Description:FIELD OF THE INVENTION
This invention relates to Smart IoT-Edge AI System for Real-Time Brain Tumor Detection and Segmentation using Conventional Neural Networks
BACKGROUND OF THE INVENTION
Diagnosis of brain tumors is one of the most lethal medical conditions and requires early and accurate diagnosis for effective treatment. Traditional methods for identifying brain tumors involve radiologists manually examining MRI/CT scans, a process that is not only labor-intensive but also prone to human error. In addition, these methods are often unavailable in remote or underdeveloped areas due to a shortage of specialized medical professionals. Although current cloud-based AI solutions for tumor detection are useful, existing solutions suffer from high latency, data protection, and off the grid usage of continuous internet connectivity that impedes their use for urgent real time diagnosis. Furthermore, conventional edge computing (NVIDIA Jetson Xavier/Orin and Raspberry Pi AI Gateway) approaches lack the ability to dynamically update AI models, which limits their adaptability to new medical data and reduces diagnostic accuracy over time. There is a need for an IoT-enabled Edge AI system that can perform real-time brain tumor detection and segmentation at the point of care, reducing diagnosis time while maintaining high accuracy. This system should facilitate quick decision-making in hospitals and remote healthcare facilities by guaranteeing low-latency inference, enabling smooth cloud-based AI model upgrades, and giving doctors secure remote access.
US7035456B2: A method (100) of locating human faces, if present, in a cluttered scene captured on a digital image (105) is disclosed. The method (100) relies on a two step process, the first being the detection of segments with a high probability of being human skin in the color image (105), and to then determine a bounday box, or other boundary indication, to border each of those segments. The second step (140) is the analysis of features within each of those boundary boxes to determine which of the segments are likely to be a human face. As human skin is not highly textured, in order to detect segments with a high probability of being human skin, a binary texture map (121) is formed from the image (105), and segments having high texture are discarded.
US11687778B2: Detection of synthetic content in portrait videos, e.g., deep fakes, is achieved. Detectors blindly utilizing deep learning are not effective in catching fake content, as generative models produce realistic results. However, biological signals hidden in portrait videos which are neither spatially nor temporally preserved in fake content, can be used as implicit descriptors of authenticity. 99.39% accuracy in pairwise separation is achieved. A generalized classifier for fake content is formulated by analyzing signal transformations and corresponding feature sets. Signal maps are generated, and a CNN employed to improve the classifier for detecting synthetic content. Evaluation on several datasets produced superior detection rates against baselines, independent of the source generator, or properties of available fake content. Experiments and evaluations include signals from various facial regions, under image distortions, with varying segment durations, from different generators, against unseen datasets, and under several dimensionality reduction techniques.

SUMMARY OF THE INVENTION
This summary is provided to introduce a selection of concepts, in a simplified format, that are further described in the detailed description of the invention.
This summary is neither intended to identify key or essential inventive concepts of the invention and nor is it intended for determining the scope of the invention.
The present invention discloses a smart IoT-Edge AI system designed for real-time brain tumor detection and segmentation. The system includes an IoT-enabled MRI acquisition module capable of transmitting MRI images to an Edge AI gateway. The Edge AI gateway is embedded with a hybrid deep learning model comprising YOLOv12 for tumor detection and Segment Anything Model 2 (SAM2) for segmentation, enabling both rapid and precise analysis. The processing unit within the gateway is optimized using TensorRT and quantization techniques to achieve low-latency inference while reducing computational expenses. A secure transmission module ensures that detected and segmented tumor data is communicated to cloud storage, where further updates and secure accessibility are enabled for medical professionals.
In one embodiment, the Edge AI gateway is built on low-power AI hardware platforms such as NVIDIA Jetson Xavier/Orin or Raspberry Pi AI gateway devices. This allows deployment in both advanced hospitals and resource-limited healthcare environments where computational efficiency and portability are critical.
The YOLOv12 architecture employed in the system incorporates an area attention mechanism, which significantly enhances the efficiency and accuracy of real-time brain tumor detection. By focusing computational resources on critical regions within MRI images, the model achieves more robust performance compared to conventional methods.
The SAM2 segmentation module generates high-resolution masks that effectively delineate tumor boundaries, overcoming limitations observed in U-Net and DeepLabV3 models. This ensures precise identification of tumors that may be small, irregularly shaped, or overlapping, thereby improving diagnostic accuracy.
The Edge AI gateway is designed to process MRI images locally and transmit only the AI-derived tumor detection and segmentation results to the cloud. This approach minimizes the transfer of sensitive raw medical data, thereby enhancing privacy compliance with HIPAA and GDPR standards while maintaining system efficiency.
The cloud service platform of the invention consists of multiple functional components. A Cloud Controller distributes computational resources and deploys updated AI models to edge devices. Service Directors manage AI operations across the hybrid cloud-edge framework. Protection Supervisors safeguard against cybersecurity threats and ensure compliance with medical data privacy regulations. Information Directors handle the storage of processed outcomes and AI-generated reports, while Service Observers monitor inference accuracy, data flow, and system performance, issuing alerts in case of failure.
Through this hybrid cloud-edge integration, remote physicians are able to securely access AI-generated MRI reports in real time. This enables timely diagnosis and treatment recommendations, even when doctors are not physically present at the healthcare facility.
To further reduce latency during transmission, fog computing environments are incorporated between the edge and cloud layers. These fog nodes provide intermediate AI processing, ensuring continuous performance and low-latency inference.
The Edge AI gateway is further designed to maintain continuous inference capability even when cloud connectivity is disrupted. This ensures uninterrupted diagnostic performance in regions with poor or unstable internet access.
Finally, the combined use of YOLOv12 for detection and SAM2 for segmentation offers a unique advantage by improving both accuracy and response time in brain tumor analysis. The system is adaptable to a wide range of medical environments, from resource-rich hospitals to remote or underdeveloped healthcare facilities, thereby democratizing access to advanced diagnostic tools.
To further clarify advantages and features of the present invention, a more particular description of the invention will be rendered by reference to specific embodiments thereof, which is illustrated in the appended drawings. It is appreciated that these drawings depict only typical embodiments of the invention and are therefore not to be considered limiting of its scope. The invention will be described and explained with additional specificity and detail with the accompanying drawings.
Both U-Net and DeepLabV3 segmentation methods face difficulties in accurately tracing boundaries, especially if the tumors are small, irregular or overlap. These challenges often lead to segmentation errors that have adverse effects on both diagnosis and treatment plan and typically lead to lower Dice score (i.e., less than optimal alignment of predicted and actual tumor regions). To address this challenge, a hybrid segmentation model consisting of CNN architectures such as Yolov12 with the Transformer model, Segment Anything Model 2 (SAM), is proposed. Since the SAM2 model is especially built for segmentation tasks that require accuracy and flexibility, it is more prone to segmentation. Excellent for creating high resolution masks for intricate borders and number of shapes. Yolov12 is a robust technique for real-time brain tumor identification since it incorporates the area attention in order to significantly improve detection efficiency and accuracy.
BRIEF DESCRIPTION OF THE DRAWINGS
The illustrated embodiments of the subject matter will be understood by reference to the drawings, wherein like parts are designated by like numerals throughout. The following description is intended only by way of example, and simply illustrates certain selected embodiments of devices, systems, and methods that are consistent with the subject matter as claimed herein, wherein:
Fig.1.Proposed Method Workflow diagram
The figures depict embodiments of the present subject matter for the purposes of illustration only. A person skilled in the art will easily recognize from the following description that alternative embodiments of the structures and methods illustrated herein may be employed without departing from the principles of the disclosure described herein.
DETAILED DESCRIPTION OF THE INVENTION
The detailed description of various exemplary embodiments of the disclosure is described herein with reference to the accompanying drawings. It should be noted that the embodiments are described herein in such details as to clearly communicate the disclosure. However, the amount of details provided herein is not intended to limit the anticipated variations of embodiments; on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the scope of the present disclosure as defined by the appended claims.
It is also to be understood that various arrangements may be devised that, although not explicitly described or shown herein, embody the principles of the present disclosure. Moreover, all statements herein reciting principles, aspects, and embodiments of the present disclosure, as well as specific examples, are intended to encompass equivalents thereof.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments. As used herein, the singular forms “a",” “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises,” “comprising,” “includes” and/or “including,” when used herein, specify the presence of stated features, integers, steps, operations, elements and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and/or groups thereof.
It should also be noted that in some alternative implementations, the functions/acts noted may occur out of the order noted in the figures. For example, two figures shown in succession may, in fact, be executed concurrently or may sometimes be executed in the reverse order, depending upon the functionality/acts involved.
In addition, the descriptions of "first", "second", “third”, and the like in the present invention are used for the purpose of description only, and are not to be construed as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. Thus, features defining "first" and "second" may include at least one of the features, either explicitly or implicitly.
Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which example embodiments belong. It will be further understood that terms, e.g., those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
Both U-Net and DeepLabV3 segmentation methods face difficulties in accurately tracing boundaries, especially if the tumors are small, irregular or overlap. These challenges often lead to segmentation errors that have adverse effects on both diagnosis and treatment plan and typically lead to lower Dice score (i.e., less than optimal alignment of predicted and actual tumor regions). To address this challenge, a hybrid segmentation model consisting of CNN architectures such as Yolov12 with the Transformer model, Segment Anything Model 2 (SAM), is proposed. Since the SAM2 model is especially built for segmentation tasks that require accuracy and flexibility, it is more prone to segmentation. Excellent for creating high resolution masks for intricate borders and number of shapes. Yolov12 is a robust technique for real-time brain tumour identification since it incorporates the area attention in order to significantly improve detection efficiency and accuracy.
The process begins with acquiring MRI images from an MRI scanner and transmitting them to an Edge AI gateway via an MRI-enabled IoT system. Real time MRI scans are processed by the Edge AI gateway, which is pre trained provided with hybrid deep learning model consisting of Yolov12 for tumour detection and SAM2 for segmentation. When the MRI images are received, tumour detection and segmentation are performed locally in the Edge AI gateway and this reduces latency as well as reliance on cloud computing. The processed images, including the detected tumour regions and segmented areas, are then securely transmitted to cloud storage. This facilitates doctors to look through and assess the processed MRI images from any place. Furthermore, the cloud storage system provides security, in that patient data can only be retrieved by secure medical people who would also support privacy and complying with the medical regulations. This streamlined workflow allows remote doctors to quickly view and assess tumour detection and segmentation results for timely diagnosis of the disease and treatment recommendation. This approach contributes to improvement of patient care in regions with poor availability of specialized radiologists and oncologists, which results in faster and more precise medical decision making.
Cloud Services: Management and improvement of AI based IoT healthcare solutions requires cloud based or edge computing services. the Cloud Controllers, distributes cloud computing resources and deploy AI models to the edge devices. AI Operations are handled by Service Directors who oversee from the cloud to the edge, how tasks are scheduled and distributed. Protection Supervisors attempt to handle all the cybersecurity and data privacy issues to keep the MRI/CT scan data secure and conform to the medical data security regulations. Processed tumour detection outcomes and AI generated reports are securely stored by Information Director and can be easily retrieved by medical professionals. With Service Observers, they track system performance, AI inference accuracy, cloud data flow, and it alerts when failures have occurred. Fog Computing Environments also provide some AI processing on the way from edge to cloud to minimize latencies and to keep AI inference continuous even when cloud is not available. Combined, these services provide an effective, scalable and secure environment for real time brain tumour detection and segmentation in medicine and offering remote healthcare.
This invention is a Hybrid IoT-Edge AI system that uses YOLOv12 for real-time detection of brain tumors and SAM2 (Segment Anything Model 2) for reliable tumor segmentation with low power Edge AI devices like the NVIDIA Jetson Xavier/Orin and Raspberry Pi AI Gateway. As compared to conventional cloud-based AI models with high latency, our system performs on device inference, and results in extremely quick tumor detection and segmentation at the edge. Ongoing AI model improvements in the cloud edge (hybrid cloud-edge) and remote physicians’ access to AI generated reports instantly through the IoT framework are facilitated. In addition to this, the deep learning inference is also achieved on edge devices through the use of TensorRT optimization and quantization techniques with the aim to reduce the computational expenses. This processing of raw medical images locally, sending back only AI derived tumor results to the cloud, keeps data privacy act by adhering to HIPAA and GDPR standards. Combining edge AI computing, IoT connectivity, cloud-based update and cutting edge deep learning model (YOLOv12 + SAM 2) create a unique solution offering low latency, high accuracy and cost-effective brain tumor detection suitable for the resource rich hospitals as well as the resource limited healthcare environments.
, Claims:1. A smart IoT-Edge AI system for real-time brain tumor detection and segmentation, comprising:
a) an IoT-enabled MRI acquisition module configured to transmit MRI images to an Edge AI gateway;
b) an Edge AI gateway embedded with a hybrid deep learning model consisting of YOLOv12 for tumor detection and SAM2 (Segment Anything Model 2) for tumor segmentation;
c) a processing unit within the Edge AI gateway optimized with TensorRT and quantization techniques for low-latency inference;
d) a secure transmission module for communicating detected and segmented tumor data to a cloud storage system; and
e) a cloud service platform configured to update AI models, store results, and provide secure access for medical professionals.
2. The system as claimed in claim 1, wherein the Edge AI gateway comprises low-power AI hardware including NVIDIA Jetson Xavier/Orin or Raspberry Pi AI gateway devices.
3. The system as claimed in claim 1, wherein YOLOv12 integrates an area attention mechanism to improve real-time tumor detection efficiency and accuracy.
4. The system as claimed in claim 1, wherein SAM2 generates high-resolution segmentation masks to delineate tumor boundaries with improved precision over conventional U-Net and DeepLabV3 models.
5. The system as claimed in claim 1, wherein the Edge AI gateway processes MRI images locally and transmits only AI-derived tumor results to the cloud, thereby enhancing data privacy and complying with HIPAA and GDPR standards.
6. The system as claimed in claim 1, wherein the cloud service platform comprises:
a) a Cloud Controller for distributing computing resources and deploying AI models to edge devices;
b) Service Directors for managing AI operations from cloud to edge;
c) Protection Supervisors for handling cybersecurity and data privacy issues;
d) Information Directors for storing processed outcomes and AI-generated reports; and
e) Service Observers for monitoring inference accuracy, system performance, and data flow.
7. The system as claimed in claim 1, wherein the hybrid cloud-edge framework facilitates remote physicians to securely access AI-generated MRI analysis reports in real time.
8. The system as claimed in claim 1, wherein fog computing environments are integrated between edge and cloud to provide intermediate AI processing and minimize latency during transmission.
9. The system as claimed in claim 1, wherein the Edge AI gateway enables continuous inference capability even in the absence of cloud connectivity.
10. The system as claimed in claim 1, wherein the combination of YOLOv12 detection and SAM2 segmentation improves accuracy and reduces latency in brain tumor detection, making the system suitable for both resource-rich hospitals and resource-limited healthcare environments.