Scribd
Upload
91 views14 pages

Diabetes Heart Disease Prediction

The document presents a project on predicting diabetes and heart disease using machine learning models, focusing on early detection through a fusion of image-based and tabular datasets. It outlines the proposed system architecture, which includes a CNN for image processing and an MLP for tabular data, achieving over 90% accuracy. The project emphasizes its relevance to public health and the potential for real-time clinical deployment.

Uploaded by

Akshay Tiwari
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
91 views14 pages

Diabetes Heart Disease Prediction

The document presents a project on predicting diabetes and heart disease using machine learning models, focusing on early detection through a fusion of image-based and tabular datasets. It outlines the proposed system architecture, which includes a CNN for image processing and an MLP for tabular data, achieving over 90% accuracy. The project emphasizes its relevance to public health and the potential for real-time clinical deployment.

Uploaded by

Akshay Tiwari
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 14

Diabetes and Heart Disease

Prediction using Machine


Learning Models
Course Code: XX123 | Team
Members: Akshay Tiwari
(21BCEXXXX) | Guide: Dr. XYZ
Outlines / Content
• 1. Introduction
• 2. Literature Review
• 3. Problem Formulation
• 4. SDG Relevance
• 5. Proposed System Architecture
• 6. Analytical and Theoretical Description
• 7. Tools and Parameters
• 8. Results Analysis
• 9. Conclusion
Introduction
• Exploring diabetes and heart disease risks
using AI/ML models.
• Multi-modal datasets used: image-based and
tabular.
• Objective: Early prediction and risk detection.
Literature Review
• ML for diabetes prediction using tabular
datasets.
• CNNs for retinal image classification in
diabetic retinopathy.
• Fusion-based models show improved
diagnostic accuracy.
Problem Formulation
• Challenge: Integrating image and tabular data.
• Goal: Improve prediction accuracy and
robustness.
• Scope: Fusion model for binary classification
(disease/no disease).
SDG Relevance
• Supports UN SDG 3: Good Health and Well-
being.
• Early disease detection reduces health costs
and mortality.
• Promotes AI-driven medical diagnosis in
underserved areas.
Proposed System / Architecture
• Three-part model: CNN for images, MLP for
tabular, Fusion layer.
• EfficientNetB3 extracts retinal features (512-
dim).
• MLP extracts tabular features (512-dim).
• Fusion: Concatenation + multiplication →
1536-dim vector → FC layer.
Analytical and Theoretical
Description
• CNN: Pretrained EfficientNetB3 with custom
FC head.
• MLP: Input → Hidden → ReLU → 512-dim
output.
• Fusion: Feature engineering with element-
wise ops + FC classification.
Tools & Simulation/Design
Parameters
• Python, PyTorch, Google Colab, VS Code.
• Libraries: torchvision, scikit-learn, matplotlib.
• Dataset sizes: PIMA (768 samples),
Retinopathy (~35k images), Cleveland (~300
samples).
Results Analysis
• Fusion model accuracy: 90%+
• EfficientNetB3 + MLP fusion outperformed
individual models.
• Confusion matrices, AUC scores, and accuracy
plots used.
Conclusion
• Fusion model effectively combines image and
tabular data.
• Promising tool for early prediction of chronic
diseases.
• Scope for real-time clinical deployment with
minimal tuning.
Contribution of Team Members
• Akshay Tiwari: MLP and CNN Implementation,
Fusion model, Web Deployment.
• Team Member 2: Dataset cleaning,
Exploratory Analysis, Model Evaluation.
• Team Member 3: Report writing, Presentation
Design, Literature Review.
Impact on Society and
Environment
• Promotes low-cost, accessible healthcare
screening.
• Reduces burden on healthcare infrastructure
via automation.
• Environmentally sustainable – minimal
hardware use.
References
• [1] UCI ML Repository – PIMA and Cleveland
datasets.
• [2] Kaggle – Diabetic Retinopathy Dataset.
• [3] Tan, M. & Le, Q. EfficientNet: Rethinking
Model Scaling for CNNs.
• [4] Recent ML papers on medical fusion
models.

You might also like