Apps iMed Web Application User Manual
- June 13, 2024
- APPs
Table of Contents
iMed User Manual
Introduction
1.1. Purpose
The purpose of this web application is to take raw information and allow
manipulating it in a manner that gives results useful in decision making. This
can be to train a model with raw data or predicting the outcome using models
and analysis.
1.2. Navigational Menu
The navigational menu on the top of the page holds all the links to get to
where you need to be. If you ever get lost, you can always click the back
arrow to get to a familiar page, go back home, or find the page you are
looking for within the navigational menu.
1.3. Account
If you don’t already have an account, you must register to use the
application. To do so, click the account button on the top right and click
register. Then enter your username, password, and email to proceed.
If you already have an account, sign in with your username and password.
Home Page
By clicking on the items on the left of the page, a description of each will appear in the middle of the page to help you understand what each one does.
iMedBot
The iMedBot application presents an interface that fosters easy user interaction with agents, enabling personalized prediction and model training. It serves as the first step towards transforming the outcomes of deep learning research into an online tool, which has the potential to spark additional research pursuits in this domain. Its respective user manual can be found here.
Data Analysis
4.1. Retrieve Subsets
This section lets the user edit their dataset. You can choose to either upload
a new dataset or use an existing one from the drop-down menu.
Once the dataset has been uploaded, you can choose what action you would like
to take by clicking one of the options on the left side menu.
4.1.1. Retrieve Subsets Based on Filters
This section allows getting a smaller subset of the original dataset based on
given filters. Choose the values that you want in the subset and then choose
the columns that you want shown in the final dataset.
4.1.2. Return Sorted Results
This returns the dataset in a sorted form. Choose the target column, sorting
order, number of rows to return, and which columns to show in the final
output.
4.1.3. Expand the Dataset
This allows the user to expand a singular column stored as a dictionary into
an actual table that the user can then manipulate. It takes a nested dataset
and moves what is required by the user into the top-most layer. First, upload
a dataset that includes a column with a nested dataset. If a column that needs
expanding is automatically detected, choose which column to expand and which
columns to extract from the nested information. Click submit and you can view
your information as columns of a table instead of nested data.
4.2. Merge Files
By selecting and uploading multiple datasets by ctrl clicking (command for
mac), this will merge them into one larger dataset than be used for something
else.
Just select all datasets and fill in the required information. This will save
the new dataset to the iMed application and is then available for download.
4.3. Plot Functions
This section lets the user plot their dataset. Choose one of the options on
the left-hand side menu and then fill in the required fields to obtain your
plot. Below are the types of plots you can make from your data:
4.4. Statistical Analysis
This section lets us run statistical tests on our dataset. Choose a test to
run from the left side menu and fill in the fields to run the tests. Below are
the types of tests that are available:
ODPAC
5.1. Learn
This page includes a brief description of each type of resource available on
this page. Clicking the button at the top of each section will link to another
page allowing the user to use or learn more about the topic.
5.1.1. Epistasis
This page lets us use MBS, a search algorithm to learn from data.
Specifically, it allows us to study epistasis, the interaction between two or
more genes that affect the phenotype. This is useful to profile diseases in
the genetic aspect. Conventional methods are not suited to handle the high-
dimensional data found in genome-wide association studies (GWAS). The Multiple
Beam Search (MBS) algorithm allows detecting interacting genes at a much
faster rate. Upload the data that you want to use and then input the required
fields. For more in-depth information, find the full paper here.
5.1.2. Risk Factors
This page lets us use the IGain package to learn interactions between data. It
specifically learns interactions from high-dimensional data using a heuristic
search. This method builds on the Exhaustive_IGain method previously developed
to learn interactions from low-dimensional data. Upload the data and then
input the required fields. More information about the IS thresholds and iGain
can be found here.
5.1.3. Prediction Models
This section allows the use of prediction models already pre-built on top of
machine learning models to expedite its use. This allows their use without the
use of coding and prior experience to predict models using their own dataset.
There a numerous prediction models available to the user including Logistic,
Regression, Support Vector Machines (SVMs), Decision Trees, and many more. The
full list of prediction methods are found on the right side of the page here.
5.2. Prediction
This section allows predictions from a shared model previously uploaded. First
upload a shared model if not done so already. Then choose the model to use for
prediction by clicking the model name. Then upload the data for the prediction
model to use. This can be done either manually using the form at the bottom of
the page or using the template available for download. If using the template,
upload the dataset file and click submit to receive the model prediction.
5.3. Decision Support
Decision support provides classification and can guide treatment choices from
information supplied to the system. It has been trained from data to recommend
the optimal treatment procedure based on a patient’s features. More
information regarding Clinical Decision Support Systems (CDSS) can be found
here.
The System Recommendation takes a patient’s features and recommends treatment
procedure and predicts the future probability of 5 year metastasis. The User
Intervention takes both the patient features and the treatment procedure to
predict the future probability of 5 year metastasis based on current treatment
instead of optimal treatment.
MBIL
The Markov Blanket and Interactive Risk Factor Learner (MBIL) is an algorithm that learns single and interactive risk factors that have a direct influence on a patient’s outcome. Click “go to MBIL” to be redirected to the Python Package Index (PyPI) for the MBIL package located here. More information about MBIL can be found at BMC Bioinformatics.
Datasets
This section allows the user to see and upload new datasets to the web
application.
7.1. See All Datasets Available
To see all datasets available, simply click “Show Available Datasets.”
7.2. Upload a Dataset
To upload a dataset, click “Share Your Datasets” and then fill in required
information as stated on the webpage. First, upload the dataset and fill in
required fields.
Then, fill in the fields below or upload a text file with the information filled in. An example of how to organize the information so the application can understand it is given below.
Models
This section allows the user to see models available to them and share a
model.
8.1. See All Models Available
To see all models available, click “Show Available Models.”
8.2. Share a Model
To share a model, click on “Share Your Models” and then upload a model file
trained by tensor flow or PyTorch.
8.2.1. Related Dataset
You should then upload the related dataset which includes the headers. The
class/label for the dataset should be in the last column.
8.2.2. Predictors and Class information
If the dataset includes all the features, the feature form can be skipped
after uploading the dataset. However, if they are not all included, this
information must be provided in the description file or within the feature
form. Choose the option from the drop down indicating how you intend to
provide the predictors and class information.
If using the description option, you can either fill in the fields or upload a text file with the information filled in. An example of how to organize the information is given below.
References
- imed.odpac.net/odpac/learn/prediction
- Leveraging Bayesian networks and information theory to learn risk factors for breast cancer metastasis | BMC Bioinformatics | Full Text
- A clinical decision support system learned from data to personalize treatment recommendations towards preventing breast cancer metastasis - PubMed
- mbil-py · PyPI
- A Fast Algorithm for Learning Epistatic Genomic Relationships - PMC
Read User Manual Online (PDF format)
Read User Manual Online (PDF format) >>