1. A system (300) for classifying brain tumors using AI-based transfer learning, comprising: a magnetic resonance imaging (MRI) module (302) configured to generate brain images showing non-tumor and tumor areas and to determine the shape and location of the tumor in the brain; a pre-processing module (304) configured to filter noise from the images and enhance image contrast using contrast limited adaptive histogram equalization; a feature extraction module (306) configured to extract spatial grey level dependency matrix features from the pre-processed images; a data division module (308) configured to divide the dataset into a training dataset and a test dataset comprising the remaining images; an image augmentation module (310) configured to manipulate and generate altered versions of images in the training dataset; a deep learning-based feature extraction module (312) configured to extract features from the training dataset using a pre-trained convolutional neural network (CNN) model; a classification module (314) configured to classify the tumor based on the extracted features and transfer learning techniques, wherein the system automatically separates brain tumor images into benign and malignant categories.

2. The system as claimed in claim 1, wherein the pre-processing module (304) filters noise from the images using Laplacian of Gaussian filtering method.

3. The system as claimed in claim 1, wherein the MRI module (302) further comprises a segmentation module configured to segment the tumor from the brain images.

4. The system as claimed in claim 1 or 2, wherein the pre-trained CNN model is selected from GoogLeNet, ResNet50, ResNet100, AlexNet, and SqueeZeNet.

5. The system as claimed in any one of claims 1 to 3, wherein the pre-processing module (306) further comprises a reshaping module configured to re-shape the images after noise removal.

6. The system as claimed any one of claims 1 to 4, wherein the data division module (308) further comprises a validation dataset comprising a portion of the training dataset for model evaluation.

7. The system as claimed in any one of claims 1 to 5, wherein the image aug-mentation module (310) further comprises techniques for varying coloring and lighting conditions to increase classifier robustness.

8. The system as claimed in any one of claims 1 to 6, wherein the classification module (314) employs a softmax layer for final classification of tumor images into benign and malignant categories.

9. The system as claimed in any one of claims 1 to 7, wherein the deep learn-ing-based feature extraction module (312) is trained to avoid overfitting and under-fitting of the model on the test dataset.

10. The system as claimed in any one of claims 1 to 8, further comprising a dis-play module configured to present the classified tumor images to a user, aiding in medical diagnosis and treatment planning.

11. The system of any one of the claims 1 to 9, wherein the classification mod-ule (314) provides output indicative of the exact position of the tumor within the brain.

Description:FIELD OF THE INVENTION
[0001] The invention pertains to the field of medical imaging and computational healthcare systems. Specifically, the invention pertains to an innovative deep learn-ing-based framework for automated classification of brain tumors using MRI imag-es and transfer learning techniques.

BACKGROUND OF THE INVENTION
[0002] A brain tumor constitutes an aggregation of abnormal tissues within the brain, resulting from unchecked cellular growth. This growth pattern is erratic and disrupts the brain's normal functions, leading to a spectrum of symptoms such as headaches, dizziness, fainting spells, and even paralysis. Often, individuals may not immediately recognize these symptoms as indicative of a serious illness, further complicating timely diagnosis and treatment. The uncontrolled proliferation of tu-mor cells can severely impede proper brain function, potentially culminating in fa-tality if left unchecked. Despite significant advancements in machine intelligence and biomedical techniques over recent years, the intricate and dynamic nature of brain tumors poses a persistent challenge. The variability and instability inherent in tumor growth render conventional diagnostic approaches inadequate, perpetuating the enduring threat this disease poses to humanity's well-being.
[0003] Brain tumor cancer represents one of the fastest-growing diseases in to-day's medical landscape. The brain, being the most intricate organ in the human body, orchestrates a multitude of vital functions, including respiration, sensory per-ception, and motor coordination, through its complex network of nerve cells and tis-sues. While these cells possess the natural capacity to proliferate, certain aberra-tions cause them to lose their regulatory mechanisms, leading to the formation of tumors within the brain. Over time, these tumors encroach upon neighboring brain regions, disrupting their normal activities and causing dysfunction. Broadly catego-rized into benign and malignant types, benign tumors pose comparatively lower risks as they are non-invasive and tend to remain localized. In contrast, malignant tumors, characterized by their cancerous nature, present a grave threat as they can metastasize to distant parts of the body if left untreated. Primary malignant tumors originate within the brain itself, whereas secondary malignant tumors, stemming from elsewhere in the body, infiltrate the brain, exacerbating the severity of the condition.
[0004] MRI (Magnetic Resonance Imaging) stands out as the preferred method for tumor detection due to its safety and exceptional image quality. The clarity and high resolution of MRI images facilitate precise identification and delineation of tumor-affected areas during image processing. Operating on the principles of magnetic fields, MRI capitalizes on the fact that approximately 70% of the human body com-prises water molecules. When subjected to a magnetic field, the protons within these water molecules align at specific angles. Upon removal of the magnetic field, the protons return to their relaxed state, with the time taken for alignment and relaxa-tion differing in regions with varying water content, such as those affected by tu-mors. By measuring this time differential, MRI generates detailed brain images that highlight both tumor and non-tumor regions. While MRI effectively captures brain imagery, the crux lies in discerning tumor types and precisely determining their lo-cation within the brain. This underscores the critical role of MRI in aiding clinicians in diagnosing brain tumors accurately and facilitating targeted treatment planning.
[0005] One of the primary challenges in diagnosing brain tumors lies in accurately determining the extent and composition of tumor cells using MRI imaging. To ad-dress this challenge, researchers and clinicians employ a combination of conven-tional machine learning techniques and advanced deep learning methodologies, par-ticularly leveraging convolutional neural networks (CNNs). CNNs represent a class of deep learning models specifically designed for image analysis tasks.
[0006] At the heart of CNNs are convolutional layers, which play a pivotal role in automatically learning intricate patterns and features from input data, such as MRI brain scans. These layers consist of learnable filters, or kernels, which are con-volved with the input image data by sliding across its spatial dimensions. Through this process, CNNs extract local features from the input image, while preserving spatial relationships within the data. This inherent capability of CNNs to capture both low-level and high-level features makes them particularly well-suited for tasks requiring nuanced analysis of complex image data, such as identifying and charac-terizing brain tumors.
[0007] By harnessing the power of CNNs and deep learning techniques, research-ers and medical practitioners aim to enhance the accuracy and efficiency of brain tumor detection and characterization from MRI scans. These advanced computa-tional approaches offer promising avenues for overcoming the inherent challenges associated with traditional methods, ultimately facilitating more precise diagnosis and personalized treatment strategies for patients with brain tumors.
[0008] In view of the foregoing disadvantages, there is a pressing need for the de-velopment and implementation of advanced computational methodologies and algo-rithms that can dynamically adapt to evolving datasets, thereby improving the accu-racy, efficiency, and reliability of brain tumor detection and classification from MRI images.

SUMMARY OF THE INVENTION
[0009] To address the foregoing problems, in whole or in part, and/or other problems that may have been observed by persons skilled in the art, the present disclosure provides compositions and methods as described by way of example as set forth below.
[0010] The principal object of the present invention is to enhance the accuracy and precision of brain tumor classification and localization using MRI imaging and advanced deep learning techniques, thereby facilitating early and accurate diagnosis.
[0011] Another object of the invention is to develop a fully automated system capable of efficiently processing MRI images, extracting relevant features, and classifying brain tumors into benign and malignant categories without the need for manual intervention.
[0012] Another object of the invention is to optimize the utilization of transfer learning strategies, particularly leveraging pre-trained convolutional neural network (CNN) models, to expedite the training process and improve the performance of the classification system, especially in scenarios with limited available data.
[0013] Another object of the invention is to streamline the workflow of medical professionals by providing a reliable and user-friendly tool for brain tumor diagnosis, enabling timely treatment planning and intervention to improve patient outcomes.
[0014] In view of the foregoing, the present invention provides a system designed for the precise classification of brain tumors utilizing AI-based transfer learning methodologies. Central to the system is a magnetic resonance imaging (MRI) module, which is tasked with generating detailed brain images delineating non-tumor and tumor areas, as well as accurately determining the shape and location of tumors within the brain. Accompanying this module is a pre-processing component, responsible for filtering noise from the images and enhancing image contrast through contrast limited adaptive histogram equalization techniques. Subsequently, a feature extraction module extracts spatial grey level dependency matrix features from the pre-processed images, facilitating nuanced analysis. The dataset is then partitioned by a data division module into a training dataset and a test dataset, with the latter comprising remaining images for evaluation purposes. An image augmentation module manipulates and generates altered versions of images within the training dataset to enhance classifier robustness. Leveraging a deep learning-based feature extraction module, features are extracted from the training dataset using a pre-trained convolutional neural network (CNN) model. Finally, a classification module employs these extracted features and transfer learning techniques to automatically classify brain tumor images into benign and malignant categories, providing clinicians with a reliable tool for accurate diagnosis and treatment planning.
[0015] In another aspect, the pre-processing module filters noise from the images using Laplacian of Gaussian filtering method.
[0016] In another aspect, the MRI module further comprises a segmentation mod-ule configured to segment the tumor from the brain images.
[0017] In another aspect, the pre-trained CNN model is selected from GoogLeNet, ResNet50, ResNet100, AlexNet, and SqueeZeNet.
[0018] In another aspect, the pre-processing module further comprises a reshaping module configured to reshape the images after noise removal.
[0019] In another aspect, the data division module further comprises a validation dataset comprising a portion of the training dataset for model evaluation.
[0020] Additional features of the invention will be or will become apparent to one with skill in the art upon examination of the following figures and detailed descrip-tion. It is intended that all such additional features and advantages be included with-in this description, be within the scope of the invention, and be protected by the ac-companying claims.

BRIEF DESCRIPTION OF THE DRAWINGS
[0021] Having thus described the subject matter of the present invention in general terms,
reference will now be made to the accompanying drawings, which are not necessarily drawn to
scale, and wherein:
[0022] Figure 1 Shows a block diagram of fundamental thought of transfer learning, in accordance with an embodiment of the present invention;
[0023] Figure 2 shows a method flow diagram of a system for classifying brain tumors using AI-based transfer learning, in accordance with an embodiment of the present invention;
[0024] Figure 3 shows a block diagram of the system for classifying brain tumors using AI-based transfer learning, in accordance with an embodiment of the present invention;
[0025] Those skilled in the art will appreciate that elements in the drawings are illustrated for simplicity and may not have necessarily been drawn to scale. Furthermore, the drawings may show only those specific details that are pertinent to understanding the embodiments of the present invention so as not to obscure the drawings with details that will be readily apparent to those of ordinary skill in the art having the benefit of the description herein.

DETAILED DESCRIPTION OF THE INVENTION
[0026] The subject matter of the present invention now will be described more fully hereinafter with reference to the accompanying drawings, in which some, but not all embodiments of the subject matter of the present invention are shown. Like numbers refer to like elements throughout. The subject matter of the present invention may be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will satisfy applicable legal requirements. Indeed, many modifications and other embodiments of the subject matter of the present invention set forth herein will come to mind to one skilled in the art to which the subject matter of the present invention pertains having the benefit of the teachings presented in the foregoing descriptions and the associated drawings. All illustrations of the drawings are for the purpose of describing selected versions of the present invention and are not intended to limit the scope of the present invention. Therefore, it is to be understood that the subject matter of the present invention is not to be limited to the specific embodiments disclosed and that modifications and other embodiments are intended to be included within the scope of the appended claims.
[0027] As a preliminary matter, it will readily be understood by one having ordinary skill in the relevant art that the present disclosure has broad utility and application. As should be understood, any embodiment may incorporate only one or a plurality of the above-disclosed aspects of the disclosure and may further incorporate only one or a plurality of the above-disclosed features. Furthermore, any embodiment discussed and identified as being “preferred” is considered to be part of a best mode contemplated for carrying out the embodiments of the present disclosure. Other embodiments also may be discussed for additional illustrative purposes in providing a full and enabling disclosure. Moreover, many embodiments, such as adaptations, variations, modifications, and equivalent arrangements, will be implicitly disclosed by the embodiments described herein and fall within the scope of the present disclosure.
[0028] Accordingly, while embodiments are described herein in detail in relation to one or more embodiments, it is to be understood that this disclosure is illustrative and example of the present disclosure and are made merely for the purposes of providing a full and enabling disclosure. The detailed disclosure herein of one or more embodiments is not intended, nor is to be construed, to limit the scope of patent protection afforded in any claim of a patent issuing here from, which scope is to be defined by the claims and the equivalents thereof. It is not intended that the scope of patent protection be defined by reading into any claim a limitation found herein that does not explicitly appear in the claim itself.
[0029] Thus, for example, any sequence(s) and/or temporal order of steps of various processes or methods that are described herein are illustrative and not restrictive. Accordingly, it should be understood that, although steps of various processes or methods may be shown and described as being in a sequence or temporal order, the steps of any such processes or methods are not limited to being carried out in any particular sequence or order, absent an indication otherwise. Indeed, the steps in such processes or methods generally may be carried out in various different sequences and orders while still falling within the scope of the present invention. Accordingly, it is intended that the scope of patent protection is to be defined by the issued claim(s) rather than the description set forth herein.
[0030] Additionally, it is important to note that each term used herein refers to that which an ordinary artisan would understand such term to mean based on the contextual use of such term herein. To the extent that the meaning of a term used herein - as understood by the ordinary artisan based on the contextual use of such term - differs in any way from any particular dictionary definition of such term, it is intended that the meaning of the term as understood by the ordinary artisan should prevail.
[0031] Furthermore, it is important to note that, as used herein, “a” and “an” each generally denotes “at least one”, but does not exclude a plurality unless the contextual use dictates otherwise. When used herein to join a list of items, “or” denotes “at least one of the items”, but does not exclude a plurality of items of the list. Finally, when used herein to join a list of items, “and” denotes “all of the items of the list”.
[0032] The present disclosure provides a system aimed at improving the classification of brain tumors through the integration of AI-based transfer learning methodologies. At its core, the system utilizes magnetic resonance imaging (MRI) technology to generate detailed brain images highlighting both non-tumor and tumor areas. By leveraging advanced image processing techniques, such as contrast limited adaptive histogram equalization, the system enhances image quality by filtering noise and improving contrast, thus providing clearer and more informative images for analysis.
[0033] To further refine the analysis process, the system employs a feature extraction module, which extracts spatial grey level dependency matrix features from the pre-processed images. These features serve as crucial indicators for identifying and characterizing different types of brain tumors. Additionally, the system incorporates a data division module to partition the dataset into a training dataset and a test dataset, facilitating robust model training and evaluation.
[0034] Key innovation of the system lies in its utilization of transfer learning techniques, particularly through a deep learning-based feature extraction module. By leveraging pre-trained convolutional neural network (CNN) models, the system is able to extract meaningful features from the training dataset efficiently and effectively. These features are then utilized by a classification module, which employs transfer learning techniques to automatically classify brain tumor images into benign and malignant categories. This automated classification process not only enhances accuracy but also expedites the diagnosis and treatment planning process for medical professionals.
[0035] In an embodiment, the system represents an advanced approach to classifying brain tumors using AI-based transfer learning methodologies. At its core, the system employs a magnetic resonance imaging (MRI) module, which is tasked with generating detailed brain images that highlight both non-tumor and tumor areas, while also determining the shape and precise location of the tumor within the brain. This module serves as the foundation for subsequent analysis and classification processes.
[0036] To prepare the MRI images for analysis, a pre-processing module is utilized. This module is specifically designed to filter out noise from the images and enhance image contrast using contrast limited adaptive histogram equalization techniques. By improving the quality and clarity of the images, this pre-processing step ensures that subsequent analysis steps are performed on reliable and informative data.
[0037] Following pre-processing, a feature extraction module is employed to extract spatial grey level dependency matrix features from the pre-processed images. These features capture important spatial relationships and patterns within the images, providing valuable information for tumor classification. Additionally, a data division module is utilized to partition the dataset into a training dataset and a test dataset, facilitating model training and evaluation.
[0038] To further enhance the robustness of the classification model, an image augmentation module is incorporated into the system. This module manipulates and generates altered versions of images within the training dataset, thereby increasing the diversity of the training data and improving the model's ability to generalize to unseen data.
[0039] Further, with respect to the classification process, there is provided a deep learning-based feature extraction module, which extracts features from the training dataset using a pre-trained convolutional neural network (CNN) model. Leveraging the knowledge encoded in the pre-trained CNN model, this module efficiently extracts relevant features from the MRI images, enabling accurate and effective tumor classification. Finally, a classification module utilizes the extracted features and transfer learning techniques to automatically classify brain tumor images into benign and malignant categories. This automated classification process streamlines the diagnostic workflow, enabling timely and accurate diagnosis and treatment planning for patients.
[0040] In an embodiment, within the MRI module, a segmentation module plays a crucial role in the classification process by accurately delineating and isolating the tumor from the surrounding brain tissue within the images. This segmentation process involves identifying and defining the boundaries of the tumor region, enabling precise analysis and characterization. By effectively separating the tumor from the rest of the brain images, the segmentation module enhances the system's ability to extract relevant features and classify the tumor accurately. This segmentation step is essential for facilitating subsequent analysis and ensuring that the classification model can focus specifically on the tumor region, thereby improving diagnostic accuracy and aiding in treatment planning for patients with brain tumors.
[0041] In accordance with an embodiment of the present invention, Figure 1 shows a block diagram of fundamental thought of transfer learning. This figure shows the core concept of transfer learning, illustrating a sequential flow from the source to the target domain. Initially, the source data, comprising labeled examples from a related task or domain, is utilized to train a source model. Concurrently, corresponding source labels are employed to guide the learning process. Subsequently, the knowledge acquired by the source model is transferred to the target domain, involving the adaptation of the model's learned features and parameters to accommodate the target data and labels. This transfer of knowledge enables the refinement of a target model, tailored to the specific characteristics and nuances of the target domain. Ultimately, the target model, along with the newly associated target labels and data, embodies the culmination of the transfer learning process, poised to address the unique challenges and requirements of the target task or domain.
[0042] In accordance with an embodiment of the present invention, Figure 2 shows a method flow diagram of a system for classifying brain tumors using AI-based transfer learning. This figure illustrates the sequential steps involved in the system for classifying brain tumors using AI-based transfer learning. It commences with sample MRI images, which undergo preprocessing to enhance data quality. This preprocessing includes noise removal utilizing the Laplacian of Gaussian Filtering method, resulting in smoothed MRI images. Subsequently, image enhancement is performed using contrast limited adaptive histogram equalization to improve visualization. Feature extraction is then executed utilizing spatial grey level dependency matrix, followed by feature selection to identify pertinent characteristics. The extracted features are utilized for image classification employing a CNN-based classifier, which distinguishes between normal and abnormal brain images. Abnormal images undergo tumor segmentation and classification via transfer learning techniques, ultimately culminating in the classification of tumors as benign or malignant. This comprehensive method flow ensures the accurate and efficient classification of brain tumors, leveraging advanced AI techniques and transfer learning methodologies.
[0043] In accordance with an embodiment of the present invention, Figure 3 shows a block diagram of the system for classifying brain tumors using AI-based transfer learning. The block diagram shows a Magnetic Resonance Imaging (MRI) module 302, which generates brain images highlighting both tumor and non-tumor areas while determining the tumor's shape and location within the brain. Subsequently, the images undergo preprocessing by a pre-processing module 304, wherein noise is filtered out and image contrast is enhanced via contrast limited adaptive histogram equalization. The feature extraction module 306 then identifies spatial grey level dependency matrix features from the preprocessed images, facilitating the extraction of relevant characteristics for tumor classification. The dataset is divided into training and test datasets by the data division module 308, ensuring robust model training and evaluation. The image augmentation module 310 manipulates training images to enhance model robustness, while the deep learning-based feature extraction module 312 leverages a pre-trained convolutional neural network (CNN) model to extract pertinent features. Finally, the classification module 314 employs transfer learning techniques to classify tumors based on extracted features, automatically categorizing brain tumor images into benign and malignant categories. Through this cohesive integration of modules, the system streamlines the process of brain tumor classification, offering accurate and efficient diagnosis crucial for patient treatment and care.
[0044] In an embodiment, the selection of the pre-trained convolutional neural network (CNN) model from a pool including GoogLeNet, ResNet50, ResNet100, AlexNet, and SqueeZeNet is a critical step in the process, as it determines the framework upon which the transfer learning process will be based. Each of these models offers unique architectures and capabilities, allowing for the extraction of different features and patterns from the input data. By carefully selecting the most suitable pre-trained CNN model, the system can effectively leverage the knowledge and insights embedded within the model to expedite the learning process and enhance the accuracy of tumor classification. This selection process ensures that the system can adapt to the intricacies of brain tumor images and extract relevant features efficiently, ultimately improving the diagnostic capabilities and clinical outcomes of the system.
[0045] In an embodiment, incorporated within the pre-processing module, the reshaping component serves a pivotal role by optimizing the structure of the images subsequent to noise removal. Following the application of noise reduction techniques, such as Laplacian of Gaussian filtering, the reshaping module adjusts the dimensions or aspect ratios of the images, ensuring uniformity and compatibility with subsequent stages of analysis. This reshaping process is crucial for standardizing the image inputs, facilitating consistent and reliable feature extraction and classification. By preparing the images in a standardized format, the reshaping module enhances the effectiveness of subsequent processing steps, ultimately contributing to improved accuracy and reliability in the classification of brain tumors from MRI images.
[0046] Further, embedded within the data division module, the validation dataset plays a pivotal role in assessing the performance and generalization ability of the trained model. Derived from a portion of the training dataset, the validation dataset serves as an independent subset used for model evaluation during the training process. By withholding a portion of the training data for validation, the system can gauge the model's performance on unseen data, thereby detecting issues such as overfitting or underfitting. This validation dataset enables iterative refinement of the model's parameters and architecture, ensuring that it achieves optimal performance on new, unseen data. Overall, the inclusion of the validation dataset within the data division module enhances the robustness and reliability of the classification model, ultimately improving its effectiveness in accurately identifying and categorizing brain tumors from MRI images.
[0047] Further, in an embodiment, within the image augmentation module, diverse techniques are implemented to introduce variations in coloring and lighting conditions, thereby fortifying the classifier's resilience to different imaging scenarios. By systematically altering the hue, saturation, brightness, and contrast levels of the images, the module simulates a wide spectrum of real-world conditions encountered during image acquisition. Additionally, techniques such as random rotations, translations, and flips may be employed to introduce spatial variations, further diversifying the dataset. These augmentations serve to mitigate the risk of model overfitting by exposing it to a more comprehensive range of image variations during training. As a result, the classifier becomes adept at recognizing and distinguishing tumor features across a multitude of imaging conditions, enhancing its robustness and generalization capabilities. In essence, the image augmentation techniques integrated within the module bolster the classifier's performance and reliability in accurately classifying brain tumors from MRI images, even in the presence of varying coloring and lighting conditions.
[0048] In an embodiment, the classification module lies the utilization of a softmax layer, a pivotal component tasked with the final categorization of tumor images into benign and malignant classifications. This layer operates by computing the probabilities associated with each class, enabling the model to assign a likelihood score to every potential outcome. Through the application of sophisticated mathematical operations, such as the softmax function, the layer normalizes these scores, resulting in a probability distribution across all classes. Subsequently, the class with the highest probability is deemed the most likely classification for the given tumor image. By employing the softmax layer, the classification module delivers a decisive and accurate determination of whether a brain tumor is benign or malignant, thereby empowering clinicians with vital diagnostic insights crucial for informed treatment decisions and patient care.
[0049] In an embodiment, the deep learning-based feature extraction module undergoes rigorous training to mitigate the risks of both overfitting and underfitting when applied to the test dataset. Overfitting occurs when the model excessively learns the intricacies of the training data, resulting in reduced generalization performance on unseen data, while underfitting signifies a failure of the model to capture the underlying patterns within the dataset adequately. To counter these challenges, the module is trained using techniques such as regularization, dropout, and cross-validation, which help prevent the model from memorizing noise and instead focus on learning meaningful features. By fine-tuning the model's parameters and architecture iteratively, the module achieves optimal performance, striking a balance between complexity and generalization ability. Ultimately, this approach ensures that the deep learning-based feature extraction module effectively captures pertinent features from the data, facilitating robust and accurate classification of brain tumors while maintaining optimal performance on unseen test data.
[0050] In an embodiment, transfer Learning (TL) stands as a pivotal technique within the context of the present invention, offering a powerful approach to enhancing the efficiency and accuracy of brain tumor classification. TL involves leveraging the knowledge encoded within a pre-trained convolutional neural network (CNN) model, originally developed for a different application, and adapting it to the task of brain tumor classification. This approach is notably advantageous, as it allows for the rapid acquisition of new features from the dataset, significantly expediting the model development process compared to starting from scratch.
[0051] By utilizing TL on a pre-trained CNN model, the system can efficiently learn and adapt to the unique features present in MRI images of brain tumors. This enables the system to quickly identify relevant patterns and features within the data, facilitating more accurate tumor classification. Moreover, TL enables the system to overcome limitations associated with traditional learning approaches, which often require substantial time and resources to train independent models from the ground up.
[0052] One of the key benefits of TL is its ability to leverage previously learned features and weights from the pre-trained model, even when faced with limited data for the new task. This ensures that valuable information from past training sessions is effectively utilized, enhancing the system's ability to generalize and make accurate predictions on new data. Unlike traditional learning approaches, which often require starting anew for each task, TL allows for the seamless transfer of knowledge between related tasks, resulting in faster and more precise model development.
[0053] Transfer Learning (TL) stands out as a transformative approach, diverging from conventional methods by enabling the retention and utilization of valuable data, features, and weights from past trained models. Unlike traditional machine learning approaches where data is not preserved between tasks, TL allows for the seamless transfer of knowledge from one task to another. This means that features and weights learned from previous models can be repurposed and applied to new assignments, even in scenarios with limited data availability. By incorporating knowledge from past training sessions, TL empowers the system to adapt more efficiently to new tasks, accelerating the learning process and enhancing its performance in addressing challenges such as data scarcity or imbalance. Overall, TL facilitates a more agile and effective learning paradigm, enabling the system to capitalize on existing knowledge and insights to achieve superior performance in various applications, including brain tumor classification using MRI imaging.
[0054] In an embodiment, in the validation process of the methodology, the invention employs five distinct and well-established convolutional neural network (CNN) models: GoogLeNet, ResNet50, ResNet100, AlexNet, and SqueeZeNet. These models are chosen for their diverse architectures and proven effectiveness in image recognition tasks. Leveraging magnetic resonance imaging (MRI) images of brain tumors, the invention applies transfer learning strategies on the provided dataset.
[0055] Transfer learning involves utilizing pre-trained models, originally developed for unrelated tasks, and fine-tuning them for the specific task at hand. In this case, the pre-trained CNN models are repurposed to extract meaningful features from MRI images of brain tumors. By leveraging the knowledge and insights encoded within these pre-trained models, the system can expedite the learning process and achieve superior performance, even when faced with limited training data.
[0056] The transfer learning process involves retraining the pre-trained CNN models on the provided dataset, allowing them to adapt to the unique characteristics of brain tumor images. This process effectively removes outwardly recognizable and fundamental features, enhancing the model's ability to capture subtle and nuanced patterns indicative of different tumor types. Finally, the classification of these features is carried out using a softmax layer, which assigns probabilities to each class, enabling the system to categorize brain tumor images into benign and malignant categories with high accuracy. Overall, the utilization of transfer learning in conjunction with established CNN models represents a powerful and effective approach to brain tumor classification, offering improved accuracy and efficiency in diagnosis and treatment planning.
[0057] The process initiates with a dataset comprising MRI images meticulously collected and categorized into benign and malignant classes. This dataset undergoes a systematic procedure consisting of several key stages: preprocessing, segmentation, and feature extraction, followed by deep learning-based extraction of features, culminating in tumor classification. During preprocessing, noise is filtered and image contrast is enhanced to optimize data quality. Subsequently, segmentation techniques are applied to delineate tumor regions within the images, facilitating precise analysis. Feature extraction is then performed to capture salient characteristics from the segmented regions, leveraging deep learning methodologies to extract complex patterns and features relevant to tumor classification. Finally, utilizing the extracted features, the system executes tumor classification, effectively discerning between benign and malignant tumors. This comprehensive process ensures the robust and accurate classification of brain tumors, thereby facilitating timely diagnosis and treatment planning for improved patient outcomes.
[0058] In an embodiment, the process for classifying brain tumors begins with a dataset comprising MR brain images containing both benign and malignant MR slices. The second step involves extracting images based on distinct features to identify the shape and position of tumors within the MRI images. Following this, preprocessing techniques are applied to enhance data quality by filtering out noise using methods such as Laplacian of Gaussian filtering and reshaping the images. Additionally, image labels are loaded to facilitate machine recognition, and contrast limited adaptive histogram equalization is utilized for image enhancement.
[0059] Subsequently, feature extraction is performed using spatial grey level dependency matrix, capturing relevant characteristics from the preprocessed images. Data division is then conducted to partition the dataset into training and test sets, typically with the training set comprising 70-80% of the images. The model is trained on the training dataset, and its performance is evaluated on the test dataset to ensure accuracy and generalization.
[0060] To enhance the robustness of the model, image augmentation is employed, generating altered versions of images in the training dataset to account for variations in coloring and lighting conditions. A deep-trained model is then applied to the training dataset, leveraging deep learning techniques for feature extraction. If the model demonstrates high accuracy without overfitting or underfitting, it is utilized for classification.
[0061] Upon classification, the trained model effectively separates brain tumors, yielding normal and abnormal images. Abnormal images are further segmented or classified using transfer learning techniques, allowing for the accurate identification of tumor types. Finally, the classification results provide a clear delineation between benign and malignant tumors, facilitating timely diagnosis and treatment planning for improved patient outcomes. Overall, this comprehensive process ensures the accurate and efficient classification of brain tumors, leveraging advanced imaging and deep learning methodologies.

[0062] Some of the non-limiting advantages of the present invention are:
? Enhanced Accuracy: The system improves the accuracy of brain tumor classification by leveraging advanced AI-based transfer learning methodologies, resulting in more precise and reliable diagnostic outcomes compared to traditional methods.
? Automated Processing: With its fully automated workflow, the system streamlines the process of brain tumor classification, reducing the need for manual intervention and enabling faster diagnosis and treatment planning.
? Improved Efficiency: By efficiently processing MRI images and extracting relevant features, the system enhances the efficiency of the diagnostic process, enabling medical professionals to make informed decisions more quickly and effectively.
? Transfer Learning Benefits: Leveraging transfer learning techniques, the system harnesses the knowledge from pre-trained convolutional neural network (CNN) models, allowing for faster model training and improved classification accuracy, especially in scenarios with limited data.
? Customizable and Scalable: The modular design of the system allows for easy customization and scalability to accommodate varying data and clinical requirements. This flexibility ensures that the system can be adapted to suit different healthcare settings and patient populations, maximizing its utility and impact.

[0063] Terms and phrases used in this document, and variations thereof, unless otherwise expressly stated, should be construed as open-ended as opposed to limit-ing. As examples of the foregoing: the term “including” should be read as mean “in-cluding, without limitation” or the like; the term “example” is used to provide ex-emplary instances of the item in discussion, not an exhaustive or limiting list there-of; and adjectives such as “conventional,” “traditional,” “standard,” “known” and terms of similar meaning should not be construed as limiting the item described to a given time period or to an item available as of a given time, but instead should be read to encompass conventional, traditional, normal, or standard technologies that may be available or known now or at any time in the future. Likewise, a group of items linked with the conjunction “and” should not be read as requiring that each and every one of those items be present in the grouping, but rather should be read as “and/or” unless expressly stated otherwise. Similarly, a group of items linked with the conjunction “or” should not be read as requiring mutual exclusivity among that group, but rather should also be read as “and/or” unless expressly stated otherwise. Furthermore, although item, elements or components of the disclosure may be de-scribed or claimed in the singular, the plural is contemplated to be within the scope thereof unless limitation to the singular is explicitly stated. The presence of broad-ening words and phrases such as “one or more,” “at least,” “but not limited to” or other like phrases in some instances shall not be read to mean that the narrower case is intended or required in instances where such broadening phrases may be absent.
[0064] For the purposes of this specification and appended claims, unless otherwise indicated, all numbers expressing amounts, sizes, dimensions, proportions, shapes, formulations, parameters, percentages, quantities, characteristics, and other numerical values used in the specification and claims, are to be understood as being modified in all instances by the term “about” even though the term “about” may not expressly appear with the value, amount, or range. Accordingly, unless indicated to the contrary, the numerical parameters set forth in the following specification and attached claims are not and need not be exact, but may be approximate and/or larger or smaller as desired, reflecting tolerances, conversion factors, rounding off, measurement error and the like, and other factors known to those of skill in the art depending on the desired properties sought to be obtained by the subject matter of the present invention. For example, the term “about,” when referring to a value can be meant to encompass variations of, in some embodiments ± 100%, in some embodiments ± 50%, in some embodiments ± 20%, in some embodiments ± 10%, in some embodiments ± 5%, in some embodiments ± 1%, in some embodiments ± 0.5%, and in some embodiments ± 0.1% from the specified amount, as such variations are appropriate to perform the disclosed methods or employ the disclosed compositions.
[0065] Further, the term “about” when used in connection with one or more numbers or numerical ranges, should be understood to refer to all such numbers, including all numbers in a range and modifies that range by extending the boundaries above and below the numerical values set forth. The recitation of numerical ranges by endpoints includes all numbers, e.g., whole integers, including fractions thereof, subsumed within that range (for example, the recitation of 1 to 5 includes 1, 2, 3, 4, and 5, as well as fractions thereof, e.g., 1.5, 2.25, 3.75, 4.1, and the like) and any range within that range.
[0066] All publications, patent applications, patents, and other references mentioned in the specification are indicative of the level of those skilled in the art to which the presently disclosed subject matter pertains. All publications, patent applications, patents, and other references are herein incorporated by reference to the same extent as if each individual publication, patent application, patent, and other reference was specifically and individually indicated to be incorporated by reference. It will be understood that, although a number of patent applications, patents, and other references are referred to herein, such reference does not constitute an admission that any of these documents forms part of the common general knowledge in the art. Although the foregoing subject matter has been described in some detail by way of illustration and example for purposes of clarity of understanding, it will be understood by those skilled in the art that certain changes and modifications can be practiced within the scope of the appended claims.

, Claims:I/We Claim:

1. A system (300) for classifying brain tumors using AI-based transfer learning, comprising:
a magnetic resonance imaging (MRI) module (302) configured to generate brain images showing non-tumor and tumor areas and to determine the shape and location of the tumor in the brain;
a pre-processing module (304) configured to filter noise from the images and enhance image contrast using contrast limited adaptive histogram equalization;
a feature extraction module (306) configured to extract spatial grey level dependency matrix features from the pre-processed images;
a data division module (308) configured to divide the dataset into a training dataset and a test dataset comprising the remaining images;
an image augmentation module (310) configured to manipulate and generate altered versions of images in the training dataset;
a deep learning-based feature extraction module (312) configured to extract features from the training dataset using a pre-trained convolutional neural network (CNN) model;
a classification module (314) configured to classify the tumor based on the extracted features and transfer learning techniques, wherein the system automatically separates brain tumor images into benign and malignant categories.
2. The system as claimed in claim 1, wherein the pre-processing module (304) filters noise from the images using Laplacian of Gaussian filtering method.
3. The system as claimed in claim 1, wherein the MRI module (302) further comprises a segmentation module configured to segment the tumor from the brain images.
4. The system as claimed in claim 1 or 2, wherein the pre-trained CNN model is selected from GoogLeNet, ResNet50, ResNet100, AlexNet, and SqueeZeNet.

5. The system as claimed in any one of claims 1 to 3, wherein the pre-processing module (306) further comprises a reshaping module configured to re-shape the images after noise removal.
6. The system as claimed any one of claims 1 to 4, wherein the data division module (308) further comprises a validation dataset comprising a portion of the training dataset for model evaluation.
7. The system as claimed in any one of claims 1 to 5, wherein the image aug-mentation module (310) further comprises techniques for varying coloring and lighting conditions to increase classifier robustness.
8. The system as claimed in any one of claims 1 to 6, wherein the classification module (314) employs a softmax layer for final classification of tumor images into benign and malignant categories.
9. The system as claimed in any one of claims 1 to 7, wherein the deep learn-ing-based feature extraction module (312) is trained to avoid overfitting and under-fitting of the model on the test dataset.
10. The system as claimed in any one of claims 1 to 8, further comprising a dis-play module configured to present the classified tumor images to a user, aiding in medical diagnosis and treatment planning.
11. The system of any one of the claims 1 to 9, wherein the classification mod-ule (314) provides output indicative of the exact position of the tumor within the brain.