File size: 2,750 Bytes
87bbe58
0ff826a
 
 
 
87bbe58
8f61851
87bbe58
 
0ff826a
87bbe58
 
0ff826a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8f61851
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
---
title: CKD Prediction System
emoji: 🫘
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 5.23.3
app_file: app.py
pinned: false
license: mit
---

# Chronic Kidney Disease (CKD) Prediction System

**Research Project | PAF-IAST | HEC-Funded**  
**Principal Investigator:** Dr. Muhammad Rizwan

---

## Overview

This web application provides an AI-powered screening tool for **Chronic Kidney Disease (CKD)** 
based on routine blood test parameters. It was developed as part of a peer-reviewed research study 
conducted at the **Pakistan Air Force Institute of Aerospace Technology (PAF-IAST)** under an 
HEC-funded project.

The application accepts 21 blood parameters as input and returns a binary prediction 
(**CKD / No CKD**) along with class probabilities.

---

## Model

| Property | Details |
|---|---|
| Algorithm | Random Forest Classifier |
| Accuracy | ~95.5% |
| F1 Score | ~95.4% |
| ROC-AUC | ~98.5% |
| Validation | 5-fold Stratified Cross-Validation |
| Explainability | SHAP + LIME |

The Random Forest model was selected after benchmarking 8 classifiers including 
XGBoost, CatBoost, Gradient Boosting, AdaBoost, Logistic Regression, Decision Tree, 
and K-Nearest Neighbors.

---

## Input Features

The model uses the following 21 blood test parameters:

**Demographic:** Sex, Age  
**CBC:** Haemoglobin, Hematocrit, RBC Count, WBC Count, Platelet Count  
**Red Cell Indices:** MCH, MCHC, MCV, RDW-SD, MPV, PCT, PDW  
**Differential Count:** Lymphocytes, Monocytes, Polymorphs, Eosinophils  
**Kidney Function:** Creatinine, Blood Urea, GFR  

---

## Key Findings (from SHAP Analysis)

Top predictive features (in order of importance):
1. **Creatinine** β€” strongest CKD indicator
2. **GFR** (Glomerular Filtration Rate) β€” inverse relationship with CKD
3. **Urea** β€” elevated in kidney dysfunction
4. **RDW-SD** β€” red cell distribution width
5. **MPV** β€” mean platelet volume

---

## Dataset

- **Samples:** 398 patients
- **Class distribution:** CKD: 241 (60.6%) | No CKD: 157 (39.4%)
- **Missing values:** < 1.5% (imputed with zero-fill)
- **No duplicates**

---

## Files in this Space

```
app.py               ← Gradio application
requirements.txt     ← Python dependencies
Random_Forest.pth    ← Trained Random Forest model (joblib format)
README.md            ← This file
```

---

## Disclaimer

> This application is intended **for research and educational purposes only**.  
> It does **not** constitute medical advice, diagnosis, or treatment.  
> All clinical decisions must be made by a licensed and qualified physician.

---

## Citation

If you use this work, please cite the associated research paper (details to be added upon publication).

**Contact:** Dr. Muhammad Rizwan | PAF-IAST