Device for personalized radiotherapy with AI-controlled dosimetry

Abstract

Device for personalized radiotherapy using AI-controlled dosimetry, comprising: - a patient data acquisition unit configured to receive real-time data on patient-specific health information, including medical imaging data from the patient's CT, MRI and PET scans, biometric data from wearable or hospital-grade monitoring devices, electronic health records (EHRs) and previous treatment history; - The medical image processing unit, which analyzes and processes the captured image data, performs various functions, from tumor segmentation and identification of organs at risk to the spatial alignment of multimodal images for radiotherapy planning; - a CPU is configured to preprocess other collected data by (a) normalizing imaging and physiological data, (b) aggregating imaging and physiological data, (c) encoding imaging and physiological data for subsequent analysis, and (d) ensuring the compatibility and consistency of data from different sources; - an artificial intelligence (AI) controlled dosimetry machine that is communicatively connected to the central processing unit, wherein the AI-controlled dosimetry machine is configured to use machine learning algorithms to evaluate patient-specific parameters, optimize the distribution of a radiation dose and predict responses to treatment; - a visualization and treatment planning interface coupled with the AI-driven dosimetry machine, the visualization interface being configured to provide radiation dose distributions, patient-specific treatment recommendations, and comparative dose plans that enable data-driven adjustments by clinicians; and - a visualization interface with a feedback mechanism that allows real-time input from practicing clinicians and continuous automatic recording of patient response, which is used for iterative improvement of both the dosimetry model and the treatment plan.

Description

Field of the present invention

[0001] The invention relates in particular to an automated method that uses an intelligent system derived from artificial intelligence (AI) algorithms, which combines real-time imaging and patient-specific data to optimize radiation dose distribution. Background of the invention

[0002] Radiation therapy is one of the most important treatment modalities for patients with cancer and other diseases that require targeted radiation to specific anatomical locations. The success of radiation therapy depends largely on irradiating the tumor with an appropriate dose while sparing healthy tissue during total lung irradiation (VTA). However, standard protocols for radiation therapy planning and dosimetry do not fully account for individual patient characteristics, the heterogeneity of tumor response dynamics, and actual physiological changes. This non-personalized approach can lead to a suboptimal radiation dose, which impairs treatment efficacy and increases side effects such as radiation toxicity, organ damage, and reduced quality of life for the patient.

[0003] Current radiation dosimetry systems use pre-treatment imaging techniques (e.g., computed tomography (CT), magnetic resonance imaging (MRI), or positron emission tomography (PET)) to calculate pre-treatment plans. These plans are created for a fixed radiation duration and are based on the static assumption that the patient's anatomy and tumor characteristics remain unchanged during treatment. In reality, however, tumors shrink, move, or change their density in response to radiation, while the movement of internal organs due to respiration, digestion, or physiological fluctuations makes it difficult to deliver radiation as intended. Changes in patient position during multiple treatment sessions introduce additional uncertainties and can result in unintended radiation being delivered to healthy (non-target) tissue.

[0004] To account for these fluctuations, radiation oncologists and medical physicists interactively update treatment plans based on image-mediated adaptation (IMA) intervals. Accurate dose calculation is not only time-consuming and prone to error, but also provides the latest information on inaccuracies. Furthermore, modern treatment planning algorithms are computationally intensive and hardware-limited, resulting in conservative dosimetry that may not maximize the therapeutic effect. Accurate foundations for a personalized and automated dosimetry system that adapts in real time to patient-specific changes and provides data to improve the precision of radiotherapy are urgently needed.

[0005] However, recent advances in artificial intelligence (AI) and machine learning (ML) have the potential to revolutionize how we address these problems, with AI-powered dosimetry systems that optimize radiation treatment in real time. Machine learning thrives on large and persistent datasets and is well-suited to describing intertwined imaging and physiological data, recognizing patterns, and generating highly accurate figures for growth response and radiation dose distribution. These models can incorporate real-time imaging, treatment history, and biological markers to continuously optimize radiation parameters, delivering the most effective dose while minimizing exposure to healthy tissue.AI-based dosimetry, which uses deep learning algorithms, reinforcement learning agents or predictive analytics, can provide better clinical decision support, minimize the consequences of manual interventions and potentially improve the therapeutic ratio.

[0006] Although these advances have been achieved through considerable effort, significant obstacles to the widespread adoption of AI-based treatment planning in clinical practice still appear to exist. To date, predictive algorithms, particularly AI models, have been trained on retrospective datasets that may not fully represent the diverse patient population in the real world, potentially leading to biased treatment predictions. They argue that the lack of standardization and interoperability between different radiotherapy platforms and imaging systems hinders the seamless integration of AI-based dosimetry into clinical workflows. Furthermore, data privacy and security considerations pose a separate challenge, as radiotherapy data is sensitive and requires robust protection to ensure medical compliance.Similarly, AI models must be comprehensively validated before they can be used in critical medical applications, a requirement demanded by both clinicians and regulatory authorities.

[0007] We believe these challenges justify a novel AI-based dosimetry device that enables patient-specific adaptive radiotherapy in real time. Such a system would combine AI algorithms with real-time imaging, physiological monitoring, and predictive modeling to adaptively control the radiation dose distribution throughout the treatment process. The invention aims to improve the precision of radiotherapy, reduce the risk of radiation toxicity, and increase treatment efficacy by employing deep learning-based tumor segmentation, motion tracking, and adaptive treatment planning, all integrated into the system. Furthermore, the system is designed to integrate directly into existing clinical workflows to enhance accessibility and benefits for healthcare providers.

[0008] This new approach represents a major advancement in the field of radiation oncology, as it creates an intelligent and automated dosimetry system for treatment. This invention could transform personalized cancer treatment by enabling AI-driven dose optimization in real time, thus helping to ensure that every cancer patient receives the most effective and safest radiation therapy package possible. Summary of the invention

[0009] An AI-driven dosimetry device for personalized radiotherapy is presented, enabling a new approach to planning, monitoring, and adjusting radiation treatment in real time. Instead of conventional radiotherapy techniques, where treatment is guided by imaging prior to administration and static calculations are performed to estimate the delivered dose, this system combines artificial intelligence (AI) and machine learning (ML) with real-time imaging and physiological monitoring to dynamically optimize radiation dose delivery throughout the entire treatment period.

[0010] It utilizes deep learning-based image analysis techniques that enable automatic segmentation of tumors, surrounding tissue, and OARs, allowing for highly precise and accurate radiation targeting. With the help of AI-driven algorithms that analyze real-time image data, patient-specific physiological changes, and treatment history, clinical oncologists are able to continuously adjust radiation dose calculations, taking into account tumor shrinkage, organ movement, and anatomical shifts. This ensures that the radiation dose is always optimal, naturally protecting healthy tissue while simultaneously maximizing therapeutic efficacy.

[0011] Among the unique aspects of this device is its AI-based motion tracking system, which takes into account involuntary patient movements such as breathing, digestion, and muscle contractions. In conventional radiotherapy, the patient must remain still throughout the entire treatment, which is not always possible and leads to inaccuracies in dose delivery. This invention also employs real-time motion compensation algorithms to ensure precise irradiation of the tumor despite patient movement.

[0012] Furthermore, the system uses reinforcement learning and predictive analytics to forecast tumor response to radiation therapy and proactively adjust treatment protocols. The system can then tailor the therapy to the individual patient, adjusting the radiation dose according to tumor biology and expected treatment outcomes.

[0013] This device integrates seamlessly with all existing radiation equipment, such as LINACs, proton therapy modules, and adaptive radiotherapy (ART) modules. It enables cloud-based computing and secure, AI-driven data processing, adhering to HIPAA, GDPR, and other governance standards to protect patient data. Furthermore, a flexible interface allows clinicians to visualize treatment in real time and make AI-assisted decisions, enabling radiation oncologists to review, modify, or approve suggested AI-based dose adjustments before treatment is actually administered.

[0014] The invention significantly reduces the workload for clinicians, minimizes the risk of human error, and improves overall treatment precision by automating complex dosimetry calculations and enabling adaptive, real-time control of radiation therapy. The end product is a successful radiotherapy system that is more efficient, safer, and more patient-centered, ultimately improving therapeutic outcomes while minimizing radiation-induced side effects.

[0015] This is a groundbreaking, paradigm-changing advance in the field of radiation oncology, leading to "intelligent" and personalized radiotherapy delivered at the speed of light with highly precise adaptive dose delivery. This pioneering invention utilizes state-of-the-art AI technologies to ensure that each patient receives the ideal radiation dose at the optimal time, improving treatment effectiveness as well as long-term survival and patients' quality of life. Brief description of the image:

[0016] As in Fig. A personalized radiotherapy system with AI-controlled dosimetry was demonstrated. The System 100 includes: The Patient Data Acquisition Unit 102 is configured to capture patient-specific data in real time, such as medical imaging (CT, MRI, PET scans), biometric data from wearable devices or hospital-specific monitoring equipment, electronic health records (EHRs), and historical treatment data. This enables the ingestion of data from multiple sources for optimized radiotherapy planning through analysis.

[0017] The medical image processing unit 104, which is operationally coupled with the patient data acquisition unit, electronically cleans the homonymous raw image data and also performs several preprocessing operations, such as noise reduction, segmentation of tumors and organs at risk, and spatial alignment of multimodal images. It captures essential anatomical and physiological features to facilitate personalized dosimetric calculations.

[0018] The central processing unit (CPU) 106 is responsible for preprocessing the collected patient data for use by the AI ​​engine. This includes normalizing, aggregating, and coding medical imaging and physiological data for analysis. The CPU also combines historical patient data with real-time imaging data and ensures that all input data is both standardized and in a computationally modelable format.

[0019] The Data Preprocessor 104 supports data cleaning, transformation, aggregation, and normalization to prepare the data for further processing; the AI-Driven Dosimetry Engine 108 is a computing unit that uses machine learning-based models (e.g., deep learning, reinforcement learning, statistical modeling) to learn from historical data and predict an optimal radiation dose for a new patient; and the Longitudinal Patient Repository 106 is a data repository containing patient experience data. This system considers various factors such as tumor size, shape, and location, radiation sensitivity and previous treatment responses, as well as patient risk factors. The predictive modeling system regularly recalibrates its recommendations by evaluating them against past, current, and new patient data, as well as real-time treatment responses, thereby increasing the algorithm's depth and breadth.

[0020] The AI-powered dosimetry engine is used by the Visualization and Treatment Planning Interface 110, which is connected to it and provides an interactive and user-friendly environment for radiation oncologists and medical physicists. It is based on existing image processing resources, such as radiation dose distribution maps, 3D visualizations of predicted / expected tumor responses, and comparative treatment plans for clinical decision-making. It assists professionals in manually fine-tuning patient-specific treatment via the interface.

[0021] Integration with the feedback mechanism 112 via the visualization interface enables real-time user input and treatment adjustments based on the patient's actual observed reactions. The AI-assisted treatment planning by DNA can be further refined by radiation oncologists to arrive at a final, optimal plan; in addition, patient feedback regarding side effects or discomfort during treatment can be used to iteratively optimize the dosimetric recommendations with the help of AI.

[0022] Automated radiation therapy system 114, trained to execute AI-generated treatment plans with high accuracy using advanced radiation therapy techniques, including intensity-modulated radiation therapy (IMRT), volumetric modulated arc therapy (VMAT), and stereotactic body radiation therapy (SBRT). Typically, the automated system monitors treatment accuracy in real time and makes immediate adjustments to beam position and intensity to achieve optimal dose conformity and reduced exposure to healthy tissue.

[0023] In one embodiment, the patient data acquisition unit 102, for example, consists of several medical imaging modalities (CT, MRI, PET) as well as physiological monitoring devices (heart rate, blood pressure, and oxygen saturation sensors) that together provide a holistic approximation of the patient's profile. The medical image processing unit 104 uses AI segmentation to accurately delineate tumors, adjacent organs, and important structures for targeted radiation therapy. To improve the output of decision support functions, a data cleansing module in the CPU 106 filters imaging and biometric data for noise, inconsistencies, or other problems, ensuring that AI-driven dosimetry calculations receive high-quality input.

[0024] The AI-powered dosimetry engine 108 incorporates supervised and unsupervised machine learning methods, such as convolutional neural networks (CNNs) for image-based dose planning, reinforcement learning for adaptive radiation planning, and decision trees for risk stratification. With the goal of improving patient outcomes, the system refines its predictive models based on new patient data, leading to optimal radiation therapy recommendations over time.

[0025] The Visualization Interface 110 includes interactive dashboards that allow oncologists to visualize dose distribution and patient responses. Insights gained from AI enable physicians to adjust parameters such as radiation intensity, beam angle, and fractionation plans. AI-based plans can be compared in real time with conventional dose planning strategies, leading to more informed decision-making via the interface.

[0026] This feedback mechanism 112 can be used for continuous monitoring of therapy adherence and patient-reported outcomes, allowing the radiation dose to be dynamically titrated down based on real-time physiological, imaging, and patient-reported data. In some embodiments, a portable system monitors patient-reported side effects and adjusts the radiation delivery schedule accordingly.

[0027] In this way, the 114 radiation therapy system automates radiation delivery by instantly adjusting the beam's intensity and direction based on prior knowledge of potential anatomical variations and the observed position of the physiological target at the time of treatment. To ensure treatment accuracy, the system also integrates robot-assisted positioning and motion tracking functions.

[0028] The proposed system, which integrates HIS, PACS, and radiotherapy equipment, leverages existing software and hardware to streamline workflows and mitigate data transfer challenges. The device consists of an external enclosure housing all processing units and high-speed interfaces for connecting to imaging devices, EHR databases, and radiotherapy equipment. A visualization interface is available via dedicated workstations and mobile applications, enabling remote monitoring and treatment protocol customization.

[0029] Artificial intelligence also plays a crucial role in the proposed dosimetry system, making it an excellent solution for optimizing individualized treatment plans with regard to treatment time and patient safety. Utilizing this invention's new version of real-time patient data, AI-powered predictive modeling, interactive visualization, and automated radiation delivery would make treatment more precise and avoid radiation toxicity for patients, thus improving treatment outcomes. The predictive models, which leverage the power of AI, represent a turning point in personalized oncology, enabling a more targeted approach to planning and delivering radiotherapy.

Claims

[1] Device for personalized radiotherapy using AI-guided dosimetry, comprising: - a patient data acquisition unit configured to receive real-time data on patient-specific health information, including medical imaging data from the patient's CT, MRI and PET scans, biometric data from wearable or hospital-grade monitoring devices, electronic health records (EHRs), and previous treatment history; - the medical image processing unit, which analyses and processes the acquired image data, performs various functions, from tumour segmentation and identification of organs at risk to spatial alignment of multimodal images for radiotherapy planning; - a CPU is configured to preprocess other collected data by (a) normalizing imaging and physiological data, (b) aggregating imaging and physiological data, (c) encoding imaging and physiological data for subsequent analysis, and (d) ensuring compatibility and consistency of data from different sources; - an artificial intelligence (AI)-controlled dosimetry engine communicatively connected to the central processing unit, wherein the AI-controlled dosimetry engine is configured to use machine learning algorithms to evaluate patient-specific parameters, optimize the distribution of a radiation dose, and predict responses to treatment; - a visualization and treatment planning interface coupled with the AI-driven dosimetry engine, the visualization interface configured to provide radiation dose distributions, patient-specific treatment recommendations, and comparative dose plans that enable data-driven adjustments by clinicians; and - a visualization interface with a feedback mechanism that allows real-time input from practicing clinicians and continuous automatic capture of patient response, which is used to iteratively improve both the dosimetry model and the treatment plan. [2] The device according to claim 1, wherein the patient data acquisition unit includes a plurality of imaging and physiological measuring devices, such that the imaging devices acquire anatomical and functional information in real time, while the biometric sensors acquire vital signs such as heart rate, respiratory rate, blood pressure and oxygen saturation to improve treatment precision. [3] The apparatus of claim 1, wherein the medical image processing unit further comprises an image segmentation module configured to automatically segment tumor contours, organ structures, and radiation-sensitive tissues using deep learning-based models for dose-optimized planning. [4] The apparatus of claim 1, wherein the central processing unit further comprises: a data cleaning module configured to eliminate noise, artifacts, and inconsistencies in the acquired imaging and physiological data to improve the accuracy of the predictive models. [5] The apparatus of claim 1, wherein the AI-driven dosimetry engine uses supervised and unsupervised machine learning algorithms with deep learning architectures including convolutional neural networks (CNNs) for dose distribution prediction, reinforcement learning for adaptive real-time treatment strategies, and probabilistic models for risk assessment of radiation exposure to healthy tissue. [6] The apparatus of claim 1, wherein the visualization and treatment planning interface comprises interactive dashboards configured to enable radiation oncologists and medical physicists to filter and modify treatment parameters input to the one or more machine learning models, compare dose plans based on AI recommendations, and visualize predicted radiation responses on actual 3D anatomy. [7] The apparatus of claim 1, wherein the feedback loop enables open communication and robust feedback from patients about their responses to treatment and any aberrant side effects (e.g., using mobile applications, chatbots, etc.), so that the system can continuously monitor physiological responses, adherence, and side effects in real time and dynamically modify radiotherapy treatment plans based on observed patient-specific tolerances and feedback from clinicians. [8] Device according to claim 1, characterized by that it further comprises an automatic delivery system capable of delivering AI-based optimized treatment plans, wherein the delivery system adjusts the intensity, angles, and arrangements of the beams using imaging and motion tracking data in real time during treatment [9] The apparatus of claim 1, wherein the AI-driven dosimetry engine is additionally capable of periodically refining its predictive models based on additional input patient data to enable real-time recalibration of the radiation dose calculation taking into account changes in tumor response and / or improvement in the patient's physiological changes. [10] The device of claim 1, comprising a tiering of dosimetry predictions consisting of a list of alert notifications for healthcare providers and patients when critical dosimetry predictions change, and healthcare providers have the opportunity to intervene and personalise or modify treatment to improve clinical outcomes.

Images