[Stage] Proposition of automatic tumor detection framework
Stage · Stage M2 · 6 mois Bac+5 / Master Oncodesign Precision Medicine · DIJON (France)
clinical imaging tumor detection framework cancer DL segmentation
Topic: Proposition of automatic tumor detection framework
Duration: 3 to 6 months start on February
Location: Oncodesign HQ – Dijon
Benefits: Monthly indemnity + meal Ticket
OPM is a technological company specialized in precision medicine. OPM's mission is to bring innovative therapeutic
and diagnostic solutions to treat therapeutic resistance and metastasis evolution. The patient is at the center of our
reflection, of our unique innovative model, and our investments. For OPM "our collective success is paramount",
there can be no value creation without exchange, without dialogue. The value creation resulting for us from
reciprocity, i.e. balanced and fair exchanges at all levels, whether between internal collaborators, or with our
partners, therapists, patients, experts and investors.
Standard of care hospital practice includes several exams to diagnose and follow up patient’s cancer. One of the
routine exams is the imaging, including radiology (CT, MRI) and/or nuclear medicine (PET, scintigraphy) scans. Those
scans allow for the detection of internal anomalies (not visible on the surface) and their characterization. In the
case of oncology, so-called anomalies are cancer cell mass that accumulates in a tissue to form the primary tumor. Those cells can propagate and migrate though the bloodstream into a new tissue to form metastasis.
Collected images can be used for (i) the evaluation of the patient’s clinical outcome and (ii) fundamental research.
In this master thesis proposal, we want to focus on the fundamental research of collected standard of care images
from a clinical trial that is managed by OPM. This will involve image processing, including segmentation (i.e. the
extraction of anatomical structures from images) of the primary tumor and/or metastasis.
Even if deep learning has achieved remarkable results in many computer vision tasks, deep neural networks
typically need a large amount of training data to avoid overfitting. Unfortunately, in our case, labeled data for
clinical applications are limited. By improving the quantity and diversity of training data, data augmentation
algorithms  have become an inevitable part of deep learning model training with image data.
The aim of this internship is to study and implement automated segmentation methods of tumors, combining
unsupervised and supervised approaches with label uncertainty for tumor detection and segmentation in volumes
of CT data and/or MRI data in three specific pathologies (triple negative breast cancer (TNBC) , non-small cell
lung cancer (NSCLC) , pancreatic ductal adenocarcinoma (PDAC) ). The methods implemented should allow to
automatically perform metrology operations (localization of the tumor in the volume, quantification of the volume
of tumors, ...) and propose a level of confidence in the measurements performed. Finally, algorithms of data
augmentation should also be tested to study their impact on the prediction results.
The icing on the cake: Building models allowing the segmentation of the vascularity of the tumor
email@example.com 2 / 2
Missions & activities of the internship
o State of the art: Identification of different automated segmentation methods (supervised/unsupervised)
o State of the art: Identification of data augmentation algorithms (bibliography)
o Recovery of imaging datasets (annotated)
o Implementation of the most relevant algorithms
o Test data augmentation process
o Evaluation of the proposed solutions
Student expected background/Knowledge
M2 or Engineer in Computer Science / Bioinformatics with the following technical skills, or strong interest in:
o Machine Learning/Deep Learning (Tensor Flow, etc.) applied to images
o Applied mathematics
o Statistical knowledges
o Python / R languages
o Knowledge of medical imaging and/or human anatomy and/or cancer is a plus!
 Guo, Ying-Ying et al. “Breast MRI Tumor Automatic Segmentation and Triple-Negative Breast Cancer
Discrimination Algorithm Based on Deep Learning.” Computational and mathematical methods in medicine vol.
2022 2541358. 31 Aug. 2022, doi:10.1155/2022/2541358
 Primakov, S.P., Ibrahim, A., van Timmeren, J.E. et al. Automated detection and segmentation of non-small cell
lung cancer computed tomography images. Nat Commun 13, 3423 (2022). https://doi.org/10.1038/s41467-022-
 Mahmoudi, T., Kouzahkanan, Z.M., Radmard, A.R. et al. Segmentation of pancreatic ductal adenocarcinoma
(PDAC) and surrounding vessels in CT images using deep convolutional neural networks and texture descriptors. Sci
Rep 12, 3092 (2022). https://doi.org/10.1038/s41598-022-07111-9
 Chlap, P., Min, H., Vandenberg, N., Dowling, J., Holloway, L., & Haworth, A. (2021). A review of medical image
data augmentation techniques for deep learning applications. Journal of Medical Imaging and Radiation Oncology,
65(5), 545–563. doi:10.1111/1754-9485.13261
Contact: Thierry Billoué – Chief Human Resources Officer – Oncodesign Precision Medicine
Send your application (resume & motivation letter) under ref “AutoTumDet” to firstname.lastname@example.org
Date limite : None
Offre publiée le 1 décembre 2022, affichage jusqu'au 28 janvier 2023