The preparation of a test data set for testing synthetic 2D mammography
Belgium (BeMSA) - KU Leuven, Leuven
Medical Physics and Qualty Assessment in Radiology
Hilde Bosmans
Hilde Bosmas
English, Dutch, French
4 weeks
Cities/Months Jan Feb Mar Apr May Jun Jul Augt Sep Oct Nov Dec
No No No No No No Yes Yes Yes No No No
Type of Research Project
- Clinical Project without Laboratory work
What is the background of the project?
The classical approach to mammography screening is the use of 2D digital mammography. With this x-ray technique, the volume of the breast is projected in a 2D plane. All breast structures superimpose on top of each other and it may be very difficult to see any (small) breast cancers. In order to improve the situation, breast tomosynthesis has been developed: this technique creates a pseudo 3D volume. In a next stage, intelligent algorithms comprise the 3D volume again in a 2D image, yet try to preserve the lesions that are potentially better visualized in the 3D slab. As more and more intelligent (maybe artificial intelligence) algorithms are being applied, synthetic mammography cannot be tested with a simple homogeneous test object.
What is the aim of the project?
Develop a test set of tomosynthesis stacks with clinically relevant lesions of different sizes and the associated synthetic 2D images in order to compare tomosynthesis and synthetic 2D mammography in terms of performance.
What techniques and methods are used?
- Create a data base of computer models of clinically relevant breast cancers (complete our current data base of 45 models with some more models) - Simulate these lesions in tomosynthesis projections such that these hybrid images look like real - Calculate also the synthetic 2D mammograms - Set up an observer study in which a human observer scores both the tomosynthesis and the synthetic 2D mammograms for detectability of lesions - Compare both modalities
What is the role of the student?
- The tasks will be done under supervision
What are the tasks expected to be accomplished by the student?
- Familiarize himself/herself with breast cancer imaging - Learn to recognize cancers in breast tomosynthesis and 2D mammography - Make a relevant selection of cancers based on natural frequencies of cancer types - Use a simulation framework to make the images - Load the images in a reading software - Prepare an observer performance study - Score the images, and ask other co-workers and radiologists to run the study - Compare the results and make a report
Will there be any theoretical teaching provided (preliminary readings, lectures, courses, seminars etc)
What is expected from the student at the end of the research exchange? What will be the general outcome of the student?
- The student will prepare a scientific report
- The student’s name will be mentioned in a future publication
What skills are required of the student? Is there any special knowledge or a certain level of studies needed?
- Some basic computer skills (excel) - Being able to work carefully
Are there any legal limitations in the student’s involvement
Type of students accepted
This project accepts:
- Medical students
- Graduated students (less than 6 months)
- Pre-Medical students from the American-British system
- Students in biomedical fields
- Impact of compressed breast thickness and dose on lesion detectability in digital mammography: FROC study with simulated lesions in real mammograms. Salvagnini E; Bosmans H; Van Ongeval C; Van Steen A; Michielsen K; Cockmartin L; Struelens L; Marshall NW. Med Phys. 2016 Sep;43(9):5104