A system for genetic classification of gliomas using artificial intelligence and a method thereof

Abstract

A system (100) for genetic classification of gliomas using artificial intelligence is disclosed. The system includes a processing subsystem which includes a receiving module (108), a sequence identification module (130), a classifying module (112), and an output module (124). The receiving module receives one or more sequences of a plurality of brain scans of a patient. The sequence identification module is configured to be trained to identify the plurality of received magnetic resonance imaging sequences. The classifying module classifies each of the plurality of brain scans by using an artificial intelligence model. The classifying module is configured to estimate the probability of having genetic subtype of glioma using convolutional neural network models. The output module is configured to display the resulting probability of genetic classification of the Glioma using a hierarchical convolution neural network model (128).

Description

[0001] A SYSTEM FOR GENETIC CLASSIFICATION OF GLIOMAS USING ARTIFICIAL INTELLIGENCE AND A METHOD THEREOF

[0002] EARLIEST PRIORITY DATE

[0003] This Application claims priority from a Complete patent application filed in India having Patent Application No. 202321051745, filed on August 01, 2023, and titled “A SYSTEM FOR GENETIC CLASSIFICATION OF GLIOMAS USING ARTIFICIAL INTELLIGENCE AND A METHOD THEREOF”.

[0004] FIELD OF INVENTION

[0005] Embodiments of the present disclosure relate to the field of medicine and diagnosis and more particularly to a system for genetic classification of gliomas using artificial intelligence and a method thereof.

[0006] BACKGROUND

[0007] Glioma refers to a tumor arising from glial cells. It is the most common primary tumor within the brain parenchyma. The world health organization (WHO) classifies Glioma according to different biological characteristics. There are different treatment methods based on this classification. Gliomas are a diverse group of common intracranial tumors that are associated with significant mortality and morbidity. Historically, brain tumors were classified based on histological evaluation.

[0008] Currently, the deep learning field for differentiating tumors such as gliomas has mainly focused on correlating with individual genetic markers. Even though machine learning provides a new perspective to glioma characterization, the selection of features and overfitting of the model made translation into the clinical workflow is difficult. Currently, signs like the T2- fluid-attenuated inversion recovery (FLAIR) mismatch sign even though highly specific, but do not have good sensitivity by the human reader to identify the sign. There is a need for a system that may use artificial intelligence to classify and identify the gliomas. There is no artificial intelligence system currently that analyses multiparametric magnetic resonance imaging (MRI) sequences including diffusion weighted imaging (DWI), Tl, T1 post contrast, FLAIR / T2 to classify gliomas by genetic subtypes.

[0009] Hence, there is a need for a system for genetic classification of gliomas using artificial intelligence and a method thereof that addresses the aforementioned issues.

[0010] OBJECTIVE OF THE INVENTION

[0011] An objective of the present invention is to provide a system for classifying gliomas using artificial intelligence by analyzing multiparametric routine magnetic resonance imaging (MRI) sequences.

[0012] Another objective of the present invention is to reduce lag of diagnosis by providing radio-genomic correlation based on non-invasive imaging of magnetic resonance imaging.

[0013] Further, another objective of the present invention is to demarcate the tumor portion using artificial intelligence.

[0014] Furthermore, another objective of the present invention is to classify the glioma in one of three genetic subtypes using a novel hierarchical deep learning model which comprises of multiple convoluted neural networks analyzing multiparametric magnetic resonance imaging brain sequences. BRIEF DESCRIPTION

[0015] In accordance with one embodiment of the disclosure, a system for genetic classification of gliomas using artificial intelligence is provided. The system includes a processing subsystem hosted on a server and configured to execute on a network to control bidirectional communications among a module. The plurality of modules includes a receiving module, a classifying module, and an output module. The receiving module configured to receive one or more sequences of a plurality of magnetic resonance imaging brain scan of a patient. The sequence identification module is operatively coupled with the receiving module. The sequence identification module includes a sequence identification convolution neural network (CNN) configured and to be trained to identify the plurality of received magnetic resonance imaging sequences. The classifying module is operatively coupled with the receiving module. The classifying module classifies glioma in each of the magnetic resonance imaging brain sequence by using an artificial intelligence model. The classifying module is configured to estimate the probability of identifying glioma in plurality of magnetic resonance imaging sequences by using a convolutional neural network model for respective sequences. Each convolutional neural network model is a trained deep learning model on particular magnetic resonance imaging brain sequence. Each convolutional neural network model includes a plurality of layers of neural networks. The convolutional neural network model for assessment of T1 post-contrast sequence is a central necrosis convolutional neural network model. The convolutional neural network model for assessment of diffusion-weighted sequence is diffusion-weighted imaging convolutional neural network model. The convolutional neural network model for assessment of T2 and fluid-attenuated inversion recovery sequences is T2- fluid- attenuated inversion recovery convolutional neural network model, and convolutional neural network model for assessment T1 sequence is, continuous cortex convolutional neural network model. The output module is operatively coupled with the classifying module. The output module is configured to display the resulting probability of genetic classification of the glioma using a hierarchical convolution neural network model.

[0016] In accordance with another embodiment, a method for genetically classifying gliomas using artificial intelligence is provided. The method includes receiving, by a receiving module of a processing subsystem, one or more sequences of a plurality of magnetic resonance imaging brain scans of a patient. The method also includes training, by a sequence identification convolution neural network to identify the plurality of received magnetic resonance imaging sequences Further, the method includes classifying, by a classifying module of the processing subsystem, each of the plurality of magnetic resonance imaging brain scans by using an artificial intelligence model. Furthermore, the method includes estimating, by the classifying module of the processing subsystem, a probability of identifying glioma in plurality of magnetic resonance imaging sequences by using a convolutional neural network model for respective sequences. Furthermore, the method includes providing, by the classifying module of the processing subsystem, the convolutional neural network model for assessment of diffusion-weighted sequence is, diffusion-weighted imaging convolutional neural network model. Furthermore, the method includes providing, by the classifying module of the processing subsystem, the convolutional neural network model for assessment of T2 and fluid-attenuated inversion recovery sequences is T2- fluid- attenuated inversion recovery convolutional neural network model. Furthermore, the method includes providing, by the classifying module of the processing subsystem, the convolutional neural network model assessment T1 sequence is, continuous cortex convolutional neural network model. Furthermore, the method includes displaying, by an output module of the processing subsystem, the resulting probability of genetic classification of the glioma using a hierarchical convolution neural network model.

[0017] To further clarify the advantages and features of the present disclosure, a more particular description of the disclosure will follow by reference to specific embodiments thereof, which are illustrated in the appended figures. It is to be appreciated that these figures depict only typical embodiments of the disclosure and are therefore not to be considered limiting in scope. The disclosure will be described and explained with additional specificity and detail with the appended figures.

[0018] BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The disclosure will be described and explained with additional specificity and detail with the accompanying figures in which:

[0020] FIG. 1 is a block diagram representing a system for genetic classification of gliomas using artificial intelligence in accordance with an embodiment of the present disclosure;

[0021] FIG. 2 is a flow chart representing an exemplary embodiment of the system for genetic classification of gliomas using artificial intelligence of FIG. 1 in accordance with an embodiment of the present disclosure;

[0022] FIG. 3 a is a schematic representation of a resulting probability by the classifying module of FIG. 1 in accordance with another embodiment of the present disclosure;

[0023] FIG. 3b is a schematic representation of an exemplary embodiment of the resulting probability by the classifying module of FIG. 1 in accordance with another embodiment of the present disclosure;

[0024] FIG. 3c is a schematic representation of another exemplary embodiment of the resulting probability by the classifying module of FIG. 1 in accordance with another embodiment of the present disclosure;

[0025] FIG. 4 is a block diagram of a computer or a server for the system for genetic classification of gliomas using artificial intelligence in accordance with an embodiment of the present disclosure; and FIG. 5 is a flow chart representing steps involved in a method for genetically classifying gliomas using artificial intelligence in accordance with an embodiment of the present disclosure.

[0026] Further, those skilled in the art will appreciate that elements in the figures are illustrated for simplicity and may not have necessarily been drawn to scale. Furthermore, in terms of the construction of the device, one or more components of the device may have been represented in the figures by conventional symbols, and the figures may show only those specific details that are pertinent to understanding the embodiments of the present disclosure so as not to obscure the figures with details that will be readily apparent to those skilled in the art having the benefit of the description herein.

[0027] DETAILED DESCRIPTION

[0028] For the purpose of promoting an understanding of the principles of the disclosure, reference will now be made to the embodiment illustrated in the figures, and specific language will be used to describe them. It will nevertheless be understood that no limitation of the scope of the disclosure is thereby intended. Such alterations and further modifications in the illustrated system, and such further applications of the principles of the disclosure as would normally occur to those skilled in the art are to be construed as being within the scope of the present disclosure.

[0029] The terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process or method that comprises a list of steps does not include only those steps but may include other steps not expressly listed or inherent to such a process or method. Similarly, one or more devices or subsystems or elements or structures or components preceded by "comprises... a" does not, without more constraints, preclude the existence of other devices, subsystems, elements, structures, components, additional devices, additional subsystems, additional elements, additional structures or additional components. Appearances of the phrase "in an embodiment", "in another embodiment" and similar language throughout this specification may, but not necessarily do, all refer to the same embodiment.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this disclosure belongs. The system, methods, and examples provided herein are only illustrative and not intended to be limiting.

[0031] In the following specification and the claims, reference will be made to a number of terms, which shall be defined to have the following meanings. The singular forms "a", "an", and "the" include plural references unless the context clearly dictates otherwise.

[0032] Embodiments of the present disclosure relate to a system for genetic classification of gliomas using artificial intelligence. The system includes a processing subsystem hosted on a server and configured to execute on a network to control bidirectional communications among a module. The plurality of modules includes a receiving module, a classifying module, and an output module. The receiving module configured to receive one or more sequences of a plurality of magnetic resonance imaging brain scan of a patient. The sequence identification module is operatively coupled with the receiving module. The sequence identification module includes a sequence identification convolution neural network configured and has been trained to identify the plurality of received magnetic resonance imaging sequences. The classifying module is operatively coupled with the receiving module and sequence identification module. The classifying module classifies each of the magnetic resonance imaging brain sequence by using sequence identification convolution neural network. The classifying module is configured to estimate the probability of identifying glioma in plurality of magnetic resonance imaging sequences by using a convolutional neural network model for respective sequences. Each convolutional neural network model is trained deep learning model on particular magnetic resonance imaging brain sequence. Each convolutional neural network model includes a plurality of layers of neural networks. The convolutional neural network model for assessment of T1 post-contrast sequence is a central necrosis convolutional neural network model. The convolutional neural network model for assessment of diffusion-weighted sequence is diffusion- weighted imaging convolutional neural network model. The convolutional neural network model for assessment of T2 and fluid-attenuated inversion recovery sequences is T2- fluid-attenuated inversion recovery convolutional neural network model, and convolutional neural network model for assessment T1 sequence is, continuous cortex convolutional neural network model. The output module is operatively coupled with the classifying module. The output module is configured to display the resulting probability of genetic classification of the glioma using a hierarchical convolution neural network model.

[0033] FIG. 1 is a block diagram representing the system for genetic classification of gliomas using artificial intelligence in accordance with an embodiment of the present disclosure. The system includes aprocessing subsystem (102), hosted on a server (104), and configured to execute on a network (106) to control bidirectional communications among a plurality of modules. In one embodiment, the server (104) may include a cloud server. In another embodiment, the server (104) may include a local server. The processing subsystem (102) is configured to execute on a network (106) to control bidirectional communications among a plurality of modules. In one embodiment, the network (106) may include a wired network such as a local area network (LAN). In another embodiment, the network may include a wireless network such as Wi-Fi, Bluetooth, Zigbee, near-field communication (NFC), infrared communication (RFID), or the like. The plurality of modules includes a receiving module (108), a sequence identification module (130), a classifying module (112), and an output module (124).

[0034] The receiving module (108) is configured to receive one or more sequences of a plurality of magnetic resonance imaging brain scans (110) of a patient. In one embodiment, the plurality of brain scans (110) includes at least one magnetic resonance imaging sequence of MRI brain which may be Tl, T1 post contrast, T2, fluid- attenuated inversion recovery (FAIR) and diffusion-weighted imaging (DWI).

[0035] The sequence identification module (130) is operatively coupled with the receiving module (108). The sequence identification module (130) includes a sequence identification convolution neural network configured to be trained to identify the plurality of received magnetic resonance imaging sequences.

[0036] The classifying module (112) is operatively coupled with the receiving module (108) and sequence identification module (130). The classifying module (112) is configured to classify each of the plurality of magnetic resonance imaging brain scans (110) by using sequence identification convolution neural network. The classifying module (112) is also configured to estimate the probability of identifying a plurality of magnetic resonance imaging sequences by using a convolutional neural network model (114). The classifying module (112) classifies each of the plurality of brain scans (110) by using an artificial intelligence model. The classifying module is further configured to estimate the probability of identifying glioma in plurality of MRI sequences by using a convolutional neural network model (CNN)for respective sequences. Each CNN is trained deep learning model on particular MRI brain sequences. Each CNN comprises of a plurality of layers of neural networks. In one embodiment, the plurality of layers comprises 23 layers. The CNN for assessment of Tl post-contrast sequence is, a central necrosis CNN (116). The CNN for assessment of diffusion-weighted sequence is diffusion -weighted imaging CNN (118). CNN for assessment of T2 and fluid- attenuated inversion recovery (FLAIR) sequences is T2- FLAIR CNN (120), and CNN for assessment Tl sequence is, continuous cortex CNN (122). Expert neuroradiologists demarcated the abnormal tumor portion in the corresponding sequences to train corresponding convolutional Neural Network (CNN). This demarcated portion data was correlated with genetic data to train corresponding convolutional Neural Network (CNN) models to identify the genetic subtype of glioma in corresponding MRI sequences.

[0037] The output module (124) is operatively coupled with the classifying module (112), wherein the output module (124) is configured to display the resulting probability of genetic classification of the Glioma using a hierarchical convolution neural network model (128). In one embodiment, the output module (124) is configured to analyze each sequence with a specific weightage assigned to each sequence.

[0038] In one embodiment, the system (100) provides a hybrid diagnosis. Neuroradiologists experienced in radio genomics read the MRI scan for glioma detection. The results of the neurologists and the Glioma- Al system (100) output are compared. Inter-observer variability may be calculated with the Kappa coefficient (K). 0.2<K<0.4 indicates fair agreement, 0.4<K<0.6 indicates moderate agreement, and K>0.6 indicates substantial agreement. In one embodiment, the kappa coefficient is a statistic that is used to measure inter-rater reliability for checking the performance of Glioma-AI with expert human reader.

[0039] FIG. 2 is a flow chart representing an exemplary embodiment of the system for genetic classification of gliomas artificial intelligence of FIG. 1 in accordance with an embodiment of the present disclosure. In one embodiment, the stratifying module (112) is configured to identify an abnormal tumor portion in one or more sequences of the plurality of brain scans (110). In one embodiment, the convolution neural network model ( 114) is of twenty-three layers and configured to train a plurality of deep learning models used in the stratifying module (112). In another embodiment, the plurality of trained deep learning models differentiates between the plurality of first of gliomas and the plurality of second glioma for each sequence of the brain scans (110). In one embodiment, wherein the accuracy of the deep learning model is evaluated by a crosstable analysis performed using a precision, a recall, and a machine learning metric. In one embodiment, the system (100) includes a compilation module (126) for compiling the resulting probability obtained by the output module (124). The cross-table analysis was also performed for the deployment trial of the system (100) to assess the sensitivity and specificity of the deep learning model.

[0040] Considering a non-limiting example where a neuroradiologist identifies the glioma based on patient’s history studied by the neurologist or neurosurgeon and clinical features. The neuroradiologist then studies all sequences to assess whether the examination confines to a good quality for diagnostic purposes. The flowchart in FIG. 2 is followed to reach the final diagnosis about glioma type. Based on this analysis, a novel hierarchical deep learning model is built. In one embodiment, a logic-embedded software is designed for radiologists to answer based on history and clinical features is there a glioma, and whether the examination is of good quality for diagnostic purposes. When the Glioma is more probable diagnosis after ruling out other disease for example stroke or infection, then the images are added to the system (100). This is a pre-requisite step. The system (100) identifies one or more sequences, which may be neglected by previous deep learning studies. In the next step, the deep learning models analyses the MRI sequences which are trained to differentiate between three types of gliomas, for each sequence. The convolutional Neural Network (CNN) model (114) of 23 layers is used to train the deep learning models. In the fourth step, the system (100) is used to stack the magnetic resonance imaging (MRI) sequences with specific weightage to each sequence. The final diagnosis is compiled by using the system (100).

[0041] CNN model training: Expert neuroradiologists demarcated the abnormal tumor portion in the corresponding sequences to training corresponding convolutional Neural Network (CNN). This demarcated portion data was correlated with genetic data to train model to identify the genetic subtype of glioma. Each CNN model is trained for corresponding sequences, for example, central necrosis CNN for T1 post contrast sequence. Hierarchical models are trained for final integrated diagnosis. This is trained against the genetic subtype of glioma by stacking the CNN models. Thus, for each glioma there is confidence interval for CNN for each particular sequence and confidence interval for overall integrated diagnosis.

[0042] A rationale behind having different CNN models rather than directly having hierarchical model for final diagnosis: Patients with brain tumor are sometimes not steady in the MRI machine. They may move in between the scan or during a particular sequence. That may make assessment of particular sequence difficult. If the patient has moved in more than one sequence, then results of hierarchical CNN may be affected. In such cases neuroradiologist cross check the results of the artificial intelligence model. The artificial intelligence model may take a call whether to trust confidence interval of CNN models in sequences. The CNN model sequences are clear and omit the results of sequences which have artefacts due to motion. In such exceptional circumstances diagnosis may be predicted with individual CNN models rather than hierarchical models. That is why having individual CNN model for each sequence and hierarchical model for all CNNs, is important in analysis of multiparametric MRI sequence using artificial intelligence. This is often overlooked by previous deep learning models.

[0043] FIG. 3 a is a schematic representation of a resulting probability by the classifying module of FIG. 1 in accordance with another embodiment of the present disclosure. In this MRI scan of patient with glioma, the abnormal tumor area is demarcated by the artificial intelligence model using Central necrosis CNN (202), T2-FLAIR CNN (204), DWI CNN (206) and Continuous cortex CNN (208), in corresponding sequences, accompanied by confidence interval for each sequence available in this patient’s scan. The output module configured to display the resulting probability of genetic classification of the Glioma using a hierarchical convolution neural network model (128) is displayed as integrated diagnosis with confidence interval for that diagnosis. Diagnosis predicted in this tumor by artificial intelligence system is IDH mutant lpl9q codeleted glioma.

[0044] FIG. 3b is a schematic representation of an exemplary embodiment of the resulting probability by the classifying module of FIG. 1 in accordance with another embodiment of the present disclosure. In this MRI scan of patient with glioma, the abnormal tumor area is demarcated by the artificial intelligence model using Central necrosis CNN (210), T2-FLAIR CNN (412), DWI CNN (214) and Continuous cortex CNN (216), in corresponding sequences, accompanied by confidence interval for each sequence available in this patient’s scan. The output module configured to display the resulting probability of genetic classification of the Glioma using a hierarchical convolution neural network model (128) is displayed as integrated diagnosis with confidence interval for that diagnosis. Diagnosis predicted in this tumor by artificial intelligence system is IDH mutant lpl9q non-codeleted glioma.

[0045] FIG. 3c is a schematic representation of another exemplary embodiment of the resulting probability by the classifying module of FIG. 1 in accordance with another embodiment of the present disclosure. In this MRI scan of patient with glioma, the abnormal tumor area is demarcated by the artificial intelligence model using Central necrosis CNN (218), T2-FLAIR CNN (420), DWI CNN (222) and Continuous cortex CNN (244), in corresponding sequences, accompanied by confidence interval for each sequence available in this patient’s scan. The output module configured to display the resulting probability of genetic classification of the Glioma using a hierarchical convolution neural network model is displayed as integrated diagnosis with confidence interval for that diagnosis. Diagnosis predicted in this tumor by artificial intelligence system is IDH wildtype glioma.

[0046] FIG. 4 is a block diagram of a computer or a server for the system for genetic classification of gliomas using artificial intelligence in accordance with an embodiment of the present disclosure. The server (300) includes a processor(s) (302), and memory (306) operatively coupled to the bus (304).

[0047] The processor(s) (302), as used herein, means any type of computational circuit, such as, but not limited to, a microprocessor, a microcontroller, a complex instruction set computing microprocessor, a reduced instruction set computing microprocessor, a very long instruction word microprocessor, an explicitly parallel instruction computing microprocessor, a digital signal processor, or any other type of processing circuit, or a combination thereof.

[0048] The bus (304) as used herein refers to internal memory channels or computer network that is used to connect computer components and transfer data between them. The bus (304) includes a serial bus or a parallel bus, wherein the serial bus transmits data in a bit-serial format and the parallel bus transmits data across multiple wires. The bus (304) as used herein, may include but not limited to, a system bus, an internal bus, an external bus, an expansion bus, a frontside bus, a backside bus, and the like.

[0049] The memory (306) includes a plurality of subsystems and a plurality of modulesstored in the form of an executable program which instructs the processor (302) to perform the method steps illustrated in FIG. 1. The memory (306) is substantially similar to the system of FIG.l. The memory (306) has submodules: a receiving module (108), a sequence identification module (130), a classifying module (112), and an output module (124).

[0050] The receiving module (108) is configured to receive one or more sequences of a plurality of magnetic resonance imaging brain scans (110) of a patient. In one embodiment, the plurality of brain scans (110) includes at least one magnetic resonance imaging sequence of MRI brain which may be Tl, T1 post contrast, T2, fluid- attenuated inversion recovery (FAIR) and diffusion-weighted imaging (DWI). The sequence identification module (130) is operatively coupled with the receiving module (108). The sequence identification module (130) includes a sequence identification convolution neural network configured to be trained to identify the plurality of received magnetic resonance imaging sequences.

[0051] The classifying module (112) is configured to estimate the probability of identifying glioma in plurality of MRI sequences by using a convolutional neural network model aka CNN for respective sequences. Each CNN is trained deep learning model on particular MRI brain sequences. Each CNN comprises of 23 layers of neural networks. CNN for assessment of T1 post-contrast sequence is, a central necrosis CNN (116). CNN for assessment of diffusion-weighted sequence is, DWI CNN (118). CNN for assessment of T2 and FLAIR sequences is T2- FLAIR CNN (120), and CNN for assessment T1 sequence is continuous cortex CNN (122).

[0052] The output module (124) is operatively coupled with the classifying module (112), wherein the output module (124) is configured to display the resulting probability of genetic classification of the Glioma using a hierarchical convolution neural network model (128).

[0053] Computer memory elements may include any suitable memory device(s) for storing data and executable program, such as read-only memory, random access memory, erasable programmable read-only memory, electrically erasable programmable readonly memory, hard drive, removable media drive for handling memory cards and the like. Embodiments of the present subject matter may be implemented in conjunction with program modules, including functions, procedures, data structures, and application programs, for performing tasks, or defining abstract data types or low-level hardware contexts. An executable program stored on any of the above-mentioned storage media may be executable by the processor(s) (402). In one embodiment, a training dataset is obtained from TCIA dataset which had patient’s consent and other approval. In a pre-requisite step, logic-embedded software is designed for radiologists to answer based on history and clinical features is there a glioma, and whether the examination is of good quality for diagnostic purposes. When the Glioma is probably diagnosed after ruling out other disease such as stroke or infection, then the images are added to the system. In the next step sequence identification CNN is trained to identify the plurality of uploaded MRI sequences. In the second step, expert neuroradiologists demarcated the abnormal tumor portion in the corresponding sequences to train corresponding convolutional Neural Network (CNN). This demarcated portion data was correlated with genetic data to train model to identify the genetic subtype of glioma. Each CNN model is trained for corresponding sequences. Central necrosis CNN for T1 post contrast sequence, DWI CNN for diffusion weighted sequence, T2-FLAIR CNN for T2 and FLAIR sequence, Continuous cortex CNN for T1 sequence.

[0054] The training method also includes training of the hierarchical model for final integrated diagnosis. This hierarchical model was trained against the genetic subtype of glioma by stacking the CNN models.

[0055] FIG. 6 is a flow chart representing steps involved in a method (400) for operating the system for genetically classifying gliomas using artificial intelligence when system is deployed in clinical practice in accordance with an embodiment of the present disclosure. The method (4) receiving, by a receiving module of a processing subsystem, one or more sequence of a plurality of brain scans of a patient in step (402).

[0056] The method also includes training, a sequence identification convolution neural network to identify the plurality of received magnetic resonance imaging sequences in step (404). The method (400) also includes classifying, by a classifying module of the processing subsystem, each of the plurality of brain scans by using an artificial intelligence model in step (406). The method also includes identifying an abnormal tumor portion in one or more sequences of the plurality of brain scans.

[0057] Further, the method (400) includes estimating, by the classifying module of the processing subsystem, a probability of identifying glioma in plurality of magnetic resonance imaging sequences by using a convolutional neural network model for respective sequences in step (408). In one embodiment, each CNN is trained deep learning model on particular MRI brain sequences. Each CNN comprises of 23 layers of neural networks. CNN for assessment of T1 post-contrast sequence is, a central necrosis CNN. CNN for assessment of diffusion-weighted sequence is DWI CNN. CNN for assessment of T2 and FLAIR sequences is T2- FLAIR CNN, and CNN for assessment T1 sequence is continuous cortex CNN. Expert neuroradiologists demarcated the abnormal tumor portion in the corresponding sequences to train corresponding convolutional Neural Network (CNN). This demarcated portion data was correlated with genetic data to train corresponding convolutional Neural Network (CNN) models to identify the genetic subtype of glioma in corresponding MRI sequences.

[0058] Furthermore, the method (400) includes providing, by the classifying module of the processing subsystem, the convolutional neural network model for assessment of diffusion-weighted sequence is, diffusion-weighted imaging convolutional neural network model in step (410).

[0059] Furthermore, the method (400) includes providing, by the classifying module of the processing subsystem, the convolutional neural network model for assessment of T2 and fluid-attenuated inversion recovery sequences is T2- fluid-attenuated inversion recovery convolutional neural network model in step (412). Furthermore, the method (400) includes providing, by the classifying module of the processing subsystem, the convolutional neural network model assessment T1 sequence is, continuous cortex convolutional neural network model in step (414).

[0060] Furthermore, the method (400) includes displaying, by an output module of the processing subsystem, the resulting probability of genetic classification of the Glioma using a hierarchical convolution neural network model in step (416). Hierarchical model was trained for final integrated diagnosis. This is trained against the genetic subtype of glioma by stacking the CNN models. Thus, for each glioma there is confidence interval for CNN for each particular sequence and confidence interval for overall integrated diagnosis.

[0061] The method also includes receiving one or more sequences of a plurality of magnetic resonance imaging brain scan of a patient from training dataset. In one embodiment a pre-requisite step is designing, a logic-embedded software for radiologists to answer based on history and clinical features is there a glioma. The method also includes training, a sequence identification CNN to identify the plurality of uploaded MRI sequences. The method also includes expertizing, the neuroradiologists demarcated the abnormal tumour portion in the corresponding sequences to train corresponding CNN. The demarcated portion data is correlated with genetic data to train model to identify the genetic subtype of glioma. The method also includes training, each CNN model for corresponding sequences with respect to central necrosis CNN for T1 post contrast sequence, DWI CNN for diffusion weighted sequence, T2-FLAIR CNN for T2 and FLAIR sequence, Continuous cortex CNN for T1 sequence. This hierarchical model is trained against the genetic subtype of glioma by stacking the CNN models.

[0062] In one embodiment, the method includes submitting the radiology scans by the user for glioma characterization. The MRI data may be anonymized in the uploaded sequences. The user may get a hybrid diagnosis result after results by the output module are checked by the neuroradiologists. This is done to ensure artificial intelligence fails safe in exceptional circumstances when MRI sequences can have artefacts due to motion or other reasons for example, hemorrhage, external devices, wrap around artefact etc.

[0063] Various embodiments of the present disclosure provide the genetic classification of gliomas using artificial intelligence. The system disclosed in the present disclosure decreased the diagnosis lag by providing timely radio-genomic correlation based on non-invasive MRI imaging. The system disclosed in the present disclosure applies an artificial intelligence model which significantly impacts on patient management especially when stereotactic biopsy and genetic testing are limited by co-morbidities or financial constraints. The invention in the present disclosure, stratifies gliomas with an imaging-based classification method with radio genomics correlation. The system enables the user to upload the results of histopathology as feedback into the system.

[0064] While specific language has been used to describe the disclosure, any limitations arising on account of the same are not intended. As would be apparent to a person skilled in the art, various working modifications may be made to the method in order to implement the inventive concept as taught herein.

[0065] The figures and the foregoing description give examples of embodiments. Those skilled in the art will appreciate that one or more of the described elements may well be combined into a single functional element. Alternatively, certain elements may be split into multiple functional elements. Elements from one embodiment may be added to another embodiment. For example, the order of processes described herein may be changed and is not limited to the manner described herein. Moreover, the actions of any flow diagram need not be implemented in the order shown; nor do all of the acts need to be necessarily performed. Also, those acts that are not dependent on other acts may be performed in parallel with the other acts. The scope of embodiments is by no means limited by these specific examples.

Claims

I CLAIM:

1. A system (100) for genetic classification of gliomas using artificial intelligence comprising: a processing subsystem (102), hosted on a server (104), and configured to execute on a network (106) to control bidirectional communications among a module comprising: a receiving module (108) configured to receive one or more sequences of a plurality of magnetic resonance imaging brain scans (110) of a patient; a sequence identification module ( 130) operatively coupled with the receiving module (108), wherein the sequence identification module (130) comprises a sequence identification convolution neural network configured to be trained to identify the plurality of received magnetic resonance imaging sequences; a classifying module (112) operatively coupled with the receiving module (108) and the sequence identification module (130), wherein the classifying module (112) classifies each of the magnetic resonance imaging brain scans (110) by using an artificial intelligence model, wherein the classifying module (112) is configured to: estimate a probability of identifying glioma in plurality of magnetic resonance imaging sequences by using a convolutional neural network model (114) for respective sequences,wherein each convolutional neural network model (114) is trained deep learning model on particular magnetic resonance imaging brain sequences, wherein each convolutional neural network model (114) comprises of a plurality of layers of neural networks, wherein the convolutional neural network model (114) for assessment of T1 post-contrast sequence is, a central necrosis convolutional neural network model (116), wherein the convolutional neural network model (114) for assessment of diffusion-weighted sequence is, diffusion- weighted imaging convolutional neural network model (118), and wherein the convolutional neural network model (114) for assessment of T2 and fluid-attenuated inversion recovery sequences is T2- fluid-attenuated inversion recovery convolutional neural network model (120), and wherein the convolutional neural network model (114) assessment T1 sequence is, continuous cortex convolutional neural network model (122); and an output module (124) operatively coupled with the classifying module (112), wherein the output module (124) is configured to display the resulting probability of genetic classification of the glioma using a hierarchical convolution neural network model (128).

2. The system (100) as claimed in claim 1, wherein the system identifies the plurality of sequences of brain scan (110), comprises of Tl, T2, fluid-attenuated inversion recovery, diffusion- weighted imaging and Tl post contrast sequences.

3. The system (100) as claimed in claim 1, wherein the classifying module (112) is configured to identify an abnormal tumor portion in one or more sequences of the plurality of brain scan (110).

4. The system (100) as claimed in claim 1, wherein each of twenty-three layers of the convolution neural network models (114) is configured to classify for corresponding magnetic resonance imaging brain scans (110).

5. The system (100) as claimed in claim 1, wherein the classifying module (112) is configured to estimate a probability of identifying glioma in plurality of magnetic resonance imaging sequences by using a convolutional neural network model for respective sequences.

6. The system (100) as claimed in claim 1, wherein each convolution neural network comprises of 23 layers of neural networks.

7. The system (100) as claimed in claim 1, wherein the output module (124) is operatively coupled with the classifying module (112), wherein the output module (124) is configured to display the resulting probability of genetic classification of the Glioma using a hierarchical convolution neural network model (128).

8. A method (400) for genetically classifying of gliomas using artificial intelligence comprises: receiving, by a receiving module of a processing subsystem, one or more sequence of a plurality of magnetic resonance imaging brain scans of a patient;training, a sequence identification convolution neural network to identify the plurality of received magnetic resonance imaging sequences; (404) classifying, by a classifying module of the processing subsystem, each of the plurality of magnetic resonance imaging brain scans by using an artificial intelligence model; (406) estimating, by the classifying module of the processing subsystem, a probability of identifying glioma in plurality of magnetic resonance imaging sequences by using a convolutional neural network model for respective sequences; (408) providing, by the classifying module of the processing subsystem, the convolutional neural network model for assessment of diffusion-weighted sequence is, diffusion-weighted imaging convolutional neural network model; (410) providing, by the classifying module of the processing subsystem, the convolutional neural network model for assessment of T2 and fluid-attenuated inversion recovery sequences is T2- fluid-attenuated inversion recovery convolutional neural network model; (412) providing, by the classifying module of the processing subsystem, the convolutional neural network model assessment T1 sequence is, continuous cortex convolutional neural network model; (414) and displaying, by an output module of the processing subsystem, the resulting probability of genetic classification of the glioma using a hierarchical convolution neural network model. (416)

Images