A lec­tur­er and research sci­en­tist at UTC, Dan Istrate works in par­tic­u­lar on con­nect­ed bio­med­ical tools. Two cur­rent projects focus on pre­vent­ing the risk of pre­ma­ture deliv­ery in child-birth and detec­tion of the risk of the elder­ly at home catch­ing infec­tious diseases.

The first, car­ried out in col­lab­o­ra­tion with Imad Rida and Amiens Uni­ver­si­ty Hos­pi­tal, con­cerns at-risk preg­nant women. “Thanks to AI tools devel­oped by Imad, the aim is to pre­dict the date of deliv­ery, enabling doc­tors to pre­pare for the birth and care of the baby under opti­mum con­di­tions. HDsEMG elec­trodes placed on the preg­nant wom­an’s low­er bel­ly pick up uter­ine mus­cle sig­nals and, depend­ing on the char­ac­ter­is­tics of the sig­nals, pre­dict that the woman should give birth 1 to 2 weeks after the mea­sure­ments. This device can be used at home. We used machine learn­ing, then deep learn­ing. Not only does deep learn­ing pro­vide bet­ter per­for­mance, but it is also capa­ble of gen­er­at­ing arti­fi­cial data that can be used to build more accu­rate mod­els,” he explains.

The sec­ond project is being car­ried out in col­lab­o­ra­tion with Vin­cent Zalc, the Toulouse Uni­ver­si­ty Hos­pi­tal and the LAAS lab­o­ra­to­ry. “This project aims to detect infec­tious res­pi­ra­to­ry or gas­troen­tero­log­i­cal dis­eases using a min­i­mum num­ber of sen­sors, while pre­serv­ing the indi­vid­u­al’s pri­va­cy. We were par­tic­u­lar­ly inter­est­ed in peo­ple liv­ing in shared accom­mo­da­tion, to avoid the spread of dis­ease. The sys­tem con­sists of a micro­phone and motion sen­sors in the liv­ing room and bath­room, plus a con­tact on the door to mon­i­tor entry and exit. A sys­tem that allows us to deter­mine a per­son­’s cough­ing, sneez­ing or mobil­i­ty,” he assures us.

This sys­tem has been in place since 2022 in a 12-stu­dio res­i­dence belong­ing to “âges sans fron­tières” and locat­ed in Brens, near Toulouse. “AI inter­venes on two lev­els in this case. The first con­cerns sound recog­ni­tion, where we use machine learn­ing. The sec­ond, which we are cur­rent­ly devel­op­ing, aims to exploit infor­ma­tion con­cern­ing both sounds and move­ments, in order to gen­er­ate alerts about a giv­en per­son­’s state of health”, con­cludes Dan Istrate.

MSD