Description:ARTIFICIAL INTELLIGENCE (AI)-BASED SYSTEM FOR ANALYSIS OF THE MEDICAL IMAGES
TECHNICAL FIELD
[0001] The present disclosure relates to medical informatics, specifically an AI-based system for analyzing colorectal images to identify cancer indicators through image processing and machine learning technologies.
BACKGROUND
[0002] Colorectal cancer (CRC) remains a critical public health issue, characterized by its high incidence and mortality rates globally. The burden of CRC is exacerbated by demographic shifts towards older populations and lifestyle changes associated with increased risk factors, making its impact on society and healthcare systems profound and far-reaching.
[0003] The established protocol for CRC screening and diagnosis is colonoscopy, a procedure that has been the cornerstone of efforts to identify and prevent the progression of colorectal cancer for decades. Colonoscopy allows for the direct visualization of the colon and rectum's inner lining, permitting the detection and removal of premalignant polyps and biopsy of uspicious lesions. Positive findings are then typically subjected to histopathological examination for a definitive diagnosis. While effective, the procedure has several drawbacks that limit its suitability as a widespread screening tool.
[0004] The invasive nature of colonoscopy often deters individuals from participating in routine screenings. Colonoscopy generally requires sedation or anesthesia, which introduces additional medical risks and necessitates recovery time, thereby incurring further indirect costs to the patient. The procedure can be uncomfortable and is associated with post-procedure discomfort and complications such as bleeding and perforation, albeit infrequently.
[0005] In terms of resource allocation, colonoscopy is resource-intensive, requiring specialized equipment and trained personnel. It is costly, contributing to the limited use in lower-resource settings and among populations with limited healthcare access. The procedure's cost-effectiveness is also a matter of ongoing debate, particularly when considering the need for regular screenings over a population's lifespan.
[0006] The skill-dependent nature of colonoscopy further complicates matters. The effectiveness of the procedure is contingent upon the experience and proficiency of the endoscopist. The reliance on manual expertise introduces variability in diagnostic accuracy, with the potential for human error leading to missed lesions or misinterpretation of findings.
[0007] From a logistical standpoint, there are significant challenges in providing colonoscopy services on a mass scale. The need for specialized facilities and trained gastroenterologists means that in many parts of the world, particularly in rural and underserved areas, access to colonoscopy is severely limited. The situation is dire in low- and middle-income countries where the healthcare infrastructure may not support such specialized services at all, let alone at the scale needed for population-wide screening.
[0008] The current approach to CRC screening with colonoscopy is further challenged by its one-size-fits-all methodology. The variability in individual risk factors, such as family history, genetics, and lifestyle choices, is not adequately addressed in a standard screening protocol, leading to over-screening or under-screening of various population segments.
[0009] Moreover, the psychological barrier to invasive diagnostic procedures cannot be understated. The fear and anxiety associated with colonoscopies can lead to avoidance behavior, which in turn results in delayed or missed diagnoses. Public perception and cultural attitudes toward invasive medical procedures also play a role in the underutilization of colonoscopies as a screening tool.
[0010] The combination of the factors - the invasiveness and risks of the procedure, the reliance on manual expertise, the high costs, and the lack of infrastructure to support widespread screenings - presents significant barriers to the effective management and prevention of CRC on a population level.
[0011] Given the limitations of colonoscopy and the escalating burden of colorectal cancer, there is an unmet need for innovative approaches that are less invasive, more cost-effective, and suitable for widespread implementation. The approaches should ideally provide accurate screening and diagnostic capabilities that are adaptable to various settings and population needs, reducing the barriers to early detection and treatment of CRC and ultimately improving outcomes.
[0012] In summary, while colonoscopy remains a valuable tool in the fight against CRC, its limitations underscore the necessity for new diagnostic methodologies. The development of alternative strategies is critical to address the diverse and growing needs of global populations in the context of CRC screening and diagnosis. Such advancements could lead to broader and more equitable access to life-saving diagnostics and a subsequent reduction in the incidence and mortality of colorectal cancer.

SUMMARY
[0013] The aim of the present disclosure is to provide an artificial intelligence (AI)-based system for analyzing medical images to enable detection of colorectal cancer.
[0014] The disclosure describes an artificial intelligence (AI)-based system tailored for the analysis of colorectal medical images. The architecture of system is dual-component, consisting of a computing device for obtaining patient images from a medical imaging device, and a server configuration interfacing through a network. The server comprises a non-transitory storage device which maintains a repository of colorectal images, each tagged with medical observation data. Central to the operation of system is a microprocessor tasked with implementing a series of executable routines. The routines are designed to first secure and denoise the incoming images, then isolate the regions of interest (ROI) within them. The core analytical engine of the system is a deep learning model, trained through the analysis of numerous reference images stored within the system. Once operational, the deep learning model scrutinizes the ROIs to identify features indicative of colorectal cancer. The outcome of the analysis is a calculated likelihood score and a set of summarized findings, which are then conveyed back to the computing device. The process aims to bolster the accuracy and efficiency of colorectal cancer detection, offering a leap forward in diagnostic methodologies.
[0015] In an embodiment, the medical observation data tagged with each colorectal reference image in the colorectal image database is selected from historical diagnostic data, patient outcomes, and annotations by medical professionals.
[0016] In an embodiment, the microprocessor utilizes one or more filters selected from the group consisting of Gaussian blur, median filtering, non-local means denoising, Block-matching and 3D filtering algorithm (BM3D) and machine learning-based denoising techniques to remove noise.
[0017] In an embodiment, the microprocessor utilizes at least one technique selected from an edge detection, thresholding, and one or more morphological operation to extract ROI from each denoised colorectal image.
[0018] In an embodiment, the display of the generated likelihood score and summarized findings is customizable based on a user preference.
[0019] In an embodiment, the server arrangement receives and incorporate additional relevant patient data selected from a genetic information, one or more lifestyle events, a demographic data, and a previous medical history.
[0020] In an embodiment, the summarized findings comprise at least one from; quantitative measurements of detected features, comparative analysis with similar cases from the colorectal image database, and the recommendations for the add-on diagnostic procedure.
[0021] In an embodiment, the server arrangement generates a report that includes a summary of detected one or more indicative features, likelihood score, and suggested next steps for treatment.
[0022] The disclosure presents a method utilizing an AI-based system for the enhanced analysis of colorectal medical images. A computing device receives patient-specific colorectal images from an imaging device and interfaces with a server arrangement through a network. The server contains a non-transitory storage device programmed with executable routines and maintains a database of colorectal reference images, each annotated with medical observations. A microprocessor executes the routines to synchronize and process both received and stored images. The process involves a noise reduction step yielding denoised images, followed by the extraction of regions of interest (ROIs) for closer examination. Utilizing the refined images, the system constructs a deep learning model, informed by the analysis of the reference image database. The deep learning model applies the learned patterns to the ROIs to identify features indicative of colorectal cancer. From the analysis, the system generates a likelihood score and a set of summarized findings that reflect the presence of cancer. The outputs are then presented on the computing device, offering an AI-enhanced tool for early diagnosis and potentially improving patient outcomes through timely cancer detection.
BRIEF DESCRIPTION OF THE DRAWINGS
[0023] The summary above, as well as the following detailed description of illustrative embodiments, is better understood when read in conjunction with the appended drawings. For the purpose of illustrating the present disclosure, exemplary constructions of the disclosure are shown in the drawings. However, the present disclosure is not limited to specific methods and instrumentalities disclosed herein.
[0024] Embodiments of the present disclosure will now be described, by way of example only, with reference to the following diagrams.
[0025] FIG. 1 illustrates an artificial intelligence (AI)-based system for analysis of the medical images in accordance with various implementations of the present disclosure;
[0026] FIG. 2 illustrates an exemplary colorectal image database in a tabular format, in accordance with the embodiments of the present disclosure;
[0027] FIG. 3 illustrates an exemplary tabular representation of likelihood score and summarized findings of multiple patients;
[0028] FIG. 4 illustrates a method for analyzing medical images using an artificial intelligence (AI)-based system, in accordance with the embodiments of the present disclosure; and
[0029] FIG. 5 depicts a systematic diagram illustrating the operational flow of an AI-based medical image analysis system, in accordance with the embodiments of the present disclosure.
DETAILED DESCRIPTION OF EMBODIMENTS
[0030] Various objects, features, aspects and advantages of the inventive subject matter will become more apparent from the following detailed description of preferred embodiments, along with the accompanying drawing figures in which like numerals represent like components.
[0031] The following discussion provides many example embodiments of the inventive subject matter. Although each embodiment represents a single combination of inventive elements, the inventive subject matter is considered to include all possible combinations of the disclosed elements. Thus, if one embodiment comprises elements A, B, and C, and a second embodiment comprises elements B and D, then the inventive subject matter is also considered to include other remaining combinations of A, B, C, or D, even if not explicitly disclosed.
[0032] In some embodiments, the numbers expressing quantities of ingredients, properties such as concentration, reaction conditions, and so forth, used to describe and claim certain embodiments of the invention are to be understood as being modified in some instances by the term “about.” Accordingly, in some embodiments, the numerical parameters set forth in the written description and attached claims are approximations that can vary depending upon the desired properties sought to be obtained by a particular embodiment. In some embodiments, the numerical parameters should be construed in light of the number of reported significant digits and by applying ordinary rounding techniques. Notwithstanding that the numerical ranges and parameters setting forth the broad scope of some embodiments of the invention are approximations, the numerical values set forth in the specific examples are reported as precisely as practicable. The numerical values presented in some embodiments of the invention may contain certain errors necessarily resulting from the standard deviation found in their respective testing measurements.
[0033] As used in the description herein and throughout the claims that follow, the meaning of “a,” “an,” and “the” includes plural reference unless the context clearly dictates otherwise. Also, as used in the description herein, the meaning of “in” includes “in” and “on” unless the context clearly dictates otherwise.
[0034] The recitation of ranges of values herein is merely intended to serve as a shorthand method of referring individually to each separate value falling within the range. Unless otherwise indicated herein, each individual value is incorporated into the specification as if it were individually recited herein. All methods described herein can be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The use of any and all examples, or exemplary language (e.g. “such as”) provided with respect to certain embodiments herein is intended merely to better illuminate the invention and does not pose a limitation on the scope of the invention otherwise claimed. No language in the specification should be construed as indicating any non-claimed element essential to the practice of the invention.
[0035] Groupings of alternative elements or embodiments of the invention disclosed herein are not to be construed as limitations. Each group member can be referred to and claimed individually or in any combination with other members of the group or other elements found herein. One or more members of a group can be included in, or deleted from, a group for reasons of convenience and/or patentability. When any such inclusion or deletion occurs, the specification is herein deemed to contain the group as modified thus fulfilling the written description of all Markush groups used in the appended claims.
[0036] FIG. 1 illustrates an artificial intelligence (AI)-based system 100 (interchangeably referred as system 100) for analysis of the medical images in accordance with various implementations of the present disclosure. The system 100 comprises a computing device 102, a server arrangement 104 and other known components of an AI based colorectal cancer detection platform/application.
[0037] In an embodiment, the computing device 102 receives the colorectal images from a medical imaging device, which can be selected from computed tomography (CT) scanners, magnetic resonance imaging (MRI) machines, and positron emission tomography (PET) scanners, etc. The primary and sole function of the computing device 102 is the reception of the high-resolution images, from the medical imaging device. The reception of colorectal images involves establishing a secure and efficient communication channel between the computing device 102 and the medical imaging device, ensuring a seamless and error-free transfer of colorectal images. The compatibility of the computing device 102 with the medical imaging device is achieved through a software and hardware interfaces, facilitating a versatile and reliable reception of colorectal images from different types of the medical imaging device.
[0038] In an embodiment, the operation within the computing device 102 is further enhanced by data handling protocols, ensuring the integrity and quality of the colorectal images during transmission. The data handling protocols are critical for preserving the diagnostic accuracy of the colorectal images and maintaining patient confidentiality, adhering to stringent medical data handling standards. The computing device 102 safeguards the patient data and ensures the fidelity of the colorectal images received.
[0039] Moreover, the functionality of the computing device 102 is not limited to mere reception of colorectal images; the computing device 102 also includes preliminary processing steps such as image validation and format standardization, preparing the colorectal images for subsequent AI-driven analysis. The approach to image reception and processing by the computing device 102 underscores vital role in the AI-based system 100, where accuracy and reliability are paramount.
[0040] In an embodiment, the server arrangement 104 is operatively coupled to the computing device 102, through the network interface. The server arrangement 104 comprises a non-transitory storage device that stores a set of executable routines. The stored set of executable routines are integral to the function of the AI-based system 100, facilitating the analysis of colorectal images, particularly focusing on colorectal cancer detection.
[0041] In another embodiment, the non-transitory storage device stores a colorectal image database that comprises the multiple colorectal reference images. Each colorectal reference image within the colorectal image database is uniquely tagged with medical observation data. The tagging of the images with medical observation data is a critical feature, as the tagging enables the AI-based system 100 to draw upon a repository of annotated medical information. The annotation plays an essential role in enhancing the accuracy and efficacy of the image analysis process.
[0042] In an embodiment, the medical observation data tagged with each colorectal reference image encompasses a diverse range of clinical information, which may include, but is not limited to, patient demographics, diagnostic details, treatment history, and outcomes. The incorporation of the medical observation data allows the AI-based system 100, through the execution of the stored routines in the non-transitory storage device, to conduct a thorough analysis of the colorectal images. The analysis enables the identification of pathological features and extends to the interpretation of the identified pathological features in the context of the patient-specific medical data.
[0043] FIG. 2 illustrates an exemplary colorectal image database in a tabular format, in accordance with the embodiments of the present disclosure. As illustrated, the table presents a dataset of colorectal reference images, each with an associated Image ID, and details pertinent to medical observation data. The columns include Image Date, Imaging Technique, Polyp Size, Histology, Lesion Location, Diagnosis, and Notes. The imaging techniques range from CT scans to MRIs, Ultrasounds, and PET scans, revealing polyps of varying sizes and histological classifications such as adenoma, carcinoma, and hyperplastic. Lesions are located in different colon segments. Diagnoses include benign and malignant findings, with notes providing directives for patient management, from no action to follow-ups and scheduled surgeries.
[0044] In an embodiment, the microprocessor acquires the received colorectal images and the multiple colorectal reference images from the computing device 102 and the non-transitory storage device, respectively. The microprocessor is programmed to seamlessly retrieve and store the received colorectal images and the multiple colorectal reference images for subsequent analytical procedures. The colorectal images obtained from computing device 102 are typically patient-specific and provide real-time, critical data necessary for diagnosis and treatment planning. Concurrently, the multiple colorectal reference images sourced from the non-transitory storage device serve as the colorectal images database, against which patient images are compared and analysed. The comparison is essential in identifying anomalies, patterns, and indicators that may signify various colorectal conditions.
[0045] In an embodiment, the microprocessor ensures the integrity and quality of the images during the acquisition process. The microprocessor can handle a wide range of image formats and resolutions, thus enhancing the applicability of system 100 across different imaging modalities and technologies. The efficiency and accuracy of microprocessor in acquiring the colorectal images and the multiple colorectal reference images are paramount, as they form the foundation upon which further AI-driven analysis is conducted.
[0046] In a preceding embodiment, the microprocessor executes the set of routines focused on the enhancement of acquired colorectal images for detection of colorectal cancer. Central to the process is the removal of noise from each acquired colorectal image, a critical step in generating a denoised colorectal image, thereby significantly improving the image quality for subsequent analysis. Noise in colorectal images may include, but is not limited to, Gaussian noise, speckle noise, salt-and-pepper noise, and Poisson noise, each potentially degrading the image quality in unique ways. To address the noise, the microprocessor employs a variety of noise removal techniques, each tailored to effectively mitigate specific types of noise. For Gaussian noise, which is characterized by statistical noise having a probability density function equal to that of the normal distribution, techniques such as Gaussian filtering or advanced deep learning-based methods are utilized. Speckle noise, commonly present in ultrasound images and known for the grainy appearance that obscures image details, is addressed using methods like speckle reducing anisotropic diffusion or wavelet-based techniques. For the removal of salt-and-pepper noise, characterized by sharp and sudden disturbances in the image signal, median filtering or adaptive median filters are effectively employed. Lastly, Poisson noise, often encountered in low-light conditions and leading to variations in pixel intensities, is mitigated using variance-stabilizing transformations or advanced denoising convolutional neural networks. Each technique is selectively applied by the microprocessor based on the specific noise characteristics observed in the colorectal images, ensuring optimal noise reduction while preserving vital diagnostic information. The result is a set of denoised colorectal images that provide enhanced clarity and detail, crucial for accurate diagnosis of colorectal cancer.
[0047] In an embodiment, the microprocessor extracts a region of interest (ROI) from each frame of the denoised colorectal image volume. The denoising of the colorectal images is achieved prior to the extraction process, ensuring clarity and precision in the subsequent analysis. The extraction of the ROI by the microprocessor involves the implementation of advanced image processing techniques, which may include, but are not limited to, edge detection algorithms, thresholding methods, and contour detection strategies. The techniques are employed to accurately identify and isolate the ROI, which is critical for detailed colorectal examination and diagnosis.
[0048] In an embodiment, edge detection algorithms utilized by the microprocessor focus on identifying significant discontinuities in the denoised colorectal image, which often correspond to the boundaries of the ROI. The edge detection algorithms are adept at discerning the edges within the colorectal images, thereby facilitating the precise demarcation of the ROI. In addition to edge detection, thresholding methods are also employed. The thresholding methods involve setting a specific threshold value, and then segregating the pixel values in the image based on the threshold, further aiding in the isolation of the ROI. Furthermore, contour detection strategies are implemented to trace the outline of the ROI, providing an accurate shape and size analysis. The combination of the aforesaid techniques by microprocessor ensures a highly accurate and efficient extraction of the ROI from each denoised colorectal image. The ROI extraction process is instrumental in enabling a thorough and detailed analysis of colorectal conditions, thereby enhancing the diagnostic capabilities of the AI-based system 100.
[0049] In an embodiment, the microprocessor executes a set of specialized routines for the development of a deep learning model, specifically tailored for the analysis of colorectal images related to colorectal health. The development process involves the critical step of analyzing a multitude of acquired colorectal reference images. Such analysis by said microprocessor is conducted through advanced image processing techniques, utilizing neural network architectures. The deep learning model, as developed, employs convolutional neural networks (CNNs), known for their efficacy in handling complex image data. Each layer of the CNNs is designed to recognize various patterns and features within the colorectal reference images, ranging from simple edges to more complex structures relevant to colorectal pathology.
[0050] In another embodiment, the microprocessor implements backpropagation algorithms to iteratively adjust the weights of the neural network, ensuring the accuracy of deep learning model in interpreting the colorectal images. In the process of developing the deep learning model, the microprocessor also applies techniques like data augmentation and transfer learning to enhance the robustness and generalizability of deep learning model. Data augmentation, executed by the microprocessor, involves the alteration of original images in a controlled manner to increase the diversity of the training dataset, thereby enabling the model to learn from a more set of variations found in colorectal images. Transfer learning, on the other hand, allows the microprocessor to utilize pre-trained models on large datasets, adapting them to the specific task of colorectal image analysis, thereby reducing the need for extensive training data specific to colorectal pathology.
[0051] In an embodiment, the microprocessor applies the developed deep learning model to each extracted ROI in the colorectal image, with the specific objective of detecting one or more indicative features. The application by the microprocessor involves a process where the deep learning model analyzes each ROI. The microprocessor is equipped to discern and highlight medical conditions or anomalies (such as colorectal cancer, etc.), signified by the one or more indicative features within the ROIs. The operation entails a detailed examination of the colorectal images, utilizing the proficiency of deep learning model in recognizing complex patterns and correlations that are not readily apparent. The identification of the features by the microprocessor is crucial, as the identification contributes significantly to accurate and early diagnosis of colorectal cancer. The precision and efficiency with which the microprocessor conducts the analysis ensure the reliability and effectiveness of the feature detection process.
[0052] In an embodiment, upon executing the stored set of routines, the microprocessor performs the critical function of generating a likelihood score that is essential in the process of colorectal cancer detection, as the likelihood score quantifies the probability of the presence of the disease based on analyzed data. Concurrently, the microprocessor enables the generation of summarized findings. The summarized findings are derived from the detection of one or more indicative features, which are identified as markers of colorectal cancer. The identification of the one or more indicative features is accomplished through the analysis of relevant data, which may include, but is not limited to, medical images, patient history, and other pertinent biomedical signals. Upon the successful identification of the one or more indicative features, the microprocessor synthesizes the information into a summary. that encapsulates the detected features and provides contextual insights that are instrumental in the diagnosis and subsequent treatment planning for colorectal cancer. The process of generating both the likelihood score and the summarized findings is characterized by a high degree of precision and accuracy, ensuring that the detection of colorectal cancer is reliable and effective. Furthermore, generation of likelihood score and the summarized findings is executed in a manner that optimizes computational efficiency, thereby facilitating timely diagnosis and intervention.
[0053] FIG. 3 illustrates an exemplary tabular representation of likelihood score and summarized findings of multiple patients. As illustrated, each patient is assigned a likelihood score reflecting their cancer risk percentage and a summary of findings. Patient 1 has the highest likelihood of cancer at 85%, noted for multiple indicative features. Patient 2 has a 30% likelihood with a moderate risk, while Patient 3 is at the lowest risk with 10%. Patient 4 has a significant risk at 70%, and Patient 5 has an intermediate risk at 55%. The evaluations suggest a method of triaging or monitoring patients based on their cancer risk factors.
[0054] In an embodiment, the microprocessor displays the generated likelihood score and the summarized findings on the computing device 102. The displaying is executed following the analysis of colorectal images, where microprocessor computes a likelihood score indicative of specific medical conditions. Simultaneously, the microprocessor synthesizes summarized findings, which provide a concise interpretation of the analyzed colorectal images. The display of both the likelihood score and the summarized findings on computing device 102 is critical, offering clear and immediate access to diagnostic information necessary for medical decision-making.
[0055] In an embodiment, the medical observation data may encompass historical diagnostic data, a key component that provides invaluable context from past diagnostic instances, thereby enriching the analytical capabilities of the system 100. Additionally, the colorectal image database integrates patient outcomes, which include varied responses to treatments and their longitudinal health impacts, adding essential dimensions to the analytical framework of system 100. Complementing the aforesaid are annotations by medical professionals, a crucial element that imparts expert insights and evaluations into the colorectal image database. The annotations, derived from seasoned healthcare practitioners, significantly augment the accuracy and relevance of the analysis. The synergistic integration of historical diagnostic data, patient outcomes, and professional annotations within the medical observation data tagged to each colorectal reference image empowers system 100. The integration facilitates a more nuanced, precise, and effective analysis of medical images, thereby enhancing diagnostic procedures and patient care in the area of colorectal health.
[0056] In an embodiment, the microprocessor may employ one or more filters for the noise removal purpose, wherein the one or more filters are specifically selected from a group that consists of Gaussian blur, median filtering, non-local means denoising, and machine learning-based denoising techniques. Each of the filters contributes uniquely to the noise reduction process. The Gaussian blur filter, applied by the microprocessor, facilitates the smoothing of image noise by averaging the pixels in a manner that approximates a Gaussian distribution, thereby reducing image granularity. Median filtering, alternatively utilized by said microprocessor, replaces value of each pixel with the median value of neighboring pixels, effectively removing salt-and-pepper noise. Furthermore, the non-local means denoising technique, also employed by such microprocessor, operates y comparing all patches in the image and averaging similar ones to preserve detailed structures while reducing noise. Lastly, machine learning-based denoising techniques, when selected by said microprocessor, utilizes set of routines to learn patterns of noise and effectively eliminate them, thereby enhancing the clarity and diagnostic utility of the medical images processed by the AI-based system 100.
[0057] In an embodiment, the microprocessor may utilize at least one technique selected from a group consisting of edge detection, thresholding, and one or more morphological operations for the purpose of extracting the ROI from each denoised colorecta

requirements or preferences through the input means. Customization options include, but are not limited to, altering the format, size, color, and layout of the displayed information. Additionally, the system 100 stores the user-defined display settings within the non-transitory storage device, ensuring that preferred configurations are readily accessible for future analyses. Furthermore, the microprocessor is adept at adjusting the display automatically based on the historical preferences of user, as determined through the input means. Such capability enhances the efficiency and user-friendliness of the AI-based system 100, providing a more efficient, personalized, and user-oriented experience in the analysis of colorectal images.
[0059] In an embodiment, the server arrangement 104 may receive and incorporate additional relevant patient data. Such patient data is selected from a group comprising genetic information, one or more lifestyle events, demographic data, and previous medical history. Upon receipt, said server arrangement 104 is adapted to process and integrate the data with the medical images for enhanced analysis. The genetic information is utilized to ascertain any predispositions or genetic markers that might influence the interpretation of the colorectal images. Lifestyle events, including but not limited to exercise habits, dietary patterns, and exposure to environmental factors, are considered by the server arrangement 104 to provide an understanding of the health status of patient. Demographic data, encompassing age, gender, ethnicity, and other relevant characteristics, is employed to refine the analysis based on population-specific trends and predispositions. Furthermore, the previous medical history is analyzed by the server arrangement 104 to identify any recurring patterns or historical health issues that may be pertinent to the current medical image analysis.
[0060] In an embodiment, the server arrangement 104 may process and analyse colorectal images. The output of the analysis includes a set of summarized findings, which are derived from a multi-faceted approach. Firstly, quantitative measurements of detected features in the colorectal images are computed. The aforesaid measurements are critical in providing objective, numerical data regarding the anomalies or characteristics identified within the images. Secondly, the server arrangement 104 conducts a comparative analysis with similar cases. The comparative analysis is achieved by accessing and utilizing the colorectal image database. Such comparative analysis aids in enhancing the accuracy of diagnosis by juxtaposing the current case with precedent cases, thereby offering a contextual understanding of the imaged condition. Lastly, the AI-based system 100, through said server arrangement 104, generates recommendations for add-on diagnostic procedures. The generated recommendations are tailored based on the unique aspects of each case, derived from the initial analysis and comparative study.
[0061] In an embodiment, the summarized findings generated by the system 100 may encompass a variety of critical elements. Such elements include, but are not limited to, quantitative measurements of detected features within the analyzed colorectal images. The server arrangement 104 calculates the measurements, ensuring precise quantification of relevant image characteristics. Additionally, comparative analysis with similar cases is conducted, utilizing the colorectal image database. The comparative analysis by the server arrangement 104 aids in identifying nuanced patterns and anomalies that may not be immediately evident. Moreover, the system 100 extends the functionality to provide recommendations for add-on diagnostic procedures. Such recommendations are formulated based on the analysis of the current colorectal image in conjunction with the comparative study, offering a holistic approach to patient diagnosis and care. The incorporation of the three key components - quantitative measurements, comparative analysis, and diagnostic recommendations - into the summarized findings by the AI-based system 100, ensures a thorough and nuanced understanding of each individual case, thereby enhancing the diagnostic accuracy and effectiveness of medical practitioners in the field of colorectal health.
[0062] In an embodiment, upon completion of the analysis, the server arrangement 104 may generate a report. The generated report comprises a summary of one or more indicative features detected within the analyzed medical images. Additionally, the server arrangement 104 calculates and includes the likelihood score within the report. The likelihood score is representative of the probability that the detected indicative features correlate with specific medical conditions or anomalies. Furthermore, the report, as generated by the server arrangement 104, suggests next steps for treatment. The suggested steps are formulated based on the analysis of the colorectal images and the corresponding likelihood score. The inclusion of the suggested next steps for treatment is aimed at assisting medical professionals in making informed decisions regarding patient care.
[0063] FIG. 4 illustrates a method 400 for analyzing medical images using an artificial intelligence (AI)-based system, in accordance with the embodiments of the present disclosure. At step 402, a computing device receives colorectal images of a patient from a medical imaging device. The step 402 involves the transmission of raw image data, which typically occurs over a secure healthcare network to ensure patient confidentiality and data integrity. At step 404, a server arrangement is coupled to the computing device through a network interface. The server arrangement includes a non-transitory storage device, which contains a set of executable routines and a colorectal image database that comprises multiple colorectal reference images, with each image tagged with corresponding medical observation data. At step 406, microprocessor executes the set of routines, wherein the first routine acquires the received colorectal images from the computing device. Simultaneously, another routine retrieves the multiple colorectal reference images from the non-transitory storage device, ensuring that all necessary image data is available for analysis. At step 408, each acquired colorectal image undergoes a denoising process to remove noise artifacts, resulting in the generation of a denoised colorectal image. The process enhances the clarity of the images, which is critical for accurate feature detection. At step 410, from each denoised colorectal image, a region of interest (ROI) is extracted. The step 410 focuses the analysis on specific areas of the image that are most likely to contain diagnostic information relevant to colorectal cancer. At step 412, a deep learning model is developed by analyzing the acquired multiple colorectal reference images. The deep learning model learns to identify patterns and features associated with colorectal cancer from the reference dataset. At step 414, the developed deep learning model is applied to each extracted ROI to detect one or more indicative features that may suggest the presence of colorectal cancer. The application involves the model scanning the ROIs to identify and interpret complex features within the images. At step 416, based on the indicative features detected by the deep learning model, a likelihood score and summarized findings are generated. The likelihood score quantifies the probability of the presence of colorectal cancer, while the summarized findings provide a concise interpretation of the image analysis. At step 418, the generated likelihood score and summarized findings are displayed on the computing device. The display provides medical practitioners with immediate access to the analysis results, enabling them to make informed decisions regarding further diagnostic or therapeutic procedures.
[0064] FIG. 5 depicts a systematic diagram illustrating the operational flow of an AI-based medical image analysis system, in accordance with the embodiments of the present disclosure. The process begins with raw CT colonography DICOM images, which undergo pre-processing to refine image quality for more accurate analysis. The pre-processed images are then utilized to generate focused patches around the inner colon and rectal surface, concurrently eliminating irrelevant areas to enhance analytical precision. The refined workflow includes a critical localization of colon and rectal regions, setting the stage for the deployment of the deep learning model tailored for polyp detection and quantification. The deep learning model performs an advanced annotation of the medical images, identifying and quantifying polyps vital for colorectal assessments. The process culminates in the automated generation of the report, encapsulating the detailed findings from the annotated images. The report is instrumental for medical professionals, providing a thorough summary that highlights the presence of polyps and suggests quantitative measures, thereby enabling informed and expedient medical interventions.
[0065] The foregoing description of the specific embodiments will so fully reveal the general nature of the embodiments herein that others can, by applying current knowledge, readily modify and/or adapt for various applications such specific embodiments without departing from the generic concept, and, therefore, such adaptations and modifications should and are intended to be comprehended within the meaning and range of equivalents of the disclosed embodiments. It is to be understood that the phraseology or terminology employed herein is for the purpose of description and not of limitation. Therefore, while the embodiments herein have been described in terms of preferred embodiments, those skilled in the art will recognize that the embodiments herein can be practiced with modification within the spirit and scope of the embodiments as described herein.
[0066] It should be apparent to those skilled in the art that many more modifications besides those already described are possible without departing from the inventive concepts herein. The inventive subject matter, therefore, is not to be restricted except in the spirit of the appended claims. Moreover, in interpreting both the specification and the claims, all terms should be interpreted in the broadest possible manner consistent with the context. In particular, the terms “comprises” and “comprising” should be interpreted as referring to elements, components, or steps in a non-exclusive manner, indicating that the referenced elements, components, or steps may be present, or utilized, or combined with other elements, components, or steps that are not expressly referenced. Where the specification claims refer to at least one of something selected from the group consisting of A, B, C … and N, the text should be interpreted as requiring only one element from the group, not A plus N, or B plus N, etc.

CLAIMS
What is claimed is:
1. An artificial intelligence (AI)-based system for analysis of the medical images, comprising:
a computing device to receive the colorectal images of a patient from a medical imaging device; and
a server arrangement coupled to the computing device via a network interface, wherein the server arrangement comprising:
a non-transitory storage device to store a set of executable routines and a colorectal image database comprising the multiple colorectal reference images, wherein each of the colorectal reference image is tagged with a medical observation data;
a microprocessor executes the set of routines to:
acquire the received colorectal images and the multiple colorectal reference images from the computing device and the non-transitory storage device, respectively;
remove noise from the each acquired colorectal image to generate a denoised colorectal image;
extract a region of interest (ROI) from the each denoised colorectal image;
develop a deep learning model by analysing the acquired multiple colorectal reference images;
apply the developed deep learning model on the each extracted ROI to detect one or more indicative features;
generate a likelihood score and the summarized findings based on the detected one or more indicative features to detect the colorectal cancer; and
display the generated likelihood score and the summarized findings at the computing device.
2. The system of claim 1, wherein the medical observation data tagged with each colorectal reference image in the colorectal image database is selected from historical diagnostic data, patient outcomes, and annotations by medical professionals.
3. The system of claim 1, wherein the microprocessor utilizes one or more filters selected from the group consisting of Gaussian blur, median filtering, non-local means denoising, and machine learning-based denoising techniques to remove noise.
4. The system of claim 1, wherein the microprocessor utilizes at least one technique selected from an edge detection, thresholding, and one or more morphological operations to extract ROI from each denoised colorectal image.

6. The system of claim 1, wherein the display of the generated likelihood score and summarized findings is customizable based on a user preference.
7. The system of claim 1, wherein the server arrangement receives and incorporates additional relevant patient data selected from a genetic information, one or more lifestyle events, a demographic data, and a previous medical history.
8. The system of claim 1, wherein the summarized findings comprise at least one from: quantitative measurements of detected features, comparative analysis with similar cases from the colorectal image database, and the recommendations for the add-on diagnostic procedure.
9. The system of claim 1, wherein the server arrangement generates a report that includes a summary of detected one or more indicative features, likelihood score, and suggested next steps for treatment.
10. A method for analyzing medical images using an artificial intelligence (AI)-based system, the method comprising the steps of:
receiving, by a computing device, colorectal images of a patient from a medical imaging device;
coupling a server arrangement to the computing device via a network interface, the server arrangement comprising a non-transitory storage device to store a set of executable routines and a colorectal image database comprising multiple colorectal reference images, each tagged with medical observation data;
executing, by a microprocessor, the set of routines to acquire the received colorectal images from the computing device and the multiple colorectal reference images from the non-transitory storage device;
removing noise from each acquired colorectal image to generate a denoised colorectal image;
extracting a region of interest (ROI) from each denoised colorectal image;
developing a deep learning model by analyzing the acquired multiple colorectal reference images;
applying the developed deep learning model on each extracted ROI to detect one or more indicative features;
generating a likelihood score and summarized findings based on the detected one or more indicative features for the detection of colorectal cancer; and
displaying the generated likelihood score and summarized findings on the computing device.

ARTIFICIAL INTELLIGENCE (AI)-BASED SYSTEM FOR ANALYSIS OF THE MEDICAL IMAGES
ABSTRACT
The present disclosure describes an artificial intelligence (AI)-based system for processing medical images to detect colorectal cancer. The system incorporates a computing device receiving patient-specific colorectal images from imaging equipment and a server connected via network interface. The server arrangement includes a non-transitory storage device that stores executable routines and a database of annotated colorectal reference images. A microprocessor executes the routines to cleanse the images of noise and extract regions of interest (ROIs) using advanced image processing and machine learning techniques. Subsequently, a deep learning model is developed by analyzing the reference images. The deep learning model is then employed to identify indicative features within the ROIs. The system calculates a likelihood score and generates summarized findings related to the presence of colorectal cancer, which are displayed on the computing device. The AI-based system aims to facilitate early detection and accurate diagnosis of colorectal cancer through advanced image analysis techniques.
Fig. 1 , Claims:CLAIMS
What is claimed is:
1. An artificial intelligence (AI)-based system for analysis of the medical images, comprising:
a computing device to receive the colorectal images of a patient from a medical imaging device; and
a server arrangement coupled to the computing device via a network interface, wherein the server arrangement comprising:
a non-transitory storage device to store a set of executable routines and a colorectal image database comprising the multiple colorectal reference images, wherein each of the colorectal reference image is tagged with a medical observation data;
a microprocessor executes the set of routines to:
acquire the received colorectal images and the multiple colorectal reference images from the computing device and the non-transitory storage device, respectively;
remove noise from the each acquired colorectal image to generate a denoised colorectal image;
extract a region of interest (ROI) from the each denoised colorectal image;
develop a deep learning model by analysing the acquired multiple colorectal reference images;
apply the developed deep learning model on the each extracted ROI to detect one or more indicative features;
generate a likelihood score and the summarized findings based on the detected one or more indicative features to detect the colorectal cancer; and
display the generated likelihood score and the summarized findings at the computing device.
2. The system of claim 1, wherein the medical observation data tagged with each colorectal reference image in the colorectal image database is selected from historical diagnostic data, patient outcomes, and annotations by medical professionals.
3. The system of claim 1, wherein the microprocessor utilizes one or more filters selected from the group consisting of Gaussian blur, median filtering, non-local means denoising, and machine learning-based denoising techniques to remove noise.
4. The system of claim 1, wherein the microprocessor utilizes at least one technique selected from an edge detection, thresholding, and one or more morphological operations to extract ROI from each denoised colorectal image.

6. The system of claim 1, wherein the display of the generated likelihood score and summarized findings is customizable based on a user preference.
7. The system of claim 1, wherein the server arrangement receives and incorporates additional relevant patient data selected from a genetic information, one or more lifestyle events, a demographic data, and a previous medical history.
8. The system of claim 1, wherein the summarized findings comprise at least one from: quantitative measurements of detected features, comparative analysis with similar cases from the colorectal image database, and the recommendations for the add-on diagnostic procedure.
9. The system of claim 1, wherein the server arrangement generates a report that includes a summary of detected one or more indicative features, likelihood score, and suggested next steps for treatment.
10. A method for analyzing medical images using an artificial intelligence (AI)-based system, the method comprising the steps of:
receiving, by a computing device, colorectal images of a patient from a medical imaging device;
coupling a server arrangement to the computing device via a network interface, the server arrangement comprising a non-transitory storage device to store a set of executable routines and a colorectal image database comprising multiple colorectal reference images, each tagged with medical observation data;
executing, by a microprocessor, the set of routines to acquire the received colorectal images from the computing device and the multiple colorectal reference images from the non-transitory storage device;
removing noise from each acquired colorectal image to generate a denoised colorectal image;
extracting a region of interest (ROI) from each denoised colorectal image;
developing a deep learning model by analyzing the acquired multiple colorectal reference images;
applying the developed deep learning model on each extracted ROI to detect one or more indicative features;
generating a likelihood score and summarized findings based on the detected one or more indicative features for the detection of colorectal cancer; and
displaying the generated likelihood score and summarized findings on the computing device.