduce post-surgical neurological deficit, and enhance the subject’s safety.

[0140]

[0141] Different combinations of brain regions can be superposed and viewed. The regions that can be visualized include tumor, GM, WM, CSF, cortical, subcortical regions, blood vessels, white matter fibre tracts, proximal tract within 10 mm of tumor, synchronously active regions of the brain in resting state, bony structure of head in any plane in multiplanar 3D slicing tool and thus can guide the neurosurgeon to precisely resect region of interest while protecting the eloquent brain regions.

[0141]

[0142] The proposed platform also helps in visualization of functional areas of the brain via MRI imaging specifically focusing on the regions that receive maximum blood flow. Users can generate a report through a speech to text converter engine which enhances efficiency by eliminating manual typing. Also, a customized feedback form with integrated speech to text converter application to full fill the requirements time to time. The application can take MRI and / or CT images from DVD, USB drive or directly from PACS for processing. Processed files obtained from this application can be feed into the neuronavigational system for image guided neurosurgery planning.

[0142]

[0143] The designed modules in the present disclosure allow visualization of brain regions in different combinations. For example, fibre tracks or tumor or parcellated brain or blood vessels or resting state networks simultaneously or in any combinations. The disclosed embodiment can be commercialized to the hospitals having neurosurgery facilities to provide assistant to the neurosurgeons in preoperative planning, intraoperative neurosurgery procedure, and postoperative neurosurgery assessment.

[0143]

[0144] It should be further appreciated that the above noted features can be implemented in various embodiments as a desired combination of one or more of hardware, execution modules, and firmware. The description is continued with respect to one embodiment in which various features are operative when execution modules are executed.

[0144]

[0145] 13. HARDWARE

[0145]

[0146] FIG. 17 is a block diagram illustrating the details of digital processing system (1700) in which various aspects of the present disclosure are operative by execution of appropriate execution modules. Digital processing system 1700 may correspond to a system executing the neurosurgery planning platform (NEUROVIZ).

[0146]

[0147] Digital processing system (1700) may contain one or more processors (such as a central processing unit (CPU) 1701), random access memory (RAM) (1702), secondary memory (1703), graphics controller (1706), display unit (1707), network interface (1708), and input interface (1709). All the components except display unit (1707) may communicate with each other over communication path (1705) which may contain several buses as is well known in the relevant arts. The components of FIG. 17 are described below in further detail.

[0147]

[0148] CPU (1701) may execute instructions stored in RAM (1702) to provide several features of the present disclosure. CPU (1701) may contain multiple processing units, with each processing unit potentially being designed for a specific task. Alternatively, CPU (1701) may contain only a single general-purpose processing unit. RAM (1702) may receive instructions from secondary memory (1703) using communication path (1705).

[0148]

[0149] Graphics controller (1706) generates display signals (e.g., in RGB format) to display unit (1707) based on data / instructions received from CPU (1701). Display unit (1707) contains a display screen (e.g., EED lights) to display the images defined by the display signals (e.g., the user interfaces of FIG.s 3, 5, 7, 9, 11, 13 and 15). Input interface (1709) may correspond to a keyboard and a pointing device (e.g., touch-pad, mouse), which enable the various inputs to be provided (e.g., inputs to interface with user interfaces of FIG.s 3, 5, 7, 9, 11, 13 and 15).

[0149]

[0150] Network interface (1708) provides connectivity to a network (e.g., using Internet Protocol), and may be used to communicate with other connected systems. Network interface (1708) may provide such connectivity over a wire (in the case of TCP / IP based communication) or wirelessly (in the case of Wi-Fi, Bluetooth based communication).

[0151] Secondary memory (1703) may contain hard drive (1703a), flash memory (1703b), and removable storage drive (1703c). Secondary memory (1703) may store the data and software instructions (e.g., for implementing the blocks of FIG.s 1 and 16, for implementing the steps of FIG.s 2, 4, 6, 8, 10, 12 and 14 and other aspects of the present disclosure), which enable digital processing system (1700) to provide several features in accordance with the present disclosure.

[0150]

[0152] Some or all of the data and instructions may be provided on removable storage unit (1704), and the data and instructions may be read and provided by removable storage drive (1703c) to CPU (1701). Floppy drive, magnetic tape drive, CD-ROM drive, DVD Drive, Flash memory, and removable memory chip (PCMCIA Card, EPROM) are examples of such removable storage drive (1703c). Removable storage unit (1704) may be implemented using storage format compatible with removable storage drive (1703c) such that removable storage drive (1703c) can read the data and instructions. Thus, removable storage unit (1704) includes a computer readable storage medium having stored therein computer software (in the form of execution modules) and / or data. However, the computer (or machine, in general) readable storage medium can be in other forms (e.g., non-removable, random access, etc.). These "computer program products" are means for providing execution modules to digital processing system (1700). CPU (1701) may retrieve the software instructions (forming the execution modules) and execute the instructions to provide various features of the present disclosure described above.

[0151]

[0153] Merely for illustration, only representative number / type of graph, chart, block, and subblock diagrams were shown. Many environments often contain many more block and sub-block diagrams or systems and sub-systems, both in number and type, depending on the purpose for which the environment is designed. While specific embodiments of the disclosure have been shown and described in detail to illustrate the inventive principles, it will be understood that the disclosed embodiment may be embodied otherwise without departing from such principles.

[0152]

[0154] Reference throughout this specification to “one embodiment”, “an embodiment”, or similar language means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present disclosure. Thus, appearances of the phrases “in one embodiment”, “in an embodiment” and similar language throughout this specification may, but do not necessarily, refer to the same embodiment.

[0153]

[0155] It should be understood that the figures and / or screen shots illustrated in the attachments highlighting the functionality and advantages of the disclosed embodiment are presented for example purposes only. The disclosed embodiment is sufficiently flexible and configurable, such that it may be utilized in ways other than that shown in the accompanying figures.

[0154]

[0156] It should be understood that the examples and embodiments described herein are for illustrative purposes only and that various modifications or changes in light thereof will be suggested to persons skilled in the art and are to be included within the spirit and purview of this application and scope of the appended claims. All publications, patents, and patent applications cited herein are hereby incorporated by reference in their entirety for all purposes.

Claims

What is claimed is:

1. A non-transitory machine -readable medium storing one or more sequences of instructions forming a neurosurgery planning platform for preoperative, intraoperative planning and postoperative evaluation, wherein execution of said one or more instructions by one or more processors contained in a digital processing system causes the neurosurgery planning platform to perform the actions of: receiving a plurality of images of a brain tumor in a brain of a subject, wherein the plurality of images includes magnetic resonance imaging (MRI) images and computed tomography (CT) scan images; pre-processing the plurality of images corresponding to different types of MR sequences to obtain a pre-processed set of images; applying a first pretrained model to the pre-processed set of images to delineate a brain tumor into distinct subregions comprising peritumoral edema, enhancing tumor and necrotic core; and displaying, on a display unit, a first set of images in axial, sagittal, and coronal planes, and in 3D (three-dimensional) view, wherein the first set of images includes images indicating the distinct subregions.

2. The non-transitory machine-readable medium of claim 1, wherein each MRI image of the plurality of MRI images is according to a Digital Imaging and Communications in Medicine (DICOM) format, wherein the pre-processing comprises: converting the plurality of images to a standard format; performing a field of view (FOV) correction to the plurality of images; removing magnetic field inhomogeneity followed by noise from background in the plurality of images; performing intensity normalization of the plurality of images; and creating a brain mask to be used for skull stripping and brain (intracranial) volume calculation.

3. The non-transitory machine -readable medium of claim 2, further comprising: classifying, using a second pretrained model, the brain tissue into a set of regions comprising white matter (WM), grey matter (GM), corticospinal fluid (CSF) regions in the image of the skull stripped brain;PLDIN2024-90258P-PCTestimating based on the set of brain masks, volumetric information of the brain and each of the set of regions, wherein the estimation is performed by first subtracting a tumor core region from GM, WM, CSF masks; and displaying, on the display unit, a second set of images in axial, sagittal, and coronal planes, and in 3D view, wherein the second set of images includes images overlaid with tissue masks (GM, WM and CSF) with the tumor core region subtracted, images overlaid with another tissue masks including the tumor core region, images showing each of the set of tissue regions along with the tumor core region.

4. The non-transitory machine -readable medium of claim 3, further comprising: performing a parcellation of the brain into a set of structural regions comprising cortical and subcortical regions; and displaying, on the display unit, a third set of images in axial, sagittal, and coronal planes, and in 3D view, wherein the third set of images includes images showing cortical parcellation, image showing subcortical parcellation, images showing the left and right frontal, temporal, parietal, occipital, and insular lobes of the brain, images showing a set of regions of interest of the brain.

5. The non-transitory machine-readable medium of claim 4, wherein the parcellation comprises atlas based parcellation delineating 48 cortical and 21 subcortical regions, brain lobe parcellation (10 regions) and model based parcellation delineates 32 regions, wherein the model based parcellation is performed using a third pretrained model, wherein the set of regions of interest includes regions corresponding to brainstem, left-putamen, right-putamen, pallidum, caudate, thalamus, hippocampus, amygdala, accumbens, lateral ventricles, inferior lateral ventricles, and cerebellum.

6. The non-transitory machine -readable medium of claim 4, further comprising: performing segmentation of cerebrovascular system in the brain, comprising: extracting, from the plurality of images, voxels corresponding to vessels based on a vessel processing algorithm, wherein each image is analyzed at multiple scales by convolving the image with Gaussian filters of varying standard deviations, taking second-order partial derivatives at each voxel to generate a matrix, and calculating eigen values from the matrix, the eigen values providing local curvature at that point; andPLDIN2024-90258P-PCTusing a threshold to remove noise from the extracted voxels; and removing voxels corresponding to disconnected vessels; and displaying, on the display unit, a fourth set of images in axial, sagittal, and coronal planes, and in 3D view, wherein the fourth set of images displays the voxels corresponding to the cerebrovascular system in the brain.

7. The non-transitory machine -readable medium of claim 6, further comprising: constructing white matter fibre tract regions of the brain using a Diffusion Tensor Imaging (DTI) deterministic method and Constrained Spherical Deconvolution (CSD) probabilistic method; and displaying, on the display unit, a fifth set of images in axial, sagittal, and coronal planes, and in 3D view, wherein the fifth set of images includes images showing a set of fiber tracts constructed using DTI deterministic method, images showing the set of fibre tracts constructed using CSD probabilistic method, images showing 3D view of the set of fiber tracts and 3D view the distinct subregions of the brain tumor indicating the proximity and / or overlap of the set of fibre tracts with the distinct subregions of the brain tumor.

8. The non-transitory machine-readable medium of claim 7, wherein the constructing using the DTI deterministic method comprises: acquiring diffusion-weighted images in multiple gradient directions with one or more b-values, that characterize the orientation of white matter pathways; computing diffusion metrics comprising fractional anisotropy, mean diffusivity, axial diffusivity, and radial diffusivity; and generating a response function with distortion correction, edema correction for accurate fibre orientation estimation, leading to improved streamline reconstruction, wherein the constructing using the CSD probabilistic method comprises: estimating Fiber Orientation Distribution (FOD); and performing normalization of the estimated FOD.

9. The non-transitory machine -readable medium of claim 7, further comprising: deriving resting state networks of the brain comprising resting-state fMRI (functional MRI), wherein the deriving comprises:PLDIN2024-90258P-PCTaligning the temporal acquisition of slices contained in the plurality of MRI images to allow accurate representation of Blood Oxygen Level-Dependent (BOLD) signal fluctuations across the brain; removing spikes from the BOLD signals and applying and motion correction to minimize head movement artifacts; performing spatial smoothing to enhance signal quality and improve the signal-to- noise ratio; conducting analysis to decompose the signal data into spatially independent components; retaining only components corresponding to resting state networks; and filtering the retained components template matching based on Pearson correlation and Dice score with standard resting state network maps; and displaying, on the display unit, a sixth set of images in axial, sagittal, and coronal planes, and in 3D view, wherein the sixth set of images includes images of a visual network, images of an occipital network and images of an auditory network.

10. The non-transitory machine-readable medium of claim 1, further comprising: converting the CT scan images in the plurality of images to the Hounsfield Unit (HU) range; displaying, on the display unit, a seventh set of images in 3D view, wherein the seventh set of CT images includes images showing the bone tissue of the brain, and images showing the internal structure of the brain.

11. The non-transitory machine-readable medium of claim 10, wherein said plurality of images includes preoperative MRI images and intraoperative CT scans, further comprising: fusing all processed results from the preoperative MRI images with the intraoperative CT scans to perform global correction of brain shift due to CSF leakage occurring during neurosuigery.

12. The non-transitory machine -readable medium of claim 11, wherein the different types of MR sequences comprise structural MRI, diffusion MRI, functional MRI, MR angiography, and CT image.PLDIN2024-90258P-PCT13. A method performed by a neurosurgery planning platform for preoperative, intraoperative planning and postoperative evaluation, the method comprising: receiving a plurality of images of a brain tumor in a brain of a subject, wherein the plurality of images includes magnetic resonance imaging (MRI) images and computed tomography (CT) scan images; pre-processing the plurality of images corresponding to different types of MR sequences to obtain a pre-processed set of images; based on the pre-processed set of images:(A) delineating the brain tumor into distinct subregions including peritumoral edema, enhancing tumor and necrotic core;(B) classifying the brain tissue into a set of regions including white matter (WM), grey matter (GM), corticospinal fluid (CSF) regions;(C) performing a parcellation of the brain into a set of structural regions comprising cortical and subcortical regions;(D) performing segmentation of cerebrovascular system of the brain;(E) constructing white matter fibre tract regions of the brain;(F) deriving resting state networks of the brain; and(G) converting CT scan images contained in the pre-processed set of images to the Hounsfield Unit (HU) range; displaying, on a display unit, an output set of images in axial, sagittal, and coronal planes, and in 3D (three-dimensional) view, wherein the output set of images comprises the pre-processed set of images overlaid with the results of (A) through (G).

14. The method as claimed in claim 13, wherein each image of the plurality of images is according to a Digital Imaging and Communications in Medicine (DICOM) format, wherein the pre-processing comprises: converting the plurality of images to a standard format; performing a field of view (FOV) correction to the plurality of images; removing magnetic field inhomogeneity followed by noise from background in the plurality of images; performing intensity normalization of the plurality of images; andPLDIN2024-90258P-PCTcreating a set of brain masks to be used for skull stripping and brain volume calculation.

15. The method as claimed in claim 14, wherein delineating the brain tumor into distinct subregions comprises applying a first pretrained model to the pre-processed set of images, wherein the output set of images includes images indicating the distinct subregions.

16. The method as claimed in claim 14, wherein classifying the brain tissue into the set of regions comprises: applying a second pretrained model to the images of the skull stripped brain; and estimating based on the set of brain masks, volumetric information of the brain and each of the set of regions, wherein the estimation is performed by first subtracting a tumor core region from the set of tissue masks, wherein the output set of images includes images overlaid with a tissue mask with the tumor core region subtracted, images overlaid with another tissue mask including the tumor core region, images showing each of the set of regions along with the tumor core region.

17. The method as claimed in claim 14, wherein the parcellation comprises atlas based parcellation, brain lobe parcellation and model based parcellation, wherein the model based parcellation is performed using a third pretrained model, wherein the output set of images includes images showing cortical parcellation, images showing subcortical parcellation, images showing the left and right frontal, temporal, parietal, occipital, and insular lobes of the brain, images showing a set of regions of interest of the brain.

18. The method as claimed in claim 14, wherein performing segmentation of cerebrovascular system of the brain comprises: extracting, from the plurality of MRI images, voxels corresponding to vessels based on a vessel processing algorithm, wherein each image is analyzed at multiple scales by convolving the image with Gaussian filters of varying standard deviations, taking second- order partial derivatives at each voxel to generate a matrix, and calculating eigen values from the matrix, the eigen values providing local curvature at that point; andPLDIN2024-90258P-PCTusing a threshold to remove noise from the extracted voxels; and removing voxels corresponding to disconnected vessels, wherein the output set of images includes images displaying the voxels corresponding to the cerebrovascular system in the brain.

19. The method as claimed in claim 14, wherein constructing white matter fibre tract regions of the brain comprises: using a Diffusion Tensor Imaging (DTI) deterministic method comprising: acquiring diffusion-weighted images in multiple gradient directions with one or more b-values, that characterize the orientation of white matter pathways; computing diffusion metrics comprising fractional anisotropy, mean diffusivity, axial diffusivity, and radial diffusivity; and generating a response function with edema correction for accurate fiber orientation estimation, leading to improved streamline reconstruction; and using a Constrained Spherical Deconvolution (CSD) probabilistic method comprising: estimating Fiber Orientation Distribution (FOD); and performing normalization of the estimated FOD, wherein the output set of images includes images showing a set of fiber tracts constructed using DTI deterministic method, images showing the set of fibre tracts constructed using CSD probabilistic method, images showing 3D view of the set of fiber tracts and 3D view the distinct subregions of the brain tumor indicating the proximity and / or overlap of the set of fibre tracts with the distinct subregions of the brain tumor.

20. The method as claimed in claim 14, wherein resting state networks of the brain comprises resting-state fMRI (functional MRI), wherein deriving comprises: aligning the temporal acquisition of slices contained in the pre-processed set of images to allow accurate representation of Blood Oxygen Level-Dependent (BOLD) signal fluctuations across the brain; removing spikes from the BOLD signals and applying and motion correction to minimize head movement artifacts; performing spatial smoothing to enhance signal quality and improve the signal-to- noise ratio;PLDIN2024-90258P-PCTconducting analysis to decompose the signal data into spatially independent components; retaining only components corresponding to resting state networks; and filtering the retained components by template matching based on Pearson correlation and Dice score with standard resting state network maps, wherein the output set of images includes images of a visual network, images of an occipital network and images of an auditory network.

21. The method as claimed in claim 14, wherein the plurality of images includes preoperative MRI images and intraoperative CT scans, the method further comprising: fusing all processed results from the preoperative MRI images with the intraoperative CT scans to perform global correction of brain shift due to CSF leakage occurring during neurosuigery.

22. A digital processing system comprising: a random access memory (RAM) to store instructions forming a neurosurgery planning platform for preoperative, intraoperative planning and postoperative evaluation; and one or more processors to retrieve and execute the instructions, wherein execution of the instructions causes the digital processing system to perform the actions of: providing, by the neurosurgery plarming platform, a user interface that enables a user to view images of a brain tumor in a brain of a subject, the images including: images showing the distinct subregions of the brain tumor including peritumoral edema, enhancing tumor and necrotic core; images showing the classification of the brain tissues into a set of regions including white matter (WM), grey matter (GM), corticospinal fluid (CSF) regions; images showing parcellation of the brain into a set of structural regions comprising cortical and subcortical regions; images showing the cerebrovascular system of the brain; images showing white matter fibre tract regions of the brain; images showing the resting state networks of the brain; and images showing a fusing of all processed results from preoperative MRI images with intraoperative CT scans.PLDIN2024-90258P-PCT

Images