Multimodal Imaging and Learning for Computional-based Medicine (MILCOM) Chair

Site sur Centrale Nantes

The following projects have been supported in part by the European Regional Development Fund (FEDER), the Pays de la Loire region on the Connect Talent scheme and Nantes Métropole (Convention 2017-10470).

Survival Analysis, PET images and Machine Learning (MILCOM, SIRIC)

This project points to the development of machine learning algorithms to assist the diagnosis and personalized treatment of patients suffering from hematological diseases such as multiple myeloma or diffuse large B-cell lymphoma (DLBCL) patients. In particular, we aim to predict a patient’s prognosis or treatment response given their full-body PET images, possibly combined with clinical data. To this end, we have proposed several types of approaches:

Graph Neural Networks for survival analysis

Self Supervised and Multitask learning

Random Survival Forests Framework

These works are done in close collaboration with the Nuclear Medicine department of the Nantes CHU and the INSERM CRCI2NA team 2. They also contribute to the SIRIC ILIAD.

---

Image Reconstruction and Deep Neural Models (MILCOM, )

Diffusion Models for Ultrasound Image reconstruction

Deep Image Priors for PET Image reconstruction

---

Learning with small datasets and few (or no) annotations (MILCOM, ANR-FULGUR)

Volumetric Segmentation / MR and ultrasound

Curriculum and Federated Learning / Fracture Classification / X-ray

---

Early Breast Cancer Detection (CIFRE HERAMI)

Industrial Collaboration with HERAMI

Multiscale Graph Neural Networks

Weakly Supervised and Multitask Learning for Anomaly Detection

---

Image registration with NN (CIFRE ICO-KEOSYS)

Collaboration with ICO and KEOSYS

Deep Image Regularization for Image Registration

---

Page template forked from evanca