A closed-loop adaptive learning-based clinical decision support system integrated with HIS and PAC systems.

Abstract

The invention encompasses a contrast optimization system with a closed-loop adaptive learning mechanism that integrates patient clinical and historical imaging data, is equipped with target HU estimation and automated phase success scoring, and aims to increase the rate of achieving the target HU value and minimize phase timing errors in computed tomography and magnetic resonance imaging examinations.

Description

1 TARIFF INTEGRATED WITH HIS AND PAC SYSTEMS, CLOSED-LOOP ADAPTIVE LEARNING BASED. CLINICAL DECISION SUPPORT SYSTEM TECHNICAL AREA 5 The invention targets HU in computed tomography and magnetic resonance imaging examinations. In order to increase the rate of achievement of the value and minimize phase timing errors, Target HU estimation, integrating the patient's clinical data and historical imaging data. and equipped with automatic phase success scoring, a closed-loop adaptive learning mechanism. It includes a contrast optimization system. 10 The invention is particularly relevant in the field of medical imaging, especially computed tomography and magnetic resonance imaging. overcoming the current limitations in the use of contrast agents in examinations It offers an innovative system and method aimed at contrast in traditional clinical practice. Most of its parameters are based on kilogram-based standard protocols, manual data entry, and operator 15 Determining the target HU value systematically based on experience This leads to inaccessibility, insufficient contrast, and phase timing errors in the image. This leads to variability in quality, the need for repeat shots, and unnecessary use of contrast. This is the case. The present invention solves this technical problem by accessing age, height, weight, and other data through a hospital information system. Clinical data such as kidney function tests and images from the past 20 years are archived through the system. a central adaptive optimization engine that automatically integrates image series It solves this through the system. The system performs automatic domain analysis on the data it integrates and A target HU estimator engine that performs HU measurements, specific to anatomical region and phase type. It determines the ideal contrast range and shoots with an automatic phase success scoring module. It objectively quantifies the phase success obtained afterwards. 25 from these modules The data is processed within a closed-loop adaptive learning engine, and the shooting results are fed into the system. The system receives feedback and adjusts the contrast volume, injection rate, and timing parameters. It updates itself through continuous improvement. Furthermore, it facilitates matching similar patients within the system. Thanks to its algorithm, it learns from successful protocols of patients with similar clinical characteristics. It offers an optimized initial plan even for patients with limited data. Thus, the invention, 30 The manual and static protocol approach is being replaced by a data-driven, target-oriented, and outcome-learning adaptive approach. by transforming it into an optimization system, the rate at which the targeted contrast level is achieved It enhances, reduces phase failures, minimizes unnecessary contrast use, and By standardizing clinical workflows, it is important for patient safety and image quality. It offers progress. 35 2 PREVIOUS TECHNIQUE Contrast agent in computed tomography and magnetic resonance imaging examinations The application, in current clinical practice, largely relies on weight-based standard protocols and manuals. injection parameter input, fixed phase timing schemes and operator experience based It is based on decisions. In a typical workflow, patient information is stored in the hospital information system 5 It is located, the examination is planned via the radiology information system, contrast parameters The syringe is inserted manually, and the resulting images are stored in an image archiving system. They are being archived. However, these systems will form a closed loop with each other, and the results... It does not work in an integrated way in which it can learn. In current practices, the patient's age, height, weight, Clinical data such as body mass index, kidney function tests, and creatinine are mostly obtained manually. 10 current laboratory results are taken into consideration in the systematic optimization process. They are not included. As a result, even with the same clinical indication and patient characteristics Significant variations in image quality occur, depending on the targeted contrast level. It is not systematically achievable. Furthermore, the concept of "good contrast enhancement" is not present in existing systems. It is generally assessed subjectively, and the target is 15, specific to the anatomical region and phase type. HU value estimation is not performed. This situation makes contrast planning measurable and This leads to the fact that it is not based on a repeatable objective. Another important aspect of the current technique is... The deficiency lies in systematically extracting past examination data from the image archiving system. The problem is that it hasn't been learned. The contrast volume, injection speed, and timing used in the previous scan were all incorrect. Automatic matching and analysis between the actual image quality obtained with the parameters 20 There is no mechanism. Systematic analysis of successful or unsuccessful phase applications. Because no learning has taken place, the same mistakes occur in different patients and at different times. It can be repeated. However, in current clinical practice, arterial, portal or late phase Its accuracy largely depends on the operator's subjective interpretation, and there is no automatic and objective phase. There is no success score mechanism. Errors in phase timing are targeted at 25. This can lead to missing the contrast enhancement pattern and a decrease in the diagnostic value of the examination. The most significant technical limitation of current systems is closed-loop adaptive optimization. The mechanism is completely absent. The image quality obtained as a result of the shooting. Because metrics and phase success information are not fed back into the system, the system malfunctions over time. They are unable to improve and learn. As a natural consequence of these shortcomings, the current 30 The inability to systematically reach the targeted HU value in applications and phase timing issues. A fundamental technical problem arises in the form of variable image quality due to errors. This technical problem arises from insufficient contrast enhancement, incorrect phase capture, and the need for a repeat shot. need, additional radiation burden, unnecessary contrast use, increased kidney risk and workflow This leads to serious clinical consequences such as inefficiency. Current systems are hardware-oriented, protocol 35 based on and dependent on manual intervention, and completely lacking the ability to learn from the outcome. It is lacking. Therefore, in contrast application, a data-driven, adaptive, and closed-loop approach is necessary. 3 There is no optimization system. This technology was developed to fill this gap. The present invention integrates clinical data with hospital information systems and image archiving systems. and automatically analyzes past imaging results, makes target HU prediction, phase objectively scores its success and, most importantly, a closed-loop adaptive learning mechanism. Thanks to its contrast optimization system, it constantly improves itself by learning from every shot result. 5 The invention presents a system that combines a manual and static protocol approach with a data-driven, target-based HU (Human Resources Unit) approach. and by transforming it into an adaptive optimization architecture that learns from the result, the targeted to increase the rate of achieving contrast enhancement level, reduce phase failures, and avoid unnecessary minimizing contrast use, eliminating the need for repeat scans, and improving clinical performance. It aims to standardize the flow. 10 Patent document US10303986B2 describes brain CT images, injury data, and machinery. A decision support system for the automated measurement of brain damage indices using learning. The system and a method implemented by a computer are being discussed. The invention relates to a patient's to automatically measure midline shift in the brain using computed tomography images 15 using and analyzing midline shift, blood volume, tissue texture pattern, and other injury data. By applying machine learning methods to features extracted from multiple sources, such as a It provides the physician with an estimate of intracranial pressure levels. The system uses Gaussian mixing. It uses a model-based hierarchical segmentation method and magnetic resonance imaging. It applies the ventricular template as preliminary information. Unlike the current invention, this system uses 20 It focuses solely on brain injuries, contrast agent optimization, and target HU value. It does not involve a predictive or closed-loop adaptive learning mechanism. The present invention, however, Directly optimizing contrast agent application, improving image quality to the target value. It offers a holistic system that adjusts and learns from the shooting results. Patent document US10192640B2 describes a fractional current reserve decision support system. The invention describes a computed tomography-based clinical decision support system. A source for calculating fractional flow reserve based on computed tomography for a patient. to determine whether or not to separate, we use available data such as coronary CT data and a To help make a decision before an examination by a radiologist or attending physician. 30 for the purpose of using a machine-learned predictor or other tool that has access to specific patient information. It implements a model. This system assesses the functional severity of coronary artery disease. It focuses on the evaluation of contrast agent injection parameters. personalized optimization, target HU-based planning, or visualization results It does not include features such as feedback within a closed loop. The present invention, 35 a directly improving contrast enhancement quality and objectively scoring phase success As a system, it fills the gap in this field. 4 Patent document number CN113223704A describes a deep learning-based computed tomography scan of the aorta. The text discusses an auxiliary diagnostic method for aneurysms. The invention relates to different hospitals. non-contrast CT scans of patients with or without aortic aneurysms from the relevant departments. It collects images and groups these images into three datasets: training, internal test, and external test sets. It separates and creates an auxiliary diagnostic tool using the Attention-Unet convolutional neural network. The system assesses the risk of aortic aneurysm and helps radiologists read and decide on films. It aims to improve its capabilities. Unlike the current invention, this system is contrast-free. They are working on the images and focusing on a specific pathology (aortic aneurysm). The present invention, however, offers a universally applicable solution for contrast-enhanced examinations, taking into account the patient's clinical and 10 integrating imaging history and optimizing contrast dose according to target HU value It is a system. Patent document US11710238B2 describes intravascular optics using deep learning. 15 systems and methods for plaque segmentation in coherence tomography images It is mentioned that the invention allows access to vascular imaging data for a patient and to present this data in advance. a deep learning machine trained to process and segment a lumen and vascular plaque. providing pre-processed data to the model and obtaining segmented vessel imaging data. The system enables the detection of plaques through intravascular imaging and facilitates clinical decision-making. It focuses on providing assistance. The present invention directly addresses 20 through contrast agent optimization. Unlike this unrelated system, CT and MRI scans target contrast parameters. Adaptively adjusting to achieve the HU value and incorporating the results into the learning cycle. It offers a layer of clinical intelligence. Patent document CN121095660A describes an asphalt mixture based on deep reasoning. The invention refers to a CT image small sample multi-scale classification method. The asphalt mixture forms a hybrid dataset consisting of CT scan images, along with aggregate. To perform multi-scale classification of asphalt mortar edges, multiple expansion ratios Many methods capture the macroscopic and mesoscopic properties of aggregates, voids, and mortar through parallel convolution. to build a scaled-down extended convolutional neural network and a deep reasoning model with weights and 30 It dynamically corrects the loss function. This system is used in the field of civil engineering. Developed for material characterization, it is used in medical imaging and contrast agents. It does not include any elements related to optimization or integration of patient data. The current system... The invention is an integrated system specifically designed for medical contrast management. 35 Patent document CN121095166A describes a deep learning-based CT scan of chronic sinusitis. The invention describes an automatic typing and intensity assessment method and system. It collects paranasal sinus CT images, preprocesses the images, and uses three improved methods. a three-dimensional residual network architecture using a typographic submodel and a convolutional block attention module It constructs a deep learning model consisting of a violence assessment sub-model based on a specific model. This system focuses on the classification of chronic sinusitis and uses contrast medium. Optimization of injection parameters, closed-loop learning, or target HU estimation, etc. It does not include features. The present invention directly improves image quality in contrast-enhanced CT scans. from this field as a contrast optimization system that improves and minimizes kidney risk. They are completely separate. Patent document CN116664904A describes 10 new technologies based on self-supervised learning. The invention refers to a medical image classification method for coronavirus infection. The masking autoencoder is based on the self-supervised method and operates independently. It creates a model supported by distillation learning. The model is computed tomography. By using scans, they aim to shorten the diagnosis time for COVID-19 patients and help radiologists make decisions. It aims to provide a support system. This system uses image 15 for infection detection. While focusing on classification, contrast agent optimization, target HU value estimation, and phase... It does not include elements such as success scoring or PACS and HIS integration. (Current) The invention optimizes contrast parameters in a personalized and adaptive way, thereby improving the image. It stands out as a system that improves quality and avoids unnecessary use of contrast. Patent document number CN116912566A describes a method based on optical coherence tomography imaging. This text discusses a classification method for the healing effect after macular hole surgery. The invention offers a solution to the differences in assessment found in various deep learning models. The adaptive model introduces an authority coefficient and model integration provides decision support. by performing this, classification of the improvement effect is made, then model hyperparameter 25 It performs deep learning model training and validation through optimization. This The system focuses on predicting surgical outcomes, while contrast agent administration is done in a closed loop. It does not include features such as adaptive learning or real-time dose optimization. The present invention enables devices that learn in real-time during contrast application and then apply them to subsequent processes. It offers an improved system architecture. 30 Patent document number CN113902702A describes a well-detected lung nodule based on computed tomography. The discussion concerns a benign-malignant auxiliary diagnostic system. The invention relates to CT scans of lung nodules. collecting images, preprocessing images, subsampling module, time The series includes a converter module consisting of a channel converter module and a gate mechanism module. In this process, a deep learning neural network is built, and the trained network is fed with CT images to be diagnosed, along with output. This system focuses on nodule diagnosis while injecting contrast medium. 6 such as optimizing parameters, estimating the target HU value, or learning from PACS data. It does not include these elements. The present invention optimizes the dose in contrast-enhanced examinations, thereby both... It directly improves both image quality and patient safety. In patent document number CN120318439A, it is stated that CT scans reveal a deep oil layer with high oil content. The text discusses a quantitative characterization method and system for learning. The invention relates to an oil. Determining CT scanning parameters through a numerical model of the fat layer, the three layers of the fat layer obtaining three-dimensional grayscale images, and comparing them with electron microscope images. by comparing, a standardized three-dimensional numerical model is created and a convolutional neural network It uses parameters such as fat content and fat saturation to extract data. This system is used for petroleum and 10 It belongs to the field of gas exploration and development, or medical imaging, contrast agent optimization, or It does not include any elements related to the integration of patient data. The present invention is purely medical. in the field, a contrast optimization system that focuses on patient safety and image quality. This is completely different from that field. Patent document CN107280697A describes lung technology based on deep learning and data fusion. The invention refers to a nodule-based stepwise decision-making method and system. It involves obtaining multiple... Pre-processing multiple CT lung images to create pre-processed CT lung nodule images. obtaining, extracting the image features of each CT lung nodule image, and Each CT lung nodule image is subject to pre-classification according to its image characteristics. holding, then imaging each pre-classified CT lung nodule image. Obtain a diagnostic model using the characteristics and corresponding comprehensive clinical examination data. It performs repeated training and iteration to achieve this. This system is used for nodule classification. while focusing, adaptive optimization of contrast agent injection parameters, closed loop It does not include features such as feedback mechanisms or target HU value estimation. The current 25 The invention targets direct image quality in contrast-enhanced imaging and improves the results of the scans. It offers a learning system. Patent document number JP2023006633A describes an optical interferometric tomography device, and this device The control method and program are discussed. The invention relates to optical interferometric tomography (30). High accuracy of the optical path length difference between the measurement light and the reference light. It enables the adjustment by providing the obtained tomography image as input. the value derived from a learned model, to a value smaller than that derived value by converting it, determining the amount of drive of a drive system and the determined amount of drive It controls the drive mechanism using this system. This system is used in optical coherence tomography devices. 35 While focusing on adjusting the optical path difference, contrast agent optimization involves patient data. It does not include any elements related to integration or clinical decision support systems. Existing 7 The invention is a holistic system that optimizes dose and volume in contrast-enhanced CT and MRI scans. They differ in this way. Patent document CN121215288A describes non-invasive prediction of thyroid cancer gene mutations. The system and method are discussed. The invention concerns surgery for patients with thyroid papillary carcinoma. prior contrast-enhanced CT image data, clinical data, and BRAFV60E gene mutation detection. It collects the results, preprocesses CT images to mark areas of interest, and extracting radiomic properties, obtaining optimal radiomic properties through feature elimination. using a machine learning algorithm based on optimal radiomic properties We are creating a radiographic prediction model and combining it with clinical risk factors to provide a clinical prediction. It creates a model. This system focuses on predicting a specific gene mutation, Optimization of contrast agent injection parameters, closed-loop adaptive learning, or It does not include elements such as improving image quality. The present invention addresses contrast. This area is characterized by its application being a system that optimizes itself and learns from every shot. They are separating. 15 Patent document KR102347496B1 describes artificial intelligence for radiation therapy planning. The invention refers to an automated multi-organ and tumor contouring system based on a single technology. computed tomography image and contour of the area of ​​interest for a patient's radiation therapy. It receives a structure file containing the information, extracts the attributes of the received contour information, and 20 using a deep learning algorithm to represent the area of ​​interest based on binary images. To create a segmentation map, we learn parameters and base our calculations on the learned parameters. It creates a decision model. This system determines the target volumes for radiation therapy. While focusing on determination, contrast agent optimization, target HU estimation, or phase success It does not include features such as scoring. The present invention improves image quality in contrast examinations by 25. Although it can indirectly contribute to radiotherapy planning by improving it, it can directly A system that optimizes contrast parameters and improves itself through closed-loop learning. They differ in this way. Patent document number CN121260410A describes an artificial intelligence-based oral disease diagnostic tool. The method, device, equipment, and environment are discussed. The invention relates to multi-source IT equipment. dynamically calibrating the density reference value and along the direction of tooth anatomy. It reduces noise through filtering, and the enamel has a high density and low decay rate. to distinguish between macroscopic contour and microscopic lesion features in the density region It uses a multi-scale convolutional neural network, based on reinforcement learning, in the root apex region 35 It dynamically balances volumetric accuracy with computational efficiency and spatial accuracy of lesions. by combining distribution, anatomical landmark locations and individualized risk factors, a priority 8 It creates a heat map. This system focuses on the diagnosis of oral diseases while using contrast. Optimization of substance injection parameters, PACS and HIS integration, or closed-loop control. The current invention does not include elements such as a loop adaptive learning mechanism. The contrasting invention... as a layer of clinical intelligence universally applicable to all CT and MRI scans They differ. 5 Patent document CN120950945A describes an artificial intelligence-based wood cutting control system. The method is described. The invention involves a high-resolution surface layer of wood to be cut. obtaining its image, X-ray tomography scan data, and acoustic wave propagation characteristics data. doing so, through a set of edge computing nodes from multidimensional sensing data 10 by creating a primary feature set through distributed feature extraction, this primary feature set is then converted into a Internode feature fusion by providing input to a federative learning framework. It performs and creates a global feature representation, then a hierarchical decision model. multi-targeted settings including cutting path, cutting depth, feed rate and spindle rotation speed. It creates an optimized cutting plan. This system belongs to the woodworking industry and is 15 related to medical imaging, contrast agent optimization, or patient data integration. It does not contain any elements. The present invention improves patient safety and image quality in the medical field. It is a special system aimed at increasing... Patent document CN120689448A describes a dynamic double-tracked positron 20 based on a physical information network. The invention refers to an emission tomography self-monitoring separation method. The invention involves a monitor. It combines kinetic modeling with deep learning technology, describing the kinetic process of the tracker. by embedding ordinary differential equations into a neural network and using a self-supervised training strategy By using this method, network parameters and kinetics can be determined without the need for large-scale clinical datasets. It updates the parameters simultaneously and provides a precise 25 of the dual-monitor PET images. This system enables the separation of tracers in PET imaging. while focusing, optimization of contrast agent injection parameters, target HU estimation or does not include elements such as phase success scoring in contrast-enhanced CT / MR examinations. Current The invention is a system specifically designed for CT and MR contrast optimization. It differs. 30 Patent document number CN116863225A describes spectral domain optical coherence tomography. diabetic retinopathy and retinal vein occlusion edema biomarker discovery method in this modality It is mentioned that the invention uses collected spectral field optical coherence tomography data to create a... training a deep learning model and identifying the focal regions in the deep model's classification decision. It shows the distribution status of biomarkers by displaying it in the form of a thermodynamic map. This system focuses on biomarker detection in eye diseases, while contrast medium 9 optimization, closed-loop adaptive learning, or integration of patient clinical data, etc. It does not include these elements. The present invention concerns dose optimization in contrast-enhanced CT and MRI examinations. It differentiates itself as a holistic system that creates and learns from each shot's outcome. Patent document number CN121071798A describes the planting and 5 aspects of roadside vegetation landscaping. The maintenance control method is discussed. The invention relates to the soil porosity distribution gradient. Spatial heat map of crumb deposit density and three-dimensional thickness of concrete pavement layer. It collects environmental data to quantitatively measure the model, and uses artificial intelligence to analyze this data. The agent provides input to the convolutional neural network model, and a residual learning module is used to process the fragments. and by analyzing the soil composition degradation index, the parameters of the detrital sediment decomposition pathway were determined, 10 The system involves creating an organic matter-based injection formula and a dynamic pressure threshold. It belongs to the fields of agriculture and landscape engineering, and includes medical imaging and contrast agent optimization. or does not include any elements relating to the integration of patient data. The present invention, In the purely medical field, patient safety and diagnostic imaging in contrast-enhanced imaging examinations. It is a system aimed at improving quality. 15 Examinations revealed the following in computed tomography and magnetic resonance imaging scans: The current technical status regarding contrast agent administration is largely based on weight-based static electricity. The protocols are based on manual parameter input and decisions derived from operator experience. These systems will form a closed loop with each other and learn from the results. It appears that they are not working in an integrated manner. Examination of existing patent documents reveals brain damage. measurement of indices, fractional flow reserve assessment, aortic aneurysm diagnosis, vascular Intravascular plaque segmentation, lung nodule classification, infection diagnosis, gene mutation prediction, in various fields such as organ contouring, oral disease diagnosis and optical coherence tomography machine learning and deep learning-based decision support systems were developed, but these 25 None of the systems offer personalized, adaptive contrast agent injection parameters. It has been determined that it is not geared towards closed-loop optimization. The existing situation in question... The systems integrate the patient's clinical data and past imaging results in a holistic manner. unable to integrate, measurable target to reach the targeted Hounsfield Unit value unable to determine or objectively score the image quality obtained after shooting. 30 And most importantly, it's a closed system that updates itself by systematically learning from the imaging results. It does not offer a cyclical mechanism. As a natural consequence of these shortcomings, the current The inability to systematically achieve the targeted HU value in applications, phase timing. errors, suboptimal contrast enhancement, need for repeat imaging, unnecessary contrast usage, and increased Serious technical problems such as renal risk arise. The present invention addresses this technical gap. 35 To fill out the form, age, height, weight, body mass index, kidney Clinical data such as functional tests and creatinine levels, along with past images, are archived through the image archiving system. Automatic region-of-interest analysis that automatically integrates DICOM data from image series. and a target HU estimator engine that performs HU measurements based on anatomical region and phase type. shooting with an automatic phase success scoring module that determines the specific ideal contrast range. objectively quantifying the phase success obtained afterwards and from all these modules By processing the data within a closed-loop adaptive learning engine, 5 from each shot result It offers a learning contrast optimization system. Furthermore, similar systems within the system... Thanks to the patient matching algorithm, successful protocols for patients with similar clinical characteristics can be selected. By learning from its data, it can generate an optimized initial plan even for patients with limited data, thus The manual and static protocol approach is being replaced by a data-driven, target-oriented, and outcome-learning adaptive approach. By transforming it into an optimization architecture, the rate at which the targeted contrast level is achieved is 10. It enhances, reduces phase failures, and minimizes unnecessary contrast use. This eliminates the need for repeat scans and standardizes clinical workflow, thus benefiting the patient. This represents a significant improvement in terms of security and image quality. Ultimately, the problems mentioned above, which cannot be solved with the current technology, are related to technical issues. 15 This has made it necessary to make an innovation in the field. A BRIEF DESCRIPTION OF THE INVENTION The present invention is a sensory device developed to eliminate the technical shortcomings mentioned above, and 20 integrated with PAC systems, closed-loop adaptive learning-based clinical decision support system. It is related. The main purpose of the invention is to improve the use of computed tomography and magnetic resonance imaging in examinations. depending on the current static and manual protocol approach to contrast agent administration The resulting systematic failure to reach the targeted HU value and phase timing 25 The aim is to offer a fundamental solution to the problem of errors. In this way, the patient's clinical data and history... Imaging results are evaluated holistically to create a personalized contrast plan. This is done so that images resulting from suboptimal contrast enhancement and incorrect phase capture are avoided. Variability in quality is eliminated. Another objective of the invention is the integration of hospital information systems and image archiving systems. The goal is to completely eliminate reliance on manual data entry. Automated clinical data acquisition and... Analysis of past DICOM images allows for the determination of age, height, weight, kidney function tests, and Quality metrics from previous contrast applications are transferred to the system accurately and completely. This minimizes operator-related errors and increases workflow efficiency. 35 11 Another aim of the invention is to estimate the target HU value specific to the anatomical region and phase type. The goal is to eliminate subjective evaluation from contrast planning. Target HU estimation engine. Thanks to this, the ideal contrast range required for the relevant anatomical region is automatically determined. This is determined by contrast volume and injection parameters, according to this measurable target. It is optimized so that image quality is standardized and becomes repeatable. 5 Another aim of the invention is to objectively digitize the phase success obtained after imaging. The aim is to provide an automated phase success scoring mechanism. Through this module, arterial, portal or whether the late phase achieves the targeted contrast enhancement criteria depends on the operator's interpretation. It is automatically evaluated without any further delay and a numerical success score is generated, thus the quality is 10. A control mechanism is established, preventing the recurrence of failed phase applications. Another aim of the invention is to enable the system to adapt to each shot using a closed-loop adaptive learning mechanism. The aim is to enable him to continuously improve himself by learning from the results. (Obtained after the scan) Quality metrics and phase success scores are fed back into the model, system contrast volume, injection 15 It updates the speed and timing parameters by refining them, thus becoming more accurate over time. It becomes efficient and adapts to different patient populations. Another aim of the invention is to match patients with similar clinical characteristics using a similar patient matching algorithm. Learning from successful contrast protocols provides a reliable starting point even for patients with limited data. The goal is to be able to create a plan. This allows for the identification of a patient of the same age, even if the patient has no prior imaging records. Successful protocols for patients with weight, renal function, and clinical indication are used as a reference. By taking these factors into account, an optimized contrast plan is created, ensuring the best possible outcome for each patient. Initial parameters are obtained. Another aim of the invention is to minimize the use of unnecessary contrast agents, thereby benefiting patients. The goal is to reduce kidney risk. Through HU-based optimization, the desired image quality is achieved. The minimum contrast volume and injection rate required to achieve the desired result are determined, thus specifying the appropriate level of contrast. In patients with impaired renal function, the risk of contrast-induced nephropathy is minimized, and the patient Security is increased. 30 Another aim of the invention is to eliminate the need for repeat scans, thus saving patients from additional radiation. The aim is to reduce the burden and workload of healthcare institutions. The target HU value in the first scan and By increasing the rate of achieving correct phase timing, inadequate contrast enhancement is achieved. This minimizes the need to repeat the examination, thus exposing patients to unnecessary radiation. 35 This avoids exposure and allows for efficient use of appointment capacity at imaging centers. 12 Another aim of the invention is to create a hardware-independent device that can be integrated into existing hospital infrastructure. It is positioned as a software-based clinical intelligence layer. From the contrast injector manufacturer and Thanks to this system, which can operate independently of the imaging device brand, hospitals' existing The system is put into operation without the need to change any hardware, thus at a low cost. And a high-tech contrast optimization solution is achieved in a short time. 5 Another aim of the invention is to standardize contrast administration processes in radiology clinics. The goal is to improve corporate quality indicators. Automated data collection, target HU estimation, phase success. Thanks to scoring and closed-loop learning, different operators and different time intervals The difference in quality between examinations is eliminated, thus improving the performance of radiology departments by 10. It becomes measurable and continuously improvable. Another aim of the invention is to develop AI-based clinical decision support systems and software in healthcare. Our aim is to offer a scalable platform architecture within the scope of medical device solutions. Modular Thanks to the system structure, radiation dose optimization will be automated in later stages. anatomical segmentation, federative learning-based multicenter adaptive modeling, and institutional Extensions such as performance analytics are enabled, so the invention is not just a single product, It becomes a constantly evolving health informatics ecosystem. All the objectives mentioned above and those that will emerge from the detailed explanation below are 20 The present invention is designed to perform computed tomography and magnetic resonance imaging. In examinations, contrast agent injection parameters are personalized and adaptive. a closed-loop adaptive learning-based clinical decision support system for optimization This system is related to the patient's demographic information, clinical indications, and laboratory data. A Clinical Data Integration Module to automatically retrieve the results of the aforementioned 25 to analyze the patient's past imaging studies in DICOM format an Image Archiving System Analysis Module, for the aforementioned Image Archiving System Automatic region of interest determination on images analyzed by the Analysis Module. A Field of Interest and HU performs Hounsfield Unit and signal-to-noise ratio measurements. The Quality Analysis Engine, the aforementioned Clinical Data Integration Module, and the aforementioned Area of ​​Interest and 30 Evaluating clinical and imaging data together, provided by the HU Quality Analysis Engine. a Central Adaptive Learning Engine, within the aforementioned Central Adaptive Learning Engine the ideal location and target to be reached according to anatomical region, phase type and clinical indication. A Target HU Prediction Engine that estimates the Hounsfield Unit range is the aforementioned Central. 35 patients with similar clinical characteristics located within the Adaptive Learning Engine A Similar Patient that performs initial optimization based on successful contrast protocols. The Matching Module ensures that the targeted contrast enhancement criteria are achieved after the shot is completed. 13 An automated system that generates a numerical phase success score by objectively assessing whether it has been reached. The Phase Success Scoring Module is provided by the aforementioned Automatic Phase Success Scoring Module. The model receives the generated quality metrics and the mentioned phase success score as feedback. updating parameters and generating optimized dosage plans for subsequent patients a Closed-Loop Adaptive Learning Engine and the aforementioned Central Adaptive Learning Engine 5 to transfer the optimum contrast plane produced by the system to the contrast injection system It includes an Injector Integration Layer. The best way to utilize the advantages of the existing invention, together with its structure and additional elements. For it to be understood, it must be evaluated together with the figures explained below. 10 BRIEF DESCRIPTION OF THE FIGURES Figure 1: Closed-loop adaptive learning-based clinical decision-making integrated with HIS and PAC systems. This is a schematic view of the support system. REFERENCE NUMBERS 1 - Clinical Decision Support System 11 - Clinical Data Integration Module 12 - Image Archiving System Analysis Module 13 - Area of ​​Interest and HU Quality Analysis Engine 20 14 - Central Adaptive Learning Engine 141 - Target HU Prediction Engine 142 - Similar Patient Matching Module - Automatic Phase Success Scoring Module 16 - Closed-Loop Adaptive Learning Engine 25 17 - Injector Integration Layer DETAILED DESCRIPTION OF THE INVENTION This detailed explanation focuses solely on the innovation in the invention to provide a better understanding of the subject matter. It is conveyed without being limited to examples. Accordingly, in the following explanation and figures, His 30 and PAC systems integrated, closed-loop adaptive learning-based clinical decision support system. It is explained. Figure 1 shows the closed-loop adaptive learning-based system integrated with the his / her and pac systems that are the subject of the invention. This is a schematic view of a clinical decision support system. Accordingly, the system consists of: 35 14  Contrast agent in computed tomography and magnetic resonance imaging examinations personalized and adaptive optimization of injection parameters A closed-loop adaptive learning-based Clinical Decision Support System (1),  Within the aforementioned Clinical Decision Support System (1) and belonging to the patient Automatically retrieve demographic information, clinical indications, and laboratory results. 5 a Clinical Data Integration Module (11),  Within the aforementioned Clinical Decision Support System (1) and the aforementioned patient Analyzing DICOM format image series from past imaging studies. Image Archiving System Analysis Module (12),  The aforementioned Clinical Decision Support System (1) includes the aforementioned Image 10 On the images analyzed by the Archive System Analysis Module (12) By automatically determining the area of ​​interest, it measures Hounsfield Unit and signal-to-noise ratio. An Area of ​​Interest and HU Quality Analysis Engine (13) that performs  Clinical Data included within the aforementioned Clinical Decision Support System (1) The Integration Module (11) and the mentioned Area of ​​Interest and HU Quality Analysis Engine (13) 15 A Central Adaptive that evaluates clinical and imaging data provided by [the organization / institution] together. Learning Engine (14),  Located within the aforementioned Central Adaptive Learning Engine (14) and anatomical The ideal Hounsfield Unit to be reached, depending on the region, phase type, and clinical indication. A Target HU Prediction Engine (141) that estimates the range, 20  Located within the aforementioned Central Adaptive Learning Engine (14) and similar starting with successful contrast protocols for patients with clinical characteristics. A Similar Patient Matching Module (142) that performs optimization,  The shooting included in the aforementioned Clinical Decision Support System (1) 25 after completion, whether the targeted contrast enhancement criteria have been achieved An Automated Phase system that generates a numerical phase success score by objectively evaluating it. Success Scoring Module (15),  The mentioned Automatic system, which is included within the aforementioned Clinical Decision Support System (1) Quality metrics produced by the Phase Success Scoring Module (15) and the mentioned phase By receiving the success score as feedback, it updates the model parameters and then the next 30 A closed-loop adaptive learning system that generates optimized dosage plans for patients. The engine (16) and  The Central part of the aforementioned Clinical Decision Support System (1) The optimum contrast plan generated by the Adaptive Learning Engine (14) contrast An Injector Integration Layer (17) 35 to transfer to the injection system It includes. The invention concerns contrast in computed tomography and magnetic resonance imaging examinations. personalized and adaptive optimization of substance injection parameters Closed-loop adaptive learning-based Clinical Decision Support System (1), patient-based clinical data and past imaging results are integrated in a closed loop. It operates on the principle of conducting a continuously improving optimization process. The system works in 5 stages. The process begins with the Clinical Data Integration Module (11) which collects the patient's demographic data. information, clinical indications and laboratory results are automatically retrieved from the hospital information system. It is initiated by receiving the data. Simultaneously, the Image Archiving System Analysis Module (12) DICOM format images of the patient's past imaging studies. analyzes the series and these images are processed by the Area of ​​Interest and HU Quality Analysis Engine (13) 10 is processed. The mentioned Area of ​​Interest and HU Quality Analysis Engine (13) automatically determines the area of ​​interest. by doing this, the Hounsfield Unit values ​​and signal-to-noise ratio in the relevant anatomical regions. It performs measurements and thus obtains the previously applied contrast parameters. It makes a match between the actual image quality and the image quality. Then, Central Adaptive Learning Engine (14), the aforementioned Clinical Data Integration Module (11) and the aforementioned Area of ​​Interest and HU 15 It evaluates clinical and image data provided by the Quality Analysis Engine (13) together. During the evaluation, the aforementioned Central Adaptive Learning Engine (14) was used. The Target HU Prediction Engine (141) is activated, taking into account the anatomical region, phase type and clinical when estimating the ideal Hounsfield Unit range to be reached according to the indication, again the same The Similar Patient Matching Module (142) located inside the engine matches 20 clinical features. An initial optimization based on successful contrast protocols for patients with this condition. It performs. The optimum produced by the Central Adaptive Learning Engine (14) mentioned above. contrast plan (volume, injection rate, timing, and phase delay), Injector The contrast is transferred to the injection system via the Integration Layer (17) and the shot is performed. After the shooting is complete, the Automatic Phase Success Scoring Module 25 (15) by objectively evaluating whether the targeted contrast enhancement criteria have been achieved It produces a numerical phase success score. Finally, the Closed Loop Adaptive Learning Engine (16), quality metrics produced by the aforementioned Automatic Phase Success Scoring Module (15) and It takes the mentioned phase success score as feedback, updates the model parameters, and this updates mentioned Central Adaptive Learning Engine (14) and Target HU 30 mentioned By passing the Prediction Engine (141) to the aforementioned Similar Patient Matching Module (142), then This enables the creation of more optimized dosage plans for patients. In this way, The aforementioned Clinical Decision Support System (1) constantly improves itself by learning from each shot result. It creates a closed-loop adaptive optimization mechanism that improves and targets It minimizes phase failures by increasing the rate at which the Hounsfield Unit value is reached. 35

Claims

16 REQUESTS 1. Closed-loop adaptive learning-based clinical decision support integrated with HIS and PAC systems. It is a system and its feature is,  Contrast agent in computed tomography and magnetic resonance imaging examinations 5 for personalized and adaptive optimization of injection parameters A Clinical Decision Support System based on closed-loop adaptive learning (1),  Within the aforementioned Clinical Decision Support System (1) and belonging to the patient Automatically retrieve demographic information, clinical indications, and laboratory results. a Clinical Data Integration Module (11),  Within the aforementioned Clinical Decision Support System (1) and the aforementioned patient 10 Analyzing DICOM format image series from past imaging studies. Image Archiving System Analysis Module (12),  The aforementioned Image is included within the Clinical Decision Support System (1) mentioned above. On the images analyzed by the Archive System Analysis Module (12) By automatically determining the area of ​​interest, Hounsfield Unit and signal-to-noise ratio measurements are obtained. An Area of ​​Interest and HU Quality Analysis Engine (13) that performs  Clinical Data included within the aforementioned Clinical Decision Support System (1) The Integration Module (11) and the mentioned Area of ​​Interest and HU Quality Analysis Engine (13) A Central Adaptive that evaluates clinical and imaging data provided by [the organization / institution] together. Learning Engine (14), 20  Located within the aforementioned Central Adaptive Learning Engine (14) and anatomical the ideal range of Hounsfield Units to be reached according to region, phase type and clinical indication A Target HU Prediction Engine (141) that predicts,  Located within the aforementioned Central Adaptive Learning Engine (14) and similar Starting point 25, referencing successful contrast protocols for patients with clinical characteristics. A Similar Patient Matching Module (142) that performs optimization,  The shooting included in the aforementioned Clinical Decision Support System (1) after completion, whether the targeted contrast enhancement criteria have been achieved. An Automated Phase system that generates a numerical phase success score by objectively evaluating it. Success Scoring Module (15), 30  The mentioned Automatic system, which is included within the aforementioned Clinical Decision Support System (1) Quality metrics produced by the Phase Success Scoring Module (15) and the mentioned phase By receiving the success score as feedback, it updates the model parameters and then... A closed-loop adaptive learning system that generates optimized dosage plans for patients. Engine (16) and 35 17  The Central Office mentioned within the Clinical Decision Support System (1) mentioned The optimum contrast plan generated by the Adaptive Learning Engine (14) contrast An Injector Integration Layer (17) to transfer to the injection system It is characterized by its inclusion.

Images