Digital avatars of patients lungs

Medi­cine isn’t just about drug treat­ments. Dif­fer­ences between patients are not lim­ited to genet­ics. At MΞDISIM, research­ers are ana­lys­ing the bio­mech­an­ics of dis­eased tis­sue to cre­ate digit­al mod­els for indi­vidu­al patients as a way bet­ter guid­ing thera­peut­ic choices.

“The lung changes shape when we breathe, and doubles in volume”, Cécile Patte, a research­er and recip­i­ent of the 2020 L’Oréal-UNESCO for Women in Sci­ence Award who recently defen­ded her doc­tor­al thes­is at MΞDISIM, points out. “The way it changes shape depends on its mech­an­ic­al prop­er­ties, such as elasti­city.” She developed a digit­al lung mod­el in order to help doc­tors at Avicenne Hos­pit­al in France to bet­ter under­stand the indi­vidu­al char­ac­ter­ist­ics of patients with idiopath­ic pul­mon­ary fibrosis – a chron­ic lung dis­ease, which is one of the long-term effects of Covid-19.

In this con­di­tion, scar tis­sue forms in the lungs, ren­der­ing them stiff, block­ing the pas­sage of oxy­gen to the blood and lead­ing to fatal res­pir­at­ory fail­ure with­in just two to five years. Sig­ni­fic­antly, the shape, poros­ity and mech­an­ic­al prop­er­ties of the lung are not the same for all patients. “In order to answer gen­er­al ques­tions, you only need a mod­el of an aver­age lung, using aver­age meas­ure­ments for the mech­an­ic­al prop­er­ties. But for a per­son­al­ised approach, you need to build an avatar of the patient’s organ,” she explains. This means acquir­ing data.

Real life data

“We want to work with exist­ing data, although it makes things more dif­fi­cult. We don’t want to sub­ject patients to extra invas­ive pro­ced­ures in order to cre­ate the avatar,” Patte says. The mod­els are there­fore based on data col­lec­ted dur­ing the nor­mal course of the patient’s treat­ment. Exams vary depend­ing on the kind of patho­logy and the organ in ques­tion. For the heart, there are blood pres­sure and elec­tro­physiology meas­ure­ments. For the lung, it’s mainly ima­ging tech­niques, X‑rays and scans. “If we could meas­ure pul­mon­ary pleur­al pres­sure, for example, the mod­el would be more accur­ate, but the test is too invas­ive.”

The patient’s spe­cif­ic inform­a­tion is fed into a digit­al avatar, enabling more detailed dia­gnos­is, which includes quan­ti­fy­ing mech­an­ic­al para­met­ers, eval­u­at­ing poten­tial treat­ments and pre­dict­ing pro­gnos­is. “This inform­a­tion should help doc­tors bet­ter dia­gnose the dis­ease and guide the patient towards drug ther­apy or transplantation.”

Cur­rently, per­son­al­ised sim­u­la­tion is only at the research, or ‘proof of concept’, stage. Much work still needs to be done before it can be applied clin­ic­ally and used as a med­ic­al device to help make thera­peut­ic decisions. The reg­u­lat­ory envir­on­ment is strict and sub­ject to EU require­ments, such as CE marking.

So far, Patte’s lung sim­u­la­tion soft­ware has only been applied to four patients. “Using the com­put­ing capa­city of the lab’s serv­er cluster, our soft­ware needs up to three full days to pro­cess one patient’s data,” she admits. Pro­cessing time will have to be shortened for the sys­tem to work in clin­ic­al practice.

A digit­al twin

The lung is not the only organ the team is mod­el­ling: a lot of work is being done on the heart with mul­tiple clin­ic­al applic­a­tions, and in dir­ect col­lab­or­a­tion with vari­ous hos­pit­als. The aim is to rep­res­ent the heart and its vari­ous char­ac­ter­ist­ics, includ­ing size, con­tractil­ity and elec­tro­physiology. Per­son­al­ised car­di­ac mod­els could enable doc­tors to test out a treat­ment vir­tu­ally, par­tic­u­larly in the case of heart fail­ure, and pre­dict wheth­er or not the patient would respond.

MΞDISIM has also designed a tool, Anaes­tAssist, to mon­it­or the car­di­ovas­cu­lar sys­tem in real time while a patient is anes­thet­ised. This soft­ware mod­els the patient’s car­di­ovas­cu­lar physiology in order to pre­dict the effects of the anaes­thet­ic and assist the doc­tor dur­ing the oper­a­tion. Oth­er labor­at­or­ies through­out the world are work­ing on mod­el­ling blood flow, bones, kid­neys, etc. Patte believes that “all organs can be stud­ied with this approach.”

It may one day be pos­sible to bring all these mod­els togeth­er in order to sim­u­late sys­tem­ic effects and cre­ate full digit­al twins of patients to help pin­point indi­vidu­al­ised treat­ments. This idea has even gained ground in indus­tri­al circles, with invest­ment from com­pan­ies such as Dassault Sys­tèmes, who cre­ated a plat­form called 3DEXPERIENCE that can help med­ic­al device man­u­fac­tur­ers optim­ize product devel­op­ment through simulation.