Heart disease prediction system on the base of numeric values(lik age,sex,chol,BP,etc) and images like CT scan and MRI

Title: Predicting Heart Disease Using Image and Numeric Data: Is it Possible?

Description: Hello everyone,

I’m currently working on a heart disease prediction system and I have a question regarding combining different types of data. Specifically, I’m interested in whether it’s possible to predict heart disease based on both medical images (such as X-rays or MRIs) and numeric/tabular data (like age, gender, cholesterol levels, blood pressure, etc.).

Has anyone worked on a similar project where they used both image processing and numeric feature analysis in the same model? What techniques or models would you recommend for handling both types of data effectively?

I would appreciate any insights or suggestions from this community. Also, if anyone has come across datasets that combine both medical images and clinical numeric data, please let me know.

Thank you!

Tags: Heart Disease Prediction, Medical Image Processing, Multimodal Data, Supervised Learning, Healthcare AI

Language: Python, TensorFlow

Product: Heart Disease Prediction Model

1 Like

Hi @Imran_Rafique

Try Kaggle for Heart disease prediction, one I am sharing here

you are missing one important way to check how blood flow is checked in heart, i.e. ECG electrocardiogram which would be one of the most important feature to use in your prediction other than images, lab diagnosis such as checking correlation to cholesterol level (HDL, LDL, VLDL) including other features you mentioned.

Heart disease also has relation to metabolic activity and other comorbidities like hypertension, hyperglycemia and cofactor like smoking and drinking habits, obese.

Regards
DP

Use of Multi-Modal Data and Machine Learning to Improve Cardiovascular Disease Care

Sources of data for digital health tools include multiple modalities such as electronic medical records (EMR), radiology images, and genetic repositories, to name a few. While historically, these data were utilized in silos, new machine learning (ML) and deep learning (DL) technologies enable the integration of these data sources to produce multi-modal insights. Data fusion, which integrates data from multiple modalities using ML and DL techniques, has been of growing interest in its application to medicine. In this paper, we review the state-of-the-art research that focuses on how the latest techniques in data fusion are providing scientific and clinical insights specific to the field of cardiovascular medicine.

2 Likes

Sir i am not asking about what are the clinical symptoms my purpose of asking question is is it possible to prepare model or does there is any data exist where the numeric values of symptoms and image like mri ctscan exist with the help of that we can predic heart diseas prediction

@Imran_Rafique I already shared the data. the reason of providing those details was to use that information when you create your prediction model. ofcourse I know you didn’t ask for symptoms but you looking for prediction model with mri/ctscan misses that most important feature to include in prediction model is to including ECG too.

the link shared gives you dataset to create your model. you can search for dataset or model on kaggle.

1 Like

Sir can you provide me dataset where it is describe on the the base of numeric values and image data that heart problem occur or not

Search on Kaggle you should surely find there the dataset and model you are looking for

Data i want isn’t avaliable.

1 Like

You need to search imran, it is there on kaggle. when you look into the various models, in dataset section you will find it.

I am sharing one more link of dataset prediction model from GitHub

I am very thankful to you on your reply
i know about models algorithms i just need data set where numeric values data as you have sent me and image data like MRI CT scan both are given and there out put is given does the the heart ptoblem is ocur or not

in the same link, go through metadata, you wil find your required dataset file named dataset.csv

Is that the dataset we’re talking about here? Because I don’t see any unstructured data in there, either. Something I missed?

1 Like

Which column is unstructured?

this is the available structed data in the link GitHub repo.

he asked dataset, not in particular of unstructured dataset and wanted a similar model prediction, so had shared.

I believe the answer to this question is clearly yes, and you can find examples of related research in the public literature. EG…

I suggest read as many of those types of papers as you can find. Look at their references and what they used as data sources. Consider reaching out directly if there is something similar enough to what you want to do. Key ideas will be fusion and ensemble. You probably already realize that personally identifiable information complicates your task considerably. Are you able to partner locally with an academic healthcare organization? Hope this helps a little. Let us know what you find?

1 Like

Can you share this data set

You may be pretty hard pressed to find a quality multimodal data set with clinical characteristics included, mainly due to regulations on the dissemination of medical info.

One of the better ones I’ve seen is AIREADI, though the imaging data available is ophthalmic and the subjects are all diabetic. It does however, include a trove of modalities from ECG and imaging to clinical tabular.

h ttps://docs.aireadi.org/docs/1/about

Another dataset you might look at is the PBT XL one put out a few years back. There isn’t any imaging, but you’ll have 12-lead ECG data along with some clinical/demographic info on the patients. Could try something like a transformer with a fusion block somewhere in it. Download is at the bottom:

h ttps://physionet.org/content/ptb-xl-plus/1.0.1/

I can’t post links for some reason so I added a space between the first letter and rest of it.

2 Likes

That comes with increased trust level through continued use of the forum.

And Welcome to the community!

1 Like

Or, as @Deepti_Prasad correctly suggests above you need to search

Using the approach suggested above, namely reading the research papers and looking for the datasets they mention, I found this in about 4 minutes…

https://stanfordaimi.azurewebsites.net/datasets/3263e34a-252e-460f-8f63-d585a9bfecfc

OL3I (pronounced olé) is a multimodal dataset of 8,139 axial computed tomography (CT) slices at the third lumbar vertebrae (L3) level of individuals along with tabular medical record data from up to one year prior to the scan. In addition, labels are provided to indicate individuals that were diagnosed with ischemic heart disease 1 or 5-years after the scan. To our knowledge, this is the first publicly available multimodal dataset. for opportunistic risk assessment following an abdominopelvic CT scan. The contrast-enhanced abdominopelvic CT images were acquired in patients who visited the emergency department of Stanford Hospital and Clinics between 2013-2018.

my emphasis added

Seems worth a look