Icon

MRI Segmentation & Prediction using Unet

Partitioning Transfer Learning Post Processing & Evaluation NotesOriginal dataset link: https://www.kaggle.com/mateuszbuda/lgg-mri-segmentationPlease download the full dataset from the link above, due to upload size purposes I was unable to compile it together with the workflow.This workflow uses Unet architecture with a VGG16 backbone and imagenet weights in order to predict with good accuracy the presence of a tumor from MRI Brain Scans.The pre-processing is mainly to append the mask in the same row as the original scan, and to see the labeling superimposed on the original MRI.The accuracy metrics for this workflow are limited to the difference between the original mask and the predicted mask, although there are many other methods you can utilise tocalculate the accuracy. Superimposed Labeling on MRI train for 100 epochs(batch size 8) withAdam Right Click >View: Interactive Segmentation View> Bounding Box Render> Show Bounding Box namesNode 322Node 323Node 326Node 379convert labelings to byteNode 381find differencesbetween imagevisually compareall segmentsconvert toBIT TYPENode 404Node 405Keras NetworkLearner InteractiveSegmentation View Partitioning Partitioning DL PythonNetwork Creator DL PythonNetwork Editor Keras NetworkExecutor Image Converter Image Normalizer Image Calculator Pre-Processing Compare Segments Image Converter Image Viewer Image Reader Partitioning Transfer Learning Post Processing & Evaluation NotesOriginal dataset link: https://www.kaggle.com/mateuszbuda/lgg-mri-segmentationPlease download the full dataset from the link above, due to upload size purposes I was unable to compile it together with the workflow.This workflow uses Unet architecture with a VGG16 backbone and imagenet weights in order to predict with good accuracy the presence of a tumor from MRI Brain Scans.The pre-processing is mainly to append the mask in the same row as the original scan, and to see the labeling superimposed on the original MRI.The accuracy metrics for this workflow are limited to the difference between the original mask and the predicted mask, although there are many other methods you can utilise tocalculate the accuracy. Superimposed Labeling on MRI train for 100 epochs(batch size 8) withAdam Right Click >View: Interactive Segmentation View> Bounding Box Render> Show Bounding Box namesNode 322Node 323Node 326Node 379convert labelings to byteNode 381find differencesbetween imagevisually compareall segmentsconvert toBIT TYPENode 404Node 405Keras NetworkLearner InteractiveSegmentation View Partitioning Partitioning DL PythonNetwork Creator DL PythonNetwork Editor Keras NetworkExecutor Image Converter Image Normalizer Image Calculator Pre-Processing Compare Segments Image Converter Image Viewer Image Reader

Nodes

Extensions

Links

Download

To use this workflow in KNIME, download it from the below URL and open it in KNIME:

Download Workflow
Created by: MSF
Created at: 2021-12-14
On NodePit since: 2021-12-30
Last update: 2024-04-27
Created with KNIME version: v4.5.0

Deploy, schedule, execute, and monitor your KNIME workflows locally, in the cloud or on-premises – with our brand new NodePit Runner.

Try NodePit Runner!