ETRO VUB
About ETRO  |  News  |  Events  |  Vacancies  |  Contact  
Home Research Education Industry Publications About ETRO

Master theses

Current and past ideas and concepts for Master Theses.

Evaluate Transfer Learning in Deep Convolutional Neural Networks for 3D Breast Microcalcification Classification

Subject

The main indicator of an early non-palpable breast cancer is the presence of micro calcifications (MCs), small calcium deposits in breast tissue which are not always restricted to malignancies as they also appear in healthy breasts. The last decades, many computer aided detection and diagnosis (CAD) systems have been proposed to diagnose breast cancer based only on MCs properties. Recently, promising results have been achieved by extracting a high amount of handcrafted radiomic features on MCs structures. Given the fact that in many cases, deep convolutional neural networks (DCNN) have outperformed systems based on hand-crafted features, we focus to evaluate the performance of DCNN on MCs classification. Among the existing deep learning approaches, using 2D slices as the input data to a DCNN has demonstrated to be an effective approach where very good results have been achieved. The thesis purpose focuses on evaluating several 2D slice-based DCNN methods whereas focus will be payed also to evaluate if using transfer learning can help models to get better performances.

Kind of work

1- Literature review about: (a) the medical problem the student is trying to solve, (b) deep convolutional neural networks, (c) transfer learning. Estimated Time of Completion (ETOC): 3 months)
2- Get familiar with the current/new dataset to be used (high resolution 3D micro-CT images). (ETOC: 1 months)
3- Evaluate the performance of several DCNN on a slice-based approach. Student should analyse also ensembling strategies (i.e.: majority voting) to combine multiple slice-results into a single volume result. (ETOC: 2 months)
4- Evaluate and analyse the influence of using transfer learning in the different DCNN models used to classify breast microcalcification. (ETOC: 3 months)
5- Writing and presentation. (ETOC: 1 month)

Framework of the Thesis

https://ieeexplore.ieee.org/document/8531134
https://www.sciencedirect.com/science/article/pii/S2405959518304934
https://ieeexplore.ieee.org/abstract/document/9345718
https://search.proquest.com/docview/2478868781?pq-origsite=gscholar&fromopenview=true
https://www.researchgate.net/publication/349021592_2DSlicesNet_A_2D_Slice-Based_Convolutional_Neural_Network_for_3D_Object_Retrieval_and_Classification

Number of Students

1

Expected Student Profile

Required qualifications:
• Following an MSc in a field related to one or more of the following: Computer Science, Electrical Engineering, Biomedical Engineering, Applied Computer Science - Digital Health.
• Experience with image processing and signal processing.
• Experience with machine learning and statistics.
• Strong programming skills (Python).
• Interest in developing state-of-the-art Deep Learning methods and conduct experiments.
• Ability to write scientific reports and communicate research results at conferences in English.

Promotor

Prof. Dr. Bart Jansen

+32 (0)2 629 1034

bjansen@etrovub.be

more info

Supervisor

Miss Redona Brahimetaj

+32 (0)2 629 2930

rbrahime@etrovub.be

more info

- Contact person

- IRIS

- AVSP

- LAMI

- Contact person

- Thesis proposals

- ETRO Courses

- Contact person

- Spin-offs

- Know How

- Journals

- Conferences

- Books

- Vacancies

- News

- Events

- Press

Contact

ETRO Department

Tel: +32 2 629 29 30

©2022 • Vrije Universiteit Brussel • ETRO Dept. • Pleinlaan 2 • 1050 Brussels • Tel: +32 2 629 2930 (secretariat) • Fax: +32 2 629 2883 • WebmasterDisclaimer