INCF presenting how to participate on Youtube
INCF presenting how to participate on Youtube
The Active Segmentation platform for ImageJ (ASP/IJ) was developed in the scope of GSoC 2016 – 2019. The plugin provides a general-purpose environment that allows biologists and other domain experts to use transparently state-of-the-art techniques in machine learning to achieve excellent image segmentation and classification. ImageJ is a public domain Java image processing program extensively used in life and material sciences. The program was designed with an open architecture that provides extensibility via plugins.
The project idea: The existing machine learning model of Active Segmentation is based on the Weka library. However, this is limited to traditional machine learning approaches. The objective of the project will be to incorporate deep-learning functionality into the platform. Deep neural nets are capable of record-breaking accuracy.
The project is a continuation of GSOC 2020. The project will start from an already available codebase implemented using Deeplearning4j. At present, there are 2 implemented architectures U-Net — an architecture for biomedical image segmentation, and SegNet — a deep learning semantic segmentation architecture. The candidate will work out the GUI integration with the rest of the ASP/IJ platform.
Minimal set of deliverables:
Desired skills: Java, Machine Learning
Mentors: Sumit Vohra, ZIB, Berlin, Germany; Dimiter Prodanov @dprodanov, INCF Belgian Node (backup)
ImageJ: https://imagej.nih.gov/ 6
Weka https://www.cs.waikato.ac.nz/ml/weka/ 3
Active Segmentation : https://github.com/sumit3203/ACTIVESEGMENTATION 6
Deeplearning4J: https://deeplearning4j.org/ 1
Tags: ImageJ, segmentation, machine learning, deep learning, GUI
We are pleased to announce that despite the difficulties imposed by the COVID-19 pandemic the Belgian Neuroinformatics Node is committed to its mission of developing software tools for the scientific community.
This year we are sponsoring two GSOC 2020 projects:
by Raghavendra Singh Chauhan
by Joanna Stachera
The Belgian Neuroinformatics Node participates in GSOC 2019 with INCF.
Follow us on Neurostars.org
Project ideas postings:
There are also pages on the node’s website:
This year the Belgian INCF Node proposed two projects under the Google Summer of Code 2018.
Weare happy to announce that the GSOC project
(3rd iteration ) by Sanjeev Dubej was successfully completed!
The project incorporated Legendre image moments and texture descriptors for the purposes of classification of different objects within similar images.
The code is available for download at:
Sanjeev was an excellent student. He was very attentive to details and feedback from our side. We appreciate very much his effort in development and testing of our platform. This allowed for solving a number of bugs on the user interface.
Sanjeev highlights his experience in the following blog post .
The lab of Daniele Marinazzo at UGent has developed a toolbox from the retrieval and deconvolution of the Hemodynamic Response Function (HRF) from resting state fMRI signals. Once that the HRF has been retrieved for each voxel, it can be deconvolved from the time series (for example to improve lag-based connectivity estimates), or one can map the shape parameters everywhere in the brain (including white matter), and use the shape as a pathophysiological indicator. The goal of this project was to make a BIDS app out of this toolbox. This involved translating the code to Python, and to make a container of it.
Neuroimaging relies upon established standard to make sense of shared data and to ensure robustness and reproducibility of the analyses. To answer these needs, the BIDS (Brain Imaging Data Structure http://bids.neuroimaging.io/) format has been proposed. A BIDS App (http://bids-apps.
Madhur Tandon is the student who completed the project. Here the link to his report.
Asier Erramuzpe Aliaga, a former GSoC student, was co-mentor of the project. Guorong Wu, Nigel Colenbier, and Sofie Van Den Bossche were also massively involved in testing and brainstorming. The app is now listed among the official BIDS app (http://bids-apps.
The Belgian node will support the development of three Google Summer of Code projects in 2018.
ImageJ is an open source Image Processing library written in Java which is being
extensively used by people working in Life and Material sciences. ImageJ supports various
application softwares through plugins. The Active Segmentation plugin provides a
general-purpose environment that allows biologists and other domain experts to use
transparently state-of-the-art techniques in machine learning to improve their image
Most of the currently used data is task-based fMRI data i.e. the data acquired when the subject performs an explicit task (such as finger-tapping). It is clear however, that the brain is active even when the subject is not performing a task. In fact, according to certain estimates, task-related changes in neuronal metabolism only account for about 5% of the brain’s total energy consumption. Resting state fMRI is a comparatively newer approach used to identify synchronous BOLD changes in multiple brain regions while the subjects lie comfortably in the scanner without performing a task. The correlations between spontaneous BOLD signals can be used to study intrinsic functional connectivity. The analysis is also useful in situations where it is advantageous to explore the brain’s functional organization in neurological or psychiatric diseases. INCF currently has a tool to retrieve the hemodynamic response function at rest from resting state fMRI, however, the code is currently written in MATLAB. In order to increase its diffusion and portability, the repository has to be converted to Python and follow the BIDS – Standards to eventually create the portable neuroimaging pipeline.
The GSOC 2017 was successfully completed by Mukesh Gupta. The aim of the project was to provide the general purpose environment that allows biologists and other domain experts to use state-of-the-art techniques in machine learning to improve segmentation of objects in different types of images. The project was completed under the supervision of Dimiter Prodanov (Belgian INCF Node Fellow) and Sumit Kumar Vohra (an Alumni of the Belgian INCF Node )
In this Google summer of code we expanded the existing functionality of the Active Segmentation plugin. Earlier this tool only used for Segmentation, now functionality has been extended to classification of the images. We used Zernike moments as a feature vector so we extracted Zernike moments from the images and then used these features to train our SVM classifier to classify images.
The code can be found here: https://mukesh14149.github.io/
Sumit Kumar Vohra completed successfully thee project Modular Machine Learning and Classification Toolbox for ImageJ supported by the Google Summer of Code. The code is accessible at the INCF repository at GitHub.
Google Summer of Code (GSoC) is a global program that offers students stipends to write code for open source projects. INCF applied as a mentoring organization in the Google Summer of Code 2016.
The plugin is called Active Segmentation and its functionality is described here.