Thank you in advance for circulating the announcement below*. *Apologies
for multiple receptions.
We are seeking to identify a candidate in connection with the following
doctoral thesis subject (Doctoral School MathSTIC, University Bretagne –
Loire, France), in the domain of Information Science and Technology.
*Thesis subject*
*Modeling and comparison of patient profiles on a large scale * *
to improve diagnosis andtherapeutic treatment.**
Contribution to advances for the medicine of the future.***
*Descripti**on**: *
The generalization of digitized files at hospital entails the generation
of large biobanks that allow to store masses of information on an
unprecedented scale. These data allow to dismiss and select various
hypotheses on diagnosis and/or avenues for treatment. Collecting data
from large cohorts provides the possibility to connect a given patient
to other patients, for which feedback about response to treatmentsis
available, for example.
In this context, recommending a treatment customized to a patient
requires the ability to connect the latter to a group of patients.
However, collecting data from multiple patient cohorts is a process that
involvesmultiple hospital institutions. Thus, the description of the
patients inthe resulting biobank is prone to be heterogeneous, both in
terms of content (raw data, text, abstracts, images, time series,
environmental data (e.g. diet, physical activity)...) and abstraction
level (reference to genomic regions, biological markers, various time
scales ...). Furthermore, such recordsmaypresentmissing values.
The thesis work will focus on diagnosis or risk prediction, and, in the
same line, therapeutic treatment definition, on the basis of the
similarity of a «query» patient to other patients. A preliminary
examination will question how the profile of a query patient can be
modeled on the fly, according to the nature and characteristics of the
observations available for both this patient and the other patients, for
which the challenge is also to model rapidly and efficiently the
profiles. A subsequent step will consist in proposing metrics, in order
to compare such profiles.
Two working assumptions will be considered. The most simpleframeworkwill
set the focus ona group of patients /a priori/deemedinteresting to shed
light on the query patient’s case. In this framework, the problem is
stated as a comparison problem, but nevertheless remains complex
(heterogeneous data, missing data, time series characterizedbyvarious
time scales, indifferent time frames …). The second assumption does not
relyon an /a//priori/and is defined as a clustering problem in which a
group of patients « sufficiently similar » to the query patient has to
be identified.
A final objective will be to quantify theconfidence in either a
diagnosis or a treatment recommendation, in function of the group of
patients targeted (first hypothesis) or of the group of patients that
were automatically identified most similarto the query patient (second
hypothesis). Conversely, it will be examined under this latter
hypothesis whether it is possible to determinea confidence threshold
related to the objective (pinpoint a diagnosis, recommend a treatment)
to guide the identification of the group of patients. In this last part,
notably, experimentations resorting to large-scale simulations will be
envisaged.
*Key-words*: Knowledge modeling, modeling of traces,
comparisonofprofiles, recommendation, machine learning, statistical
learning, metrics, uncertainty, probabilistic model, massive data,
simulations
*L**ocation*: University ofNantes, LS2N (Laboratory for Digital Science
in Nantes) / UMR CNRS 6004,
DUKe Research group(Data User Knowledge)
*Thesis supervision*: Christine Sinoquet (christine.sinoq...@univ-nantes.fr)
*F**unding*: grant from the Ministry for Higher Education and
Research(36 months)
*C**andidat**e profile* : postgraduate studentranked in the top third of
her/his class, with respect to the theoretical results obtained during
the first semester of Master’s degree.
Further information about the thesis subject is provided
athttps://uncloud.univ-nantes.fr/index.php/apps/files/?dir=/thesis_subject_april_2018&fileid=61072601
*Application information*: The interested candidates are invited to send
as soon as possible a letter of motivation, a CV, a transcript of
examinations related to the Master, including the ranking obtained in
the Master, the coordinatesfor two contacts (affiliation, email address,
phone number).The deadline is tuesday /*8th*//**//*ma*//*y*//*2018.
*//The address to send the documents is //*:
*/christine.sinoq...@univ-nantes.fr.
-------------------
Christine Sinoquet, MC HdR 2014
http://christinesinoquet.wixsite.com/christinesinoquet
Faculty of Sciences
University of Nantes
LS2N / UMR CNRS 6004
France
https://www.ls2n.fr/
_______________________________________________
uai mailing list
uai@ENGR.ORST.EDU
https://secure.engr.oregonstate.edu/mailman/listinfo/uai
