Narrow your search

Library

ULiège (1)


Resource type

dissertation (1)


Language

English (1)


Year
From To Submit

2018 (1)

Listing 1 - 1 of 1
Sort by

Dissertation
Master thesis : New tools for the analysis of Cytomine user viewing behavior
Authors: --- --- --- --- --- et al.
Year: 2018 Publisher: Liège Université de Liège (ULiège)

Loading...
Export citation

Choose an application

Bookmark

Abstract

In the medical field, doctors and researchers should be able to observe and interpret cell
samples. The most widely spread method is to observe samples with a microscope. New
methods allow for scanning samples into very large and detailed images. These images
can be uploaded to the Cytomine web application and doctors can go through the images
at ease. The application includes many functionalities, namely the ability to annotate regions
of interest. Currently the ULiege MOOC server is open to the general public (see
https//www.fun-mooc.fr). Anyone can take part in the MOOC, but it's mostly used for the
bene t of the students studying the medical field at the University of Liege. The students
study particular images and are then evaluated at the end of the year. Meanwhile, the Cytomine
app has been collecting data on the students' time spent on the website. The main
objective is to develop IT tools to analyze data collected. The bulk of the data collected
consists of where the students decided to look in the images (Gaze data). Consequently, a
tool was implemented to find correlations between students' behavior and the results they
obtain during exams. Using Machine Learning techniques, the goal is to predict a student's
grade based on how they used the application. This allows for the analysis of specific viewing
patterns that can give more insight on how the student used Cytomine. Currently, the
model contains 395 students' data with over 2000 features. Extra Trees and Random Forest
learning techniques have been applied to attempt to predict grades. Otherwise, other
tools were implemented to better visualize these patterns. The idea would be to generate
Heatmaps of the students' gaze data (Gazemap). These Gazemaps would be included in the
application and users can be given access to this information. This could give teachers the
ability to keep track of the students' work.

Listing 1 - 1 of 1
Sort by