(2) How I got into AI: Artificial intelligence in Ophthalmology

I did join Tony Adamis ' team back in 2013.

Prior that I got to know Tony Adamis very briefly.

No more than 5’.

We were closing the cardiovascular department I belonged to.?I was tired and wet, coming to the meeting straight from the airport.?Disease area heads were asked to present their asset Portfolio to the rest of the company. I was representing the nephrology discovery team. Four years of work compressed in a tiny 45’ presentation.

During the coffee break a colleague tapped on my shoulder. The guy was big. Intelligent eyes and a charming smile: “Wow, I was impressed by your work Marco. If you will ever consider joining the clinical development team, give me a call”.?

I had no idea who this guy was. I tried to spot his name on the badge: “Thanks Antony”.

He told me he was going for Tony.

Two years later I called him.

He stopped me and said: “You are the renal guy, right? If you call me, you are looking for a job, aren't you?”

Yes.

Tony is a visionary physician-scientist and a leader. A natural leader.

He was immediately attracted by the artificial intelligence work I was doing on bacteria within chemical biology. He was interested to apply the same approach to clinical development. He hired me in his innovation team. I reported first to Arlette Weichert and then to Dana McClintock .

My first meeting with Tony ended up with this request “Go, talk to people, see what you can do, come back with a plan”.?

I know that high on his agenda was ophthalmology. His first love. His eyes literally shines when he talk about Ophthalmology.

In addition to Tony’s passion for Ophtha, this medical specialty is imaging-intensive. Images contains a lot of features and AI algorithms love features. They are features-gourmand. Seemed to me the right thing to start with.

On the same floor of Tony there were at that time two medical directors. Two great ophthalmologists: Zdenka Haskova MD, PhD and Jeff Willis . On the entire floor, they were the people I was able to connect with. Zdenka and Jeff have been early AI adopters. Since then, we co-authored all my scientific works applying AI to ophthalmology.

With the help of Zdenka and Jeff we started to explore, with the help of the “bacteria crew” ( Fethallah Benmansour , Andreas Maunz and their teams), what we could do in Ophthalmology. We were quite stretched. They were in research but what I asked them to do was related to development. Not exactly their remit. We urgently looked for a dedicated resource capable to eat & breed AI & Ophtha.

This colleague came totally unexpectedly.

Fethallah Benmansour added me as a member to a selection committee for an AI/imaging scientist role. Filippo Arcadu applied for that role. His chances were very low as the role was quite competitive and he had just completed his PhD at the ETH-Zurich. He was not selected. When Fethallah told me the story, I told him: "I felt in this guy something powerful. If you agree, I will look for an alternative way to bring him in”.

He agreed.

The week after I had my biweekly meeting with Tony. I mentioned to him my interest to bring in Filippo. He told me: "If you feel he is the right person, go for it Marco".

Fethallah was supportive.

Filippo had now to convince Jeff and Zdenka.

I met Filippo 30’at the Basel railway station giving him some hints on how to face an interview with American colleagues. He kindly accepted my suggestions.

I will remember that day for a long time. There was a big steam engine locomotive filling the air with a subtle mist. It was like a movie.

Upon green light of Jeff and Zdenka, Filippo was hired as a PostDoc. Together with these colleagues and many more, we started to analyze options to apply AI to available clinical datasets. It was pioneers' time. That work was at the 'bleeding edge' as Dana liked to say.

I setup a kickoff meeting to agree on the way forward. We met in our South San Francisco Genentech headquarter.

The meeting was OK, not great. We discussed many case studies but there were way too many options on the table and, admittedly, I was not able to steer the meeting in the right direction. Which clinical question to prioritize? What dataset to use? How to bring this datasets in-house? How to proceed to the anonymization? Etc…many doubts.

I was very anxious. I asked Tony to participate into this meeting and we were almost at the end without a concrete way forward. Not good.

With the seniority that characterizes him, Tony brought us back to earth: “We have a lot of color fundus pictures and concomitant measurements of retinal thickness by OCT (optical coherent tomography): why you do not start with that?”

This was terrific: we had our entry door into the world of AI application to Ophthalmology.

This clinical question was the first we answered in times where, in clinical development, there were only first attempt to use AI.

As I said, it was pioneers' times.

The Full story is captured in IOVS and it’s readable here:? https://iovs.arvojournals.org/article.aspx?articleid=2727191

Tony continuously inspired us to apply AI to more complex questions: disease progression, treatment response and so on…you can read a tiny fraction of the work we did here, specifically focusing on disease progression in diabetic retinopathy: https://www.nature.com/articles/s41746-019-0172-3

Tony not only make me discovering the power of the application of AI to clinical studies. He inspired all of us to be ambitious about.

Next article will be on the application of AI to video colonoscopy.

An even more complex challenge.

Stay Tuned.

p.s.

if you are interested to the previous AI article (AI in antibiotics discovery) have a look here: https://www.dhirubhai.net/pulse/how-i-got-ai-small-nasty-bugs-marco-prunotto/?trackingId=nsKMcED4RgemogHHyMAJUg%3D%3D

#AI #Ophthalmology #AdvancedAnalytics #ArtificialIntelligence #teamwork #pioneering #clinicaldevelopment