Infertility affects approximately one in six people of reproductive age worldwide. In vitro fertilization (IVF) — the process of combining an egg and sperm to create an embryo in the lab, and then transferring the embryo into a uterus — can help patients who are experiencing infertility conceive a child. However, many steps in fertility care are time-consuming and inefficient, and treatment plans are often subjective. For example, conventional laboratory tests for male infertility are inaccessible in resource-constrained settings and stigmatized, hindering many men from getting tested.
In 2015, Hadi Shafiee, PhD, a starting faculty member in the Division of Engineering in Medicine at the Brigham wanted to address these challenges using smartphone-based tools and advanced computer programing approaches for medical image processing, which have already revolutionized medical care through rapid image processing. Shafiee’s first idea was to develop a low-cost, simple, rapid smartphone-based assay for male infertility assessment through on-phone semen analysis that anyone could use in the comfort of their own home.
“I was interested in the concept, but unsure if it was a good translational idea,” Shafiee said. “With limited funding and a sense of naivety, I took the leap and decided to launch the project. Encouraged by the positive feedback I received from colleagues at BWH and MGH, as well as fertility experts from the industry, I felt motivated to pursue my ideas. Fortunately, I was able to find collaborators who generously offered their support to test and validate my concepts.”
At this time, Shafiee met Charles L. Bormann, PhD, of the Department of Obstetrics, Gynecology and Reproductive Biology at Massachusetts General Hospital. Bormann, the IVF lab director, helped Shafiee develop the phone-based assay and test it on 350 semen samples. The results of this initial fertility project were published in Science Translational Medicine in 2017.
Building upon their previous collaboration, Bormann and Shafiee embarked on more ambitious projects aimed at tackling inefficiencies and inconsistencies in in-vitro fertilization (IVF) through the utilization of Artificial Intelligence (AI). The primary objective of this project was to develop AI-assisted tools for assessing and selecting patient embryos prior to transfer. This was achieved by leveraging AI algorithms to analyze microscopic images of embryos accurately, reliably, and consistently for assessment purposes.
Shafiee and his dedicated team, including Manoj Kanakasabapathy, Prudhvi Thirumalaraju, and Hemanth Kandula, through their collective efforts, developed powerful AI frameworks capable of assessing embryo and sperm morphology using static microscopic images. They showcased the performance of their AI frameworks in selecting embryos for transfer based on their implantation potential, surpassing the performance of experienced embryologists.
The collaborative efforts led to the publication of over 50 research papers, conference proceedings, and intellectual properties in esteemed journals like Nature Biomedical Engineering, eLife, The Lancet, Human Reproduction, Fertility and Sterility, Journal of Assisted Reproduction and Genetics, among others. The project was a bigger success than anyone on the team expected, garnering global attention and eventually a licensing deal with Fujifilm related to AI-based embryo assessment and selection.
“Collaborating with Charlie helped us to develop and refine new ideas and took us a step closer to clinical translation,” Shafiee explains. “He suggested we look into embryo assessment and analysis, and other steps involved with in vitro fertilization. Working together, we were able to accomplish something we could never have done alone.”
Tackling Challenges in Treatment
Bormann and Shafiee teamed up with Irene Dimitriadis, MD, a reproductive endocrinologist and infertility specialist at the Mass General Fertility Center who helps treat patients experiencing infertility.
“Social norms have led many people to delay their childbearing into their later 30s or 40s,” said Dimitriadis. “But nature never got the memo and continues to give humans the best chance for pregnancy in our early 20s. This is one of the reasons we are seeing more and more people turn to IVF as a potential treatment to achieve their family building goals.”
While physicians at both the Brigham and MGH have decades of experience in providing expert fertility care, the field faces many technological obstacles, from screening semen effectively to tracking specific embryos. The process of completing these highly technical procedures can be variable, with subjective criteria to assess the quality of an embryo or widespread differences in implantation procedures among clinicians.
Dimitriadis explains that her clinical work — evaluating a patient, retrieving their eggs and supporting their treatment — benefits from a strong relationship with embryologists like Bormann and researchers like Shafiee.
Serendipitous Collaborations Take Off
After developing the smartphone-based screening tool to assess sperm quality, Shafiee and colleagues made a 3D adapter to individually image each sperm, which is approximately 6 microns wide. Once this was successful, they tried applying the same approach to eggs and embryos, which were much bigger.
“When this imaging worked, I hurriedly called Hadi to show him the pictures,” Bormann said. “Embryo selection was actually just the beginning — the lowest-hanging fruit that people immediately recognize as a problem in IVF.”
By 2020, they developed imaging systems that could assess embryos’ genetic makeup and choose the highest-quality one available, standardizing an otherwise subjective evaluation process. Equipped with the technology, clinicians can transfer embryos that are more likely to lead to a viable pregnancy.
“If we can perfect this step, clinicians can transfer one embryo and have the highest likelihood of achieving the goal of a singleton live birth,” Dimitriadis said.
After optimizing embryo selection, the team tackled the process of witnessing, or keeping track of, embryos. The current gold standard is labeling petri dishes with identifiers for each embryo: When an embryologist transfers an embryo from dish to dish, they manually keep track of its identification.
“Artificial intelligence can assist embryologists at key steps in the process of embryo transfer, ensuring that the right embryo is always placed in the right dish,” Shafiee said. “Embryologists take great care throughout their work and some commercial platforms use bar codes to scan dishes, but AI can provide an extra layer of safety by witnessing these steps.”
The team found unique morphologic signatures on each embryo that AI imaging tools could track. These signatures were connected to each dish, adding a second layer of protection by allowing researchers to track both a dish and its contents.
Over the last six years the team has published over 50 papers and patents on integrating AI tools into IVF procedures, ranging from assisted hatching technique to quality assessment tools tracking the performance of individual embryologists and clinicians.
“Ultimately, we want improve every part of the IVF process, from start to finish,” Bormann said. “We want to automate the lab’s processes, make clinical judgements more consistent and make the entire operation safer.”
Still, the tools must be tested in clinical trials to confirm their efficacy before earning regulatory approval and proceeding to clinical implementation. Dimitriadis is leading one such effort to launch a randomized clinical trial to transfer embryos selected through machine learning to patients.
“Our specialty is very rigorous and introduces machine learning and AI into a new field,” Dimitriadis said. “The technology has the potential to improve the success of IVF procedures and increase widespread access to fertility care.”