Chronic Kidney Disease (CKD) Documentation

Inspiration

Managing CKD effectively requires continuous monitoring and personalized treatment recommendations. Our goal is to use AI to provide tailored recommendations and actionable insights for patients based on their specific data, helping improve their quality of life and health outcomes.

What it does

The Kidney Care AI system collects patient data and clinical notes, processes and analyzes this information using machine learning models, and provides personalized recommendations for managing CKD. The system also integrates with the RAG (Red, Amber, Green) model to categorize patient progress and suggest appropriate actions based on the severity of the disease.

How we built it

Data Collection and Ingestion:
- Load patient data and clinical notes from PDF documents using PyPDFDirectoryLoader.
Data Preprocessing:
- Split documents into manageable chunks using RecursiveCharacterTextSplitter.
Embedding Generation and Storage:
- Convert text data into vectors for similarity search using GoogleGenerativeAIEmbeddings.
- Store embeddings for efficient retrieval using FAISS.
Model Training and Predictions:
- Generate patient summaries and recommendations using ChatGroq and ChatPromptTemplate.
Recommendation Logic:
- Implement logic to generate recommendations based on predicted patient progress.
RAG Model Integration:
- Combine retriever and generator for relevant document context and accurate recommendations using create_retrieval_chain.
Deployment and User Interaction:
- Create an interactive web application using Streamlit.

Challenges we ran into

Ensuring the accuracy of recommendations based on limited patient data.
Integrating various components (data ingestion, preprocessing, embedding generation, and model training) seamlessly.
Handling and processing large volumes of text data efficiently.

Accomplishments that we're proud of

Successfully integrating multiple AI technologies to create a comprehensive system for CKD management.
Developing a user-friendly interface for patients and healthcare providers to interact with the system.
Achieving meaningful and actionable recommendations that can positively impact patient care.

What we learned

The importance of high-quality data for training machine learning models.
Techniques for efficiently processing and embedding large text datasets.
Best practices for integrating AI models into interactive applications.

Introduction

Chronic Kidney Disease (CKD) is a medical condition characterized by the gradual loss of kidney function over time. The kidneys are essential for filtering waste and excess fluids from the blood, which are then excreted in the urine. When CKD occurs, the kidneys are damaged and unable to perform this filtering process effectively, leading to the build-up of waste products in the body.

CKD is often progressive and can lead to end-stage renal disease (ESRD), which requires dialysis or a kidney transplant to maintain life. The stages of CKD are determined based on the glomerular filtration rate (GFR), which measures how well the kidneys are cleaning the blood.

Key Points about CKD

Risk Factors
- Diabetes: The most common cause of CKD. High blood sugar levels can damage the blood vessels in the kidneys.
- High Blood Pressure: Can damage the small blood vessels in the kidneys, affecting their ability to filter waste.
- Other Risk Factors: Heart disease, obesity, family history of kidney failure, and age over 60.
Symptoms
- Early stages of CKD typically have no symptoms.
- As the disease progresses, symptoms may include:
  - Swelling in the ankles and feet
  - Fatigue
  - Difficulty concentrating
  - Decreased appetite
  - Changes in urine output
Diagnosis
- Blood Tests: Measure GFR to assess how well the kidneys are cleaning the blood.
- Urine Tests: Look for protein or blood in the urine, which can indicate kidney damage.
- Imaging Tests: Assess kidney structure.
Treatment
- There is no cure for CKD, but treatment can slow its progression.
- Managing Underlying Conditions: Such as diabetes and hypertension.
- Lifestyle Changes: Diet, exercise, and avoiding nephrotoxic medications.

Recommendations for Managing CKD

Serious Condition (End-Stage Renal Disease - ESRD)

Immediate Actions:
- Consult your doctor immediately.
- Prepare for dialysis or consider a kidney transplant.
- Adhere strictly to prescribed medications and treatments.
Lifestyle Modifications:
- Follow a kidney-friendly diet low in sodium, potassium, and phosphorus.
- Monitor fluid intake to avoid volume overload.
Monitoring:
- Regular visits to a nephrologist.
- Frequent blood and urine tests to monitor kidney function.

Worsening Condition (Stage 3-4 CKD)

Medical Management:
- Increase the dose of medications like frusemide as advised by your doctor.
- Monitor and manage blood pressure and blood sugar levels closely.
Lifestyle Modifications:
- Reduce protein intake to lessen kidney workload.
- Decrease fluid intake by 300 ml to avoid volume overload.
Monitoring:
- Regular monitoring of kidney function.
- Watch for symptoms such as swelling, fatigue, and changes in urine output.

Stable Condition (Stage 1-2 CKD)

Medical Management:
- Continue taking prescribed medications on time.
- Manage any underlying conditions like diabetes and hypertension.
Lifestyle Modifications:
- Maintain a healthy diet low in salt and fat.
- Regular exercise to maintain overall health.
Monitoring:
- Regular check-ups with your healthcare provider.
- Periodic blood and urine tests to monitor kidney function.

Patient Data Analysis and Recommendations

Patient Data Required

To provide personalized recommendations for managing CKD, we need to gather specific patient data. Based on this data, a summary and tailored recommendations will be generated. The following patient data is required:

Age
Weight
Medical conditions
Heart rate
SpO2 levels
Blood pressure (systolic)
Ankle swelling
Breathlessness
Current medications
Average difference in key metrics
Progress
Progress category
Predicted progress

Built With

bedrock
gemini
grok
langchain
python
rag