IVF: the 2 Paths to a Million Babies a Month (AI is one)
2022 was IVF software’s coming out party, the AI in IVF world tour.
In New York, Boston, Chicago, Indian Wells, Milan, Dubrovnik, Anaheim — at home, at ESHRE, at ASRM, MRSi, PCRS and, most notably, at the first AI in Fertility meeting, a community of engaged, brilliant and committed people came together: Michelle and Don from Australia, Daniella from Tel Aviv and Yael from Haifa, Christy and Dan from Toronto, Paxton and his crew from California, Alejandro from Mexico, and Gurjeet and Sarthak and everyone else in the 32 companies of Abigail’s AI In IVF industry map, with Nikica and Christina putting us all together in one room.
We have brilliant people attacking the problem of inefficient IVF delivery, building from a foundation of ten million babies in forty years towards a goal of a million IVF babies a month, affordable, accessible, safe IVF — that you only pay for when it works.
A million babies a month, in a marketplace where you pay for outcomes, not cycles.
What’s our next step? Hardware.
In my last piece I listed the most common AI projects presented to us for funding:
1) more precisely predicting outcomes
2) stimulation protocol selection and making daily medication dosing decisions
3) choosing the day of HCG (or LH) trigger and time for egg retrieval
4) determining whether to use IVF or ICSI for fertilization
5) choosing the specific sperm cell when ICSI (intra-cytoplasmic sperm injection) is performed
6) choosing when to perform the embryo transfer
7) picking the best embryo to transfer
Each of these represents an incremental step towards upgrading the existing, somewhat pedestrian engineering of IVF to match the superb science of IVF. And, by and large, these these are software solutions.
Software is only as effective as the hardware into which it is employed, and we need to build our IVF Macs to run our emerging ivfOS; every AI solution creates a hardware need. If we want to harness data to scale aggregate IVF throughput by 10x or more, we need to build engines that can run on digital fuel.
So matching our AI solutions to hardware needed to truly exploit them creates a new shopping list for IVF innovation.
Examples:
Predictive analytics-based clinical decision support can scale IVF delivery only in combination with a closed oocyte retrieval and vitrification system to move the first half of many IVF cycles out of big box IVF clinics.
Optimized sperm selection needs to be matched with robotic fertilization.
Enhanced embryo development observation, both static and dynamic, coupled with genetic and epigenetic embryo diagnostics, both invasive and noninvasive, can only improve IVF throughout if they are matched with mechanized embryology that can digest these extra inputs without slowing down specimen, and patient, throughput.
We need to move away from exquisitely trained embryologists walking across rooms with petri dishes balanced on their palms while breathing activated carbon pre-filtered, chemical filtered, photo catalytic conversion and HEPA filtered air (thank you Giles) and towards embryology versions of DaVinci Robots, closed-system technologies that increase the positive outcomes per embryologist, fully employing their knowledge and training while minimizing the effects of fatigue and inter-operator outcomes differences.
Within each of these broad projects are dozens of smaller steps, each of which needs its own level of optimized decision-making: designing filters with just the right porousness, calculating optimal pressure differentials to effectively move cells through micro-fluidic chambers without damaging them, selecting for optimal wavelength, filtering and magnification for advanced oocyte, sperm and embryo assessment by imaging.
We’ve made some great strides already on the hardware side, some have regulatory clearance and are in use (robotic and automated management of cryo specimens, dynamic embryo monitoring and assessment of embryo development), more in development stage, employing automation, advanced optical systems, robotics and micro fluidics. We have the brain power and raw, basic biotechnology and bioengineering Lego pieces with which to get this done.
How about funding? My main concern is venture capital’s preference (outside of the IP-protected world of therapeutics) for SASS-type approaches to business-building. Yes, I understand the attraction of 99% gross margins, no need for physical inventory and annuity revenue streams, but a bear quietly writing software in the woods gets us no closer to a million babies a month — until that software directs a pipette or a laser or identifies an expression profile or lets a lens look an embryo and think “I recognize you from when you were a 2pn.”
That said, the closest comp for a robotic-assisted embryology system is a publicly trading medical device company with almost seven billion dollars in revenue and a ninety billion dollar market cap.
Innovation in IVF: the people, the platforms, the technology, and most importantly, the problems we seek to solve and they people whose lives we hope to touch make it the most exciting place in healthcare.
