How to study rare brain tumors using deep learning


(c) Tunpur Casa

Researchers from CEA-Joliot (NeuroSpin), in collaboration with the Gustave Roussy Institute, Necker Hospital and the Curie Institute (Orsay), propose an original method for analyzing MRI images of rare brain tumors, by combining detection to me (Automation control is part of engineering sciences. This discipline deals with…) of things and fragmentation (In general, the word segmentation refers to the procedure of segmentation, the fact of segmenting oneself…) by learning (Learning is the acquisition of knowledge, i.e. the process…) Deep on common tumors.

To study brain tumors from MRI images, oncologists must precisely define (demarcate) the features of the lesions or “segment” them, that is, group the pixels in the image into different groups, according to preset criteria.

There are currently several Architect (Architecture is a documentary series proposed by Frederic Campan and Richard Cobans, …) Deep learning segmentation of brain tumors. These models are only effective for types tumor (The term tumor (from the Latin tumor, for hypertrophy) refers, in medicine, to an increase in …) in which they were trained. So it is better for common tumors, such as glioblastoma multiforme (Glioblastoma multiforme (GBM), also known as grade 4 astrocytoma, is…)from rare tumors, such as glioma (Gliomas or gliomas are brain tumors that arise from the supporting tissues or …) Offside trunk (could be a trunk 🙂 Brain (children’s cancer).

However, there are some visual similarities between common and rare tumors that allow to deal with the problem in two steps: detection and then pixel classification.

This is the approach taken by NeuroSpin researchers and their partners. They provide two methods of demarcation based on:
– object detection using an automatic object detection algorithm, known for high accuracy and speed (you only look once);
– Segmentation of tumors by A network (A computer network is a group of equipment linked together to exchange information…) From convolutional neurons, they were developed for biomedical image processing (U-Net).

For each step, neural networks trained on common lesions in rare lesions were used, without adjusting additional parameters. this is strategy (Strategy – from the Greek word stratos which means “army” and Ageîn which means…) Better results are allowed when the tumor differs from the training tumor and strong delineations of infiltrating gliomas have been obtained in the brainstem.

By addressing the topic of rare tumors that do not Database (In computing, a database (Abr: “BD” or…) It cannot be built to train a deep neural segmentation network, the researchers showed that by combining “simple” object detection with tumor segmentation, good results can be obtained, without retraining or model adaptation.

References:
Antibody detection improves tumor segmentation in MRI images of rare brain tumors and cancers

Did you like this article? Would you like to support us? Share it on social networks with your friends and/or comment on it, this will encourage us to publish more similar topics!

#study #rare #brain #tumors #deep #learning

Leave a Comment

Your email address will not be published.