π Digital π Industry π Health and biotech
Digital twins: what opportunities for industry?

A digital twin of the lungs: what benefits for medicine of the future?

Martin Genet, Assistant Professor in the Mechanics Department of École Polytechnique in the MΞDISIM team (INIRIA/IP Paris)
On February 1st, 2023 |
4 min reading time
Martin Genet
Assistant Professor in the Mechanics Department of École Polytechnique in the MΞDISIM team (INIRIA/IP Paris)
Key takeaways
  • A digital twin is the digitisation of an object and its environment: it simulates the behaviour of a real object in a virtual environment.
  • By simulating an individual's organ, it allows doctors to facilitate both diagnosis and patient care.
  • The goal is to customise a generic model by adding the specific data of an individual's organ.
  • Digital twins of living tissue can be used to illustrate even the interaction of a drug molecule with a sub-network of a DNA molecule.
  • The digital twin would give doctors the ability to predict diseases and try to prevent them.

We have all seen a dia­gram of a lung in our biol­o­gy text­books. With the dif­fer­ent dia­grams, we have been able to learn how it works. How­ev­er, these inan­i­mate dia­grams can­not show how an indi­vid­u­al’s organ works, they are only gener­ic mod­els. Today, by digi­tis­ing these mod­els, it is pos­si­ble to inte­grate the data reflect­ing the organ’s activ­i­ty. The vari­ables between indi­vid­u­als are thus tak­en into account thanks to a per­son­alised mod­el. This mod­el, which by def­i­n­i­tion is more accu­rate, is called a dig­i­tal twin. 

A dig­i­tal twin is the digi­ti­sa­tion of an object and its envi­ron­ment, which is intend­ed to be true to their phys­i­cal char­ac­ter­is­tics. It thus makes it pos­si­ble to sim­u­late the real behav­iour of an object in a vir­tu­al envi­ron­ment. It is a mod­el that offers such a high degree of cus­tomi­sa­tion of the digi­tised ver­sion of the object that it is almost as if we were study­ing the actu­al object. 

The idea is to move towards ever more detailed mod­els, although we know that no mod­el is ever perfect.

This type of numer­i­cal sim­u­la­tion is used in many fields of engi­neer­ing. And it is not dif­fi­cult to see how such a spe­cif­ic mod­el of an object can be use­ful in the med­ical field. After all, when an object can be reduced to a set of equa­tions cor­re­spond­ing to its phys­i­cal char­ac­ter­is­tics, every­thing can be digi­tised, even the organ of a liv­ing being. Mar­tin Genet, a researcher in bio­me­chan­ics, and sev­er­al col­leagues in the MΞDISIM research team at the Sol­id Mechan­ics Lab­o­ra­to­ry (LMS*) of École Poly­tech­nique (IP Paris)1, are work­ing on per­son­alised lung mod­el­ling to study idio­path­ic pul­monary fibro­sis. They believe that the mechan­i­cal process of breath­ing could pro­mote the onset and devel­op­ment of the dis­ease – a kind of mechan­i­cal vicious cir­cle. This research is being done with a view to pro­vid­ing doc­tors with some kind of tool to help them diag­nose and man­age the patient. 

Customising a generic model

First, you must start with the gener­ic mod­el of the tar­get organ. In the case of Mar­tin Genet, this is the lung. Then, using this as a basis, some patient-spe­cif­ic infor­ma­tion is applied to cus­tomise the gener­ic mod­el. The aim is to cre­ate a dig­i­tal mod­el that cor­re­sponds “as close­ly as pos­si­ble” to the organ to be stud­ied. “The idea is to move towards ever more detailed mod­els, although we know that no mod­el is ever per­fect,” he explains. “A dig­i­tal twin is there­fore a per­son­alised mod­el incor­po­rat­ing the spe­cif­ic data of a patient.”

Even if obtain­ing a per­fect sim­u­la­tion seems impos­si­ble, Mar­tin Genet reminds us that per­fec­tion is not nec­es­sary to obtain rel­e­vant med­ical infor­ma­tion. “The per­fect mod­el does not exist. A dig­i­tal twin remains a mod­el, an approx­i­ma­tion of real­i­ty. So, you can’t repro­duce every pos­si­ble exper­i­ment on some­thing real,” he admits. “But a giv­en dis­ease is linked to sev­er­al phys­i­cal, chem­i­cal, and mechan­i­cal process­es. Hence, to bet­ter under­stand the dis­ease, and ulti­mate­ly to bet­ter treat it, we obvi­ous­ly must take these process­es into account, but it is not nec­es­sary to under­stand all the physi­co-chem­i­cal phe­nom­e­na that have tak­en place in the patien­t’s life,” he insists.

As far as Mar­tin Genet and the MΞDISIM team’s research is con­cerned, the dis­ease in ques­tion is pul­monary fibro­sis – a dis­ease that is strong­ly relat­ed to tis­sue mechan­ics, cor­re­spond­ing in par­tic­u­lar to an over­pro­duc­tion of col­la­gen in lung tis­sue. As col­la­gen is rigid, too much of it caus­es the tis­sue to become rigid. A per­son with excess col­la­gen will end up forc­ing his or her breath­ing, which would in turn aggra­vate the dis­ease. “This is the fun­da­men­tal aspect of our research,” explains Mar­tin Genet. “To estab­lish whether mechan­i­cal con­straints tend to favour the onset and/or the evo­lu­tion of the dis­ease, and there­fore whether the hypoth­e­sis of this vicious cir­cle is valid.”

Under­stand­ing and clas­si­fy­ing fibrosis

Per­son­alised mod­el­ling can also be used to go fur­ther in diag­no­sis, par­tic­u­lar­ly for the poten­tial clas­si­fi­ca­tion of the type of fibro­sis the patient has, or sim­ply to mon­i­tor and opti­mise the treat­ments to be admin­is­tered. For this, once again, it is not nec­es­sary to know the patien­t’s entire med­ical his­to­ry. It could be enough to focus on one aspect of the dis­ease, such as the stiff­en­ing of the tis­sue. Espe­cial­ly since “dig­i­tal twins of liv­ing tis­sue are now being devel­oped that describe even the inter­ac­tion of a drug mol­e­cule with a sub­net­work of a DNA mol­e­cule. And all this in a non-inva­sive way.”

Doc­tors are very inter­est­ed in this type of tool because there are very few effec­tive tools for this type of dis­ease today.

The next step is med­ical diag­no­sis – although at the research stage it is more appro­pri­ate to talk about dis­ease clas­si­fi­ca­tion. “Among all the types of pul­monary fibro­sis, will we be able to clas­si­fy, quan­ti­ta­tive­ly and objec­tive­ly, which one the patient in ques­tion has?” the researcher asks. If the MΞDISIM team suc­ceeds, the doc­tor will only have to see the behav­iour of the dig­i­tal lung to deter­mine the appro­pri­ate treatment.

The med­i­cine of tomorrow

More­over, such a tool also has a mul­ti­tude of oth­er ben­e­fits in the health sec­tor, and pro­fes­sion­als in this field are aware of this. “Doc­tors are very inter­est­ed in this type of bio­med­ical engi­neer­ing tool,” adds Mar­tin Genet. These tools are intend­ed to facil­i­tate the work of the doc­tor. Today, there are very few effec­tive tools for this type of dis­ease. Espe­cial­ly since, if this “fun­da­men­tal aspect” of the research were to val­i­date this “vicious cycle hypoth­e­sis”, a dig­i­tal twin would give the doc­tor the pos­si­bil­i­ty of pre­dict­ing the poten­tial devel­op­ment of fibro­sis in a patient, and there­fore, per­haps, of try­ing to pre­vent it. 

Final­ly, one of the major advan­tages of the dig­i­tal twin in the med­ical field lies in the pos­si­bil­i­ty of opti­mis­ing the treat­ments admin­is­tered. Our own organ can be mod­elled in a sim­u­la­tion that repro­duces its func­tion­ing and its inter­ac­tion with oth­er ele­ments, which are them­selves mod­elled. If one of these ele­ments is the drug to be admin­is­tered, the chem­i­cal influ­ence on the organ in ques­tion can be sim­u­lat­ed. In this way, this dig­i­tal twin of our organ will give the doc­tor a quan­ti­ta­tive indi­ca­tion of the effec­tive­ness of the treat­ment to be pre­scribed (anti-fibrosant, anti-inflam­ma­to­ry, etc.) while allow­ing them to mon­i­tor this treat­ment in quan­ti­ta­tive terms. 

“We hope to be able to use the mod­el to describe the evo­lu­tion of the dis­ease,” sum­maris­es the researcher. “Then we can opti­mise and auto­mate this tool as much as pos­si­ble, so that it can be eas­i­ly used in the clin­ic. The dream would be for the tool to be inte­grat­ed direct­ly into the hos­pi­tal’s soft­ware pipelines.”

Pablo Andres
1*LMS: a joint research unit of CNRS, École Poly­tech­nique – Insti­tut Poly­tech­nique de Paris

Our world explained with science. Every week, in your inbox.

Get the newsletter