Icon

ABF_​framework

<p><strong>ABF Framework — KNIME Workflow</strong></p><p>This KNIME workflow implements the complete&nbsp;<strong>ABF (Apparent Brain Features) framework</strong>&nbsp;for reproducible analysis and prediction.</p><p>The workflow contains the complete processing pipeline, including data preparation, feature processing, model application, and prediction/interpretation steps. The trained models are already included/configured in the workflow, so users do&nbsp;<strong>not need to retrain the models</strong>&nbsp;to test the pipeline.</p><p><strong>Input Data</strong></p><p>The original dataset used for developing and evaluating the framework cannot be distributed with this workflow because it contains FreeSurfer-processed data subject to data-sharing restrictions.</p><p>Therefore, a&nbsp;<strong>sample CSV dataset</strong>&nbsp;is provided with the workflow. The sample dataset demonstrates the required input structure and allows users to execute and explore the complete workflow.</p><p>How to Use the Workflow</p><ol><li><p><strong>Open the workflow in KNIME.</strong></p></li><li><p><strong>Run the workflow using the provided sample CSV.</strong><br>This allows you to verify that the workflow is correctly configured and to explore the different processing steps and outputs.</p></li><li><p><strong>Prepare your own dataset.</strong><br>To use the workflow with your own data, prepare a CSV file following the same structure as the provided sample dataset.</p></li><li><p><strong>Replace the sample CSV with your own dataset.</strong><br>The column names and data types should follow the structure expected by the workflow. The sample CSV can be used as a reference for the required format.</p></li><li><p><strong>Execute the workflow.</strong><br></p></li></ol><p><strong>Important Note</strong></p><p>The sample CSV is provided only to demonstrate the required input structure and to allow users to test the workflow. It does not represent the original dataset used to train and evaluate the models.</p><p>Users who want to apply the workflow to their own data should replace the sample input with a compatible CSV file while maintaining the required structure.</p><p>The workflow is intended to provide a reproducible implementation of the ABF framework within the KNIME Analytics Platform, allowing users to inspect, execute, and adapt the pipeline using their own compatible data.</p><p></p><p>Please refer to the following repositories for the missing nodes. </p><p>Reference Repository:<br>https://github.com/AliBhatti21/Data-Science-with-KNIME.git</p><p>Reference Papers:</p><p>https://link.springer.com/chapter/10.1007/978-981-95-9575-4_21</p><p></p>

ABF Framework — KNIME Workflow

This KNIME workflow implements the complete ABF (Apparent Brain Features) framework for reproducible analysis and prediction.

The workflow contains the complete processing pipeline, including data preparation, feature processing, model application, and prediction/interpretation steps. The trained models are already included/configured in the workflow, so users do not need to retrain the models to test the pipeline.

Input Data

The original dataset used for developing and evaluating the framework cannot be distributed with this workflow because it contains FreeSurfer-processed data subject to data-sharing restrictions.

Therefore, a sample CSV dataset is provided with the workflow. The sample dataset demonstrates the required input structure and allows users to execute and explore the complete workflow.

How to Use the Workflow

  1. Open the workflow in KNIME.

  2. Run the workflow using the provided sample CSV.
    This allows you to verify that the workflow is correctly configured and to explore the different processing steps and outputs.

  3. Prepare your own dataset.
    To use the workflow with your own data, prepare a CSV file following the same structure as the provided sample dataset.

  4. Replace the sample CSV with your own dataset.
    The column names and data types should follow the structure expected by the workflow. The sample CSV can be used as a reference for the required format.

  5. Execute the workflow.

Important Note

The sample CSV is provided only to demonstrate the required input structure and to allow users to test the workflow. It does not represent the original dataset used to train and evaluate the models.

Users who want to apply the workflow to their own data should replace the sample input with a compatible CSV file while maintaining the required structure.

The workflow is intended to provide a reproducible implementation of the ABF framework within the KNIME Analytics Platform, allowing users to inspect, execute, and adapt the pipeline using their own compatible data.

Please refer to the following repositories for the missing nodes.

Reference Repository:
https://github.com/AliBhatti21/Data-Science-with-KNIME.git

Reference Papers:

https://link.springer.com/chapter/10.1007/978-981-95-9575-4_21

Normalizes each FreeSurfer-derived brain ROI volume (e.g. left-ventricle hippocampus, ...) by the subject's Intracranial Volume (ICV/eTIV), removing the confound of individual head/brain size differences before any statistical modeling or feature selection is performed.
Filters the full set of candidate predictor ROIs down to a reduced, more informative subset by ranking/thresholding each predictor's Mutual Information with the target (either the target ROI in BFFS regression, or the CN/AD class label), keeping only those above a chosen threshold or top-k cutoff.

Training ABF Pipeline

Testing ABF Pipeline

Ensemble method:

Ensemble 1 (Simple Majority Voting)

Ensemble 2 (Independent ABF specific Optimized threshold)

ABF Framework & Interpretability

New Testing Pipeline
extracted only MI filter from 400+ features
Reference Column Splitter
Feature IDOriginal Feature NameMRI_Features.csv
CSV Reader
Table Transposer
RowID
metadata excluded
Column Filter
AD(45), CN(120)
GroupBy
Hold-out table
ROI models
Column Appender
ABF framework
ABF Interpretability
Column Filter
Column Filter
ABF framework
ABF Interpretability
Row Filter
Table Partitioner
Row Filter
Image View
Image View
Image View
Diag: AD and CN
Row Filter
Gender = female
Row Filter
mapping and cleaning
female subjects (1645)AD (444)CN (1201)
GroupBy
FS000 = age
Column Renamer
String to Number
FS000 = age
Column Renamer
FS000 = age
Column Renamer
included the SD005Subject ID
Column Filter
included the SD005Subject ID
Column Filter
BFFS (training set)
Hold-out table
ROI models
FS000 = age
Column Renamer
training set
Normalizer
Column Renamer
Column Filter
selected true
Row Filter
Column Renamer
test set
Normalizer (Apply)
Column Filter
Table Transposer
Top 10% features filtered
Mutual Information Filter
using linear correlation
MI feature selection
ICV Normalizer (Learner)
ICV Normalizer (Apply)
Column Filter
to get Resubstitution table
ROI models
Ensemble method
extracted only MI filter from 400+ features
Reference Column Splitter
Column Appender
feature columns name
Row to Column Names
Freesurfer 6.0Cortical and sub-cortical MeasurementsFS6.0_sample.csv
CSV Reader
double check
Row Filter
metadata excluded
Column Filter

Nodes

Extensions

Links