3IA PhD/Postdoc Seminar #12

Published on February 21, 2022 Updated on February 21, 2022
Dates

on the March 4, 2022

from 10:30am to 12:00pm


 

Program

10:30 - 11:00
Hind Dadoun (Inria)

AI-based Real Time Diagnosis of Abdominal Ultrasound Images

Abstract: Ultrasound (US) imaging is one of the most common techniques for medical diagnosis. It is the only non-invasive imaging modality, with no side effects (such as radiation-induced cancers) that can be used in real-time, making it a method of choice: whether in an emergency, in consultation for patient follow-up or during a public health screening examination. However, acquiring and interpreting an ultrasound image is a difficult and examiner-dependent task with a limited number of trained operators. As a result, despite the decrease in ultrasound hardware prices, limitations to its use persist. The focus of our study is to analyze how machine learning tools can be used for the automatic interpretation of abdominal ultrasound images, with a major setback : the absence of curated, annotated and openly available abdominal US databases. In this presentation, we will detail those challenges and point out first elements to alleviate some of them.

11:00 - 11:30
Victor Jung (UCA)

Checking Constraint Satisfaction

Abstract: We address the problem of verifying a constraint C by a set of solutions S in the context of Constraint Programming. This problem is present in almost all systems aiming at learning or acquiring constraints or constraint parameters. We propose an original approach based on MDDs. Indeed, the set of solutions can be represented by the MDD denoted by MDD(S). Checking whether S satisfies a given constraint C can be done using MDD(C), the MDD that contains the set of solutions of C, and by searching if the intersection between MDD(S) and MDD(C) is equal to MDD(S). This step is equivalent to searching whether MDD(S) is included in MDD(C). We show that this approach can be generalized for the computation of global constraint parameters satisfying C. Next, we show that the introduction of node information allows us to define a new algorithm capable to compute in only one step the set of parameters we are looking for. Finally, we present experimental results showing the interest of our approach.

11:30 - 12:00

Open discussion on the two contributions