Description:FIELD OF THE INVENTION
The present invention relates to the field of medical diagnostics and artificial intelligence (AI). More specifically, it concerns a deep learning-based system and method for long-term prediction of Chronic Kidney Disease (CKD) progression and early detection of secondary infections. The invention integrates convolutional neural networks (CNNs), recurrent neural networks such as long short-term memory (LSTM) and bidirectional LSTM (BLSTM), along with ensemble learning, to provide accurate, interpretable, and scalable predictions from multimodal medical data including laboratory reports, electronic health records, and medical imaging.
BACKGROUND OF THE INVENTION
Chronic Kidney Disease (CKD) is a significant worldwide public health concern, affecting millions of people across the globe, often culminating in disabilities and increased mortality and morbidity including End-Stage Renal Disease (ESRD). Understanding long-term progression of CKD still presents a major challenge, despite investments in medical technology. In contrast, many current models are limited to follow-ups in the short-term, which is inadequate to elucidate the long-term implications of disease evolution. In addition, current methods are frequently unable to identify secondary causes of CKD like albuminuria, which may have an important influence on treatment plans and patient outcomes. Delays in understanding the long-term effects of CKD and early detection of secondary infections can lead to delays in intervention, potentially higher healthcare costs and worse health outcomes. Thus, in order to improve CKD management and patient care, a more successful scope that works on combining long-term prediction approaches and earlier synthesis contamination identification is fundamental.
US20200005900A1: This is a machine learning based system that, by organization and analysis of the values of the estimated glomerular filtration rate (eGFR) and other biomarkers within 30-180 days, predicts the risk of CKD progression. It applies multiple ML algorithms to measure the deterioration of renal functionality. Google Patents
US20230054069A1: Proposed method of generating predictions of CKD progression using machine learning model trained on medical laboratory data, including eGFR, urine albumin-to-creatinine ratio (ACR), and other clinical parameters. The model estimates CKD progression within defined time intervals.
US9689826B2: The invention is a sensor system that utilizes gold nanoparticles that are coated with organic materials that have the property that they will bind to VOCs that signal stage 1 CKD. The ultimate goal is large-scale screening in generalist clinics.
US20210241918A1: A computer implemented method of predicting the health and therapeutic behavior of patients and making treatment plan recommendations includes the steps of: receiving patient healthcare data having one or more conditions and limiting factors; determining a therapeutic behavior pattern of patient; determining unsuccessful therapies and successful therapies for each condition based on therapeutic behavior pattern; and calculating cost quote for successful therapies based on limiting factors for time period. A computer implemented method of providing cost effective therapy for a patient is also provided and includes the steps of: receiving patient healthcare data; determining unsuccessful therapies and successful therapies; calculating probability of disease progression; calculating possible therapies ranked by probability of successful treatment; calculate cost quote for possible therapies; and paying a smart contract for a selected therapy.
US2023014055A1: The disclosure relates to in vivo and ex vivo uses of dihydronicotinamide riboside (NRH), dihydronicotinic acid riboside (NARH) and reduced derivatives thereof to treat immune-related disorders (e.g., systemic inflammatory response syndrome and sepsis), kidney disorders (e.g., acute kidney injury and hepatorenal syndrome [HRS]), liver disorders (e.g., acute liver failure and HRS), hemolytic disorders (e.g., hemolysis and hemolytic anemia), and disorders and conditions associated with oxidative stress, damage or injury (e.g., methemoglobinemia and anemia). NRH, NARH and reduced derivatives thereof can be used in vivo or ex vivo alone or in combination with one or more additional therapeutic agents, such as an anti-inflammatory agent or/and an antioxidant.
Chronic Kidney Disease is a major global health burden with millions of patients at risk of progressing to End-Stage Renal Disease. Current diagnostic and predictive models are primarily short-term in nature and fail to provide accurate insights into long-term progression. Moreover, existing systems often overlook secondary infections such as albuminuria, which significantly influence patient outcomes and treatment planning. The inability to simultaneously predict long-term disease trajectories and detect secondary infections leads to delayed interventions, higher healthcare costs, and poor patient management.
The present invention addresses these shortcomings by offering a comprehensive deep learning framework capable of modeling CKD progression over extended horizons while simultaneously detecting secondary infections. The system integrates multi-modal data sources and employs an ensemble of CNN, LSTM, and BLSTM models combined with an ensemble module to improve predictive accuracy. This dual capability provides clinicians with actionable insights for proactive intervention, thereby improving patient care, reducing complications, and enabling personalized treatment planning.
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 invention provides a comprehensive AI-driven system for long-term CKD prediction and secondary infection detection. The system incorporates data preprocessing, feature selection, deep learning modeling, ensemble prediction, and explainable outputs. It is designed to integrate clinical, laboratory, and imaging data for holistic evaluation of patient health.
The data preprocessing module cleans, normalizes, and formats multimodal patient information, handling missing values using techniques such as interpolation and imputation. A feature selection component identifies critical attributes including serum creatinine, blood pressure, urine albumin levels, and demographic information. These features are then processed by an ensemble of deep learning models: a CNN for hierarchical feature representation, an LSTM for capturing temporal dependencies, and a BLSTM for modeling both past and future dependencies in sequential data.
Outputs from these models are combined through an ensemble module that applies majority voting or weighted averaging to enhance accuracy and robustness. A dedicated secondary infection detection module monitors laboratory and imaging results to identify early signs of complications such as albuminuria. The system also incorporates explainable AI methods that provide clinicians with interpretable justifications for predictions, improving trust and clinical adoption.
By integrating long-term prediction with early infection detection, the invention supports clinical decision-making, remote patient monitoring, and healthcare research. It is scalable, adaptable to large datasets, and capable of delivering accurate and proactive insights into CKD management.
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.
This invention suggests a novel deep-learning framework that aims to improve CKD (chronic kidney disease) progression prediction over a long duration and similar occurrence of secondary infections. The system combines several deep learning algorithms, such as CNN, Long Short Term Memory (LSTM) network and deep ensemble model, to produce accurate predictions and facilitates early action in response to its forecasts, when possible, to help medics to intervene timely to improve patients’ outcome.
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:
FIGURE 1: SYSTEM ARCHITECTURE
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.
The invention provides a comprehensive deep learning framework designed to predict long-term CKD progression and detect secondary infections. The system operates on multimodal data inputs, which may include laboratory test results, electronic health records, demographic data, and medical imaging. To ensure consistency, the preprocessing module standardizes input formats, applies normalization techniques, and addresses missing values through mean imputation or interpolation.
The feature selection module identifies the most relevant attributes influencing CKD progression and infection risk. Techniques such as recursive feature elimination and mutual information analysis are employed to ensure only the most significant variables are used for modeling. Key features include estimated glomerular filtration rate, urine albumin-to-creatinine ratio, blood pressure, and serum creatinine levels.
Once data is prepared, it is fed into multiple deep learning models. The CNN model captures hierarchical structures in data, identifying complex spatial and contextual patterns linked to CKD progression. The LSTM model processes sequential health records, learning temporal dependencies that reflect gradual disease changes. The BLSTM extends this capability by incorporating both past and future temporal information, thereby improving prediction accuracy.
The outputs of these models are integrated into an ensemble learning module. This module aggregates predictions using majority voting or weighted averaging, combining the strengths of individual models to produce more reliable and accurate long-term predictions. The ensemble approach improves generalizability across diverse patient populations and datasets.
A specialized secondary infection detection module forms a critical component of the system. This module analyzes laboratory and imaging data for signs of complications such as albuminuria or urinary tract infections that commonly co-occur with CKD. Early detection of such infections allows for timely intervention and adjustment of treatment strategies, reducing risks of deterioration.
The system also incorporates explainable AI methods. Using interpretability techniques such as SHAP values, the model provides justifications for its predictions, highlighting which features most strongly influenced outcomes. This transparency builds trust among clinicians and enables more informed decision-making.
The architecture is designed to handle real-world clinical deployment. It is scalable to large datasets and supports real-time inference, allowing integration into hospital systems and electronic health records platforms. Remote monitoring capabilities extend its use to outpatient and telemedicine settings, where continuous analysis of patient data enables proactive care.
Validation of the system is achieved through multi-site clinical datasets, ensuring generalizability across patient demographics and healthcare environments. Performance is assessed using accuracy, sensitivity, specificity, and AUC metrics, demonstrating superior predictive power compared to existing single-model approaches.
In operation, the system reduces diagnostic delays by providing long-term risk predictions and identifying secondary infections early. By offering a comprehensive, integrated, and explainable framework, it bridges critical gaps in CKD management technology and provides a transformative tool for personalized healthcare.
This invention suggests a novel deep-learning framework that aims to improve CKD (chronic kidney disease) progression prediction over a long duration and similar occurrence of secondary infections. The system combines several deep learning algorithms, such as CNN, Long Short Term Memory (LSTM) network and deep ensemble model, to produce accurate predictions and facilitates early action in response to its forecasts, when possible, to help medics to intervene timely to improve patients’ outcome.
(a) System Overview: The system consists of a data preprocessing module, a feature selection component, multiple deep learning models including CNN, LSTM, and LSTM-BLSTM, an ensemble model for combining predictions, and a secondary infection detection module. It is intended to handle electronic health records (EHR), laboratory test results, and medical imaging data to forecast progression of CKD and detect secondary infections.
b) Data Collection and Preprocessing: Data are collected from various sources, Colecte EMR, Med-trace, Lab databases etc. The preprocessing module comes in place to clean the data, normalize it and make it to a format which will be similar for all. Missing values are handled using mean imputation, or interpolation, etc.
c) Feature Selection: Relevant features are selected using techniques such as Recursive Feature Elimination (RFE) and mutual information analysis. Key features include serum creatinine, blood pressure, urine albumin, and demographic information. Such features are very important in accurately predicting CKD and detecting infection.
d) Deep Learning Models:
CNN Model employs convolutional layers to learn hierarchical feature representations of structured data, accommodating for intricate patterns that represent CKD progression.LSTM model processes sequential data and captures temporal dependencies, which is important for modeling the gradual progression of CKD over time.This model is an extension of LSTM, consisting of LSTM layers integrated with Bidirectional LSTM layers to capture both past and future dependencies in sequential data and improve prediction accuracy. Use the majority voting technique to ensemble the outputs of the individual deep learning models. This combines each model's strengths to enhance prediction accuracy. The best accuracy for predicting CKD progression in 6- to 12-months demonstrated by the ensemble model.A secondary infection detection module is added to the framework for the detection of infections such as albuminuria and other related CKD complications. It helps to Monitor Laboratory Findings and Imaging Results to identify early warnings of a Secondary infection to be taken to immediate medical care. Models are evaluated and validated using the metrics, accuracy, sensitivity, specificity, and Area Under the Curve (AUC). Model robustness is evaluated using cross-validation methods. The system was validated through multiple public datasets to establish generalizability across different population types.
Best Method of Working
The best method of working involves training the framework on a large multimodal dataset containing laboratory values, medical imaging, and patient demographics. Preprocessing ensures consistency and quality of data, while feature selection identifies the most predictive parameters. The CNN, LSTM, and BLSTM models are trained in parallel, with outputs combined through the ensemble module. The secondary infection detection module continuously monitors inputs for early signs of complications. For deployment, the system is integrated with hospital electronic health records, enabling real-time patient monitoring. Predictions and infection alerts are delivered through a clinician interface that includes interpretability features, ensuring transparency. This embodiment provides the most effective and clinically valuable application of the invention.
ADVANTAGES OF THE INVENTION
Enables predictions of CKD progression over long time horizons, allowing for proactive management of at-risk individuals.Early infection detection detects the indication of secondary infection early, allowing for timely treatment and reducing the risk of complications.Integrates various sources of data and deep learning models to provide a significant view of patient health.Can be scaled to manage extensive datasets, allowing implementation in a variety of healthcare environments.
Potential Applications:
Clinical decision support: Provides healthcare providers with programs and clinical alert aim to make informed decisions on CKD management and treatment.
Remote Monitoring: Allows for continuous monitoring of patients with CKD, enabling early intervention and personalized care
Healthcare Research: Offers a tool for researchers to investigate CKD progression and secondary infection outcomes. This feat is a huge feat for CKD technology innovation by suggesting the methods of deep learning for long-term prediction and secondary infection study. This has the potential for a positive outcome for the patient by intervention at the right time and personalized approach towards the treatment strategy.
This invention proposes a Novel Predictive Methodology based on Comprehensive Deep Learning Framework with Multi-Model for Long-Term Prediction of Chronic Kidney Disease (CKD) with Secondary Infections Detection. This is a first in several important respects:
a) Combining Multiple Deep Learning Models: Conventional modalities heavily rely on a single machine learning model, as opposed to this state-of-the-art invention that illustrates combining Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTMs), and Bidirectional LSTM (BLSTM) networks to encode the spatial and temporal correlations of CKD progression.
b) Ensemble Learning for Higher Prediction Accuracy: Implementation of an ensemble model, combining the predictions from several deep learning models for better accuracy and robustness, is a novel approach and not usually utilized with present CKD prediction systems.
c) Secondary Infection Detector: Building a module to detect secondary infections linked to CKD, including those with albuminuria, introduces an additional feature to the framework, filling a gap in existing technologies.
d) Climate of Multi-Modal Data: An extremely innovative aspect of the framework is, extensive usage of processing and integrating multi-modal data, EHRs, lab examinations and medical imaging, towards holistic CKD management.
e) Long-Term Prediction: To the best of our knowledge, all existing models limit prediction to the short-term; this invention focuses on prediction over an extended period which may yield more meaningful changes to CKD progression to facilitate better patient management.
Together these new attributes provide a broad and intelligent solution for the prediction and management of CKD, which is unprecedented among other existing technologies.
 
, Claims:1. A system for long-term prediction and secondary infection detection in Chronic Kidney Disease (CKD), comprising:
a data preprocessing module configured to clean, normalize, and format multimodal medical data;
a feature selection component adapted to identify relevant attributes including laboratory values, imaging data, and demographic information;
a convolutional neural network (CNN) model configured to extract hierarchical feature representations;
a long short-term memory (LSTM) model configured to capture temporal dependencies in sequential data;
a bidirectional long short-term memory (BLSTM) model configured to capture both past and future dependencies;
an ensemble module configured to combine outputs of the CNN, LSTM, and BLSTM models to improve prediction accuracy;
a secondary infection detection module configured to identify complications such as albuminuria from laboratory and imaging data; and
an output interface for providing long-term CKD progression predictions and infection alerts.
2. The system as claimed in claim 1, wherein the data preprocessing module handles missing values using interpolation or imputation techniques.
3. The system as claimed in claim 1, wherein the feature selection component employs recursive feature elimination or mutual information analysis.
4. The system as claimed in claim 1, wherein the ensemble module applies majority voting or weighted averaging to aggregate model outputs.
5. The system as claimed in claim 1, wherein the secondary infection detection module continuously monitors laboratory and imaging data for early warning signs.
6. The system as claimed in claim 1, wherein the output interface includes explainable AI features that highlight influential parameters for predictions.
7. The system as claimed in claim 1, wherein the system integrates with hospital electronic health records for real-time clinical deployment.
8. The system as claimed in claim 1, wherein the framework is scalable to large datasets across multiple healthcare environments.
9. The system as claimed in claim 1, wherein the system supports remote patient monitoring and telemedicine applications.
10. A method for long-term prediction and secondary infection detection in Chronic Kidney Disease (CKD), comprising the steps of:
collecting and preprocessing multimodal medical data;
selecting relevant features from the data using feature selection techniques;
processing the features using CNN, LSTM, and BLSTM models to generate predictive outputs;
combining the predictive outputs using an ensemble module to improve accuracy;
detecting secondary infections using a dedicated module analyzing laboratory and imaging data; and
providing predictions and alerts through an output interface with explainable AI features.