Giovanni Turra at Copan Group S.p.a. « Machine Learning Times

By: Luba Gloukhova, Conference Chair, Deep Learning World

In anticipation of his upcoming presentation at Deep Learning World Livestream, May 24-28, 2021, we requested Giovanni Turra, Computer Vision, Machine Learning and Deep Learning Engineer at Copan Group S.p.a., a number of questions on their deployment of predictive analytics. Catch a glimpse of his presentation, Deep Learning of Microbiological Analysis inside Full Laboratory Automations, and see what’s in retailer on the DLW convention.

Q: In your work with deep studying, what do you mannequin (i.e., what’s the dependent variable, the conduct or end result your fashions predict)?

A: I’m engaged on WASPLab (Laboratory Automations) for microbiological laboratories. Specifically, I develop and apply pc imaginative and prescient and knowledge evaluation approaches to elaborate enter knowledge (and dataset) like photos of Petri dishes (or related) and non-sensitive knowledge, associated to evaluation from the identical pattern (or affected person). Output outcomes generated could be fairly heterogenous they usually rely from the kind of media (if it permits the expansion solely of particular pathogens, if it even paints a few of them in a selected method, and many others…) and pattern analyzed.

Generally, our fashions detect whether or not one thing has grown, estimate its quantity, till figuring out which pathogen (or pathogens) is being analyzed and to find out the presence of particular options.

We mix these outcomes with others from completely different sources, to generate a conclusive outcome that’s helpful and usable by microbiologists. 

Q: How does deep studying ship worth at your group  what’s one particular method wherein mannequin outputs actively drive selections or operations?

A: Implementation of deep studying methods started in 2013 and since then they elevated their success due to the foresight of managers. They have understood their potentials and appreciated the outcomes. Deep studying has allowed on the identical time to acquire higher efficiency on fields already continuing and to open new fields of microbiological evaluation and automation that beforehand had not been thought-about because of an excessive amount of complexity and lack of appropriate instruments.

To give an instance, the potential of figuring out particular bacterial species from a set of greater than 50 completely different choices required methods and purposes able to being impartial from human analysis, strong to the variability of an natural construction and able to producing in a restricted time a outcome that’s as correct as attainable with out harmful flawed identifications.

Q: Can you describe a quantitative outcome, such because the efficiency of your mannequin or the ROI of the mannequin deployment initiative?

A: An instance is by J. Bayette et al. “Evaluation of PhenoMATRIX ™ expert system for the segregation of urine specimens on CHROMIDR CPSE Elite”. In this examine, the typical choice time per pattern analyzed, utilizing certainly one of our software program developed with deep studying techniques (inside our Artificial Intelligence suite referred to as PhenoMATRIX ™) went from 100 s (with a very guide laboratory workflow) at 1.5 s per pattern (with the automated model). PhenoMATRIX ™ reached this wonderful outcome with out false detrimental errors and with a specificity rating better than 96% in a particularly advanced surroundings for the microbiological lab.

Q: What stunning discovery or perception have you ever unearthed in your knowledge?

A: Technology shouldn’t be a substitute for microbiologists and technicians however as a complement to them. In this sense, even right now, there are circumstances wherein the human behaves higher (when it’s vital to mix native and international information for instance) whereas in different instances the software program can mix the presence of options that additionally they didn’t appear very helpful to us technicians.

We are fortunate to have the ability to apply our approaches on a dataset acquired over the course of a number of years due to the collaboration of many laboratories that already used our merchandise and automations every day: regardless of the excessive diploma of standardization carried out, development conduct (of micro organism) PhenoMATRIX™ processes every day stays extraordinarily variable and for that reason our instruments have to be strong and succesful to guage conditions they’ve by no means seen essentially.

Q: What excites you most in regards to the subject of deep studying right now?

A: For positive, the number of areas the place could be utilized. Being on the identical time a analysis subject and a most important matter for giant tech firms (however not solely), it ensures there’s a steady growth of latest instruments, methods, and research to be taught. In any case, I consider its’ necessary to know which is the perfect software for particular a case to benefit from the accessible assets and reduce the event time of particular person purposes. If I need to speak about a specific subject of analysis, I consider that Transformers characterize, however above all will characterize quickly the Next Big Thing additionally in our pc imaginative and prescient subject.

Q: Sneak preview: Please inform us a take-away that you’ll present throughout your discuss at Deep Learning World.

A: During my discuss, I’ll speak about how deep studying methods inside COPAN are altering the work of microbiological laboratories by dashing up and rising their high quality. Possibilities we generate usually are not solely linked to a discount in prices or a rise in productiveness (quantitatively and qualitatively) but in addition an enchancment within the sufferers’ expertise to acquire accuratest ends in a shorter interval. Even greater than methods carried out, I consider that the actual deal of our merchandise is the decades-long microbiological information mixed with an enormous dataset of scientific instances on which to use any machine studying strategy and concept.


Don’t miss Giovanni’s presentation, Deep Learning of Microbiological Analysis inside Full Laboratory Automations, at DLW on Tuesday, May 25, 2021 from 11:55 AM to 12:15 PM. Click here to register for attendance.

By: Luba Gloukhova, Conference Chair, Deep Learning World


Please enter your comment!
Please enter your name here