A face only a mother could love? Facial Recognition Software for Donor Matching

November 1, 2012Carole 8 Comments »

One of the most surprising things I saw at the ASRM 2012 annual meeting was a new service which may revolutionize how patients pick sperm donors. When picking a donor, patients are given lots of information: race, blood type, height, hair color, eye color and maybe some information about years of education, hobbies etc. But to preserve the sperm donor’s anonymity, patients don’t typically get to see what their donor looks like. Of course there are exceptions to this rule; some banks show baby pictures of donors and some (relatively rare) donors “go public” with a video or adult pictures to help patients decide which donor to pick.  However, the vast majority of donors prefer to be and stay anonymous.

How can the anonymity of donors be preserved, yet allow patients to pick donors based on facial features?  The company Precision Donor offers patients a new tool to help them chose their sperm donor by generating something called an Eigen Map, defined as “a matrix or vector expression mapping of an existing array of data”. Umm, yeah, the important thing is that this map is a set of mathematical equations generated using facial recognition software –up to now only used by the Federal Departments of Justice, Defense and Homeland Security for identification of wanted persons. Faces that have similar facial features and bone structure generate similar Eigen maps and so can be used as a basis for matching donors.

For instance, if a patient is looking for a sperm donor that resembles her husband, she could submit her husband’s photo for analysis and find “faced-matched donors”  whose Eigen Map was most similar to her husband.  The ten best matches are offered to the patient/couple and she/they can then choose their donor based on secondary factors. Of course, if she wants her donor to look like a celebrity she could submit a celebrity photo for analysis and matching (couple’s counseling not included).

How does the analysis work? You can visit the Precision Donor  FAQ on their website for the details but I can give you a quick overview here. Basically, when an image is submitted to Precision Donor, they analyze the image using three different kinds of facial identification/recognition software.

The first software analysis pulls the part of the image that is “face” away from the rest of the image based on recognition of 80 nodal points that are common to every face such as the distance between the eyes, nose width, depth of the eye socket, jawline distance and cheekbone structure which apparently do not change throughout a person’s life regardless  of age of the patient (baby vs. adult)  or weight gain (or loss).  I interviewed Melissa Bachman, Marketing Director at Precision Donor for this post. One of the questions I had was what sort of facial issues can interfere with an accurate reading. She said that if a person had a severe facial deformity that changed the underlying bone structure or plastic surgery, the Eigen map would not be an accurate reflection of the person before the accident or before the surgery. But if the deformity were limited to one side of the face, the normal half could be mirrored by the software to get a better Eigen map for matching.

The second pass software (Surface Texture Analysis)  takes the analysis one step further by analyzing and distinguishing  between lines on the face, facial hair, skin pores and skin texture. According to the Precision Donor website, the software analysis is so detailed that it can identify and quantify the facial differences between identical  twins.

The third software uses something called a Vector Template to, according to Precision Donor’s website,  “confirm the subject and validate the facial recognition. The software identifies key features of the subject and extrapolates them to underlying bone and tissue structure. Using this method it can generate a 3-D “wire-model” of the subject’s complete face and verify the results of the first two processes. Using all three methods produces a match that is both accurate and precise“.   This resulting photo is definitely not your father’s Polaroid! Here’s a picture showing an example of the resulting wire model,  copied from Precision Donor’s website.  The patient doesn’t see the wire model, but receives a list of donors that have similarly generated wire models (a similar Eigen map) so look similar.

I still have a few questions about the development of the software for matching that probably  could be better addressed by their technical folk. How was the software validated against actual faces? At some point in development, were the original photos and matched photos compared by visual evaluation to see if subjective comparisons supported the Eigen map analysis results?  Another question I had was if this software is so precise and so discriminatory, could we actually find ourselves matched to a cousin who happens to be a sperm donor? Of course, people in the same family aren’t all cookie cutter images of each other so that probably is a non-issue but it made me wonder about the  genotype to phenotype relationship and how strict that relationship was for genes that create facial structure. Ms. Bachman mentioned that there had been interest at the ASRM meeting from gene expression scientists who also wondered about correlations between facial maps and gene expression so expect even more applications of facial recognition software going forward.

Another question I had was how good does the quality of the original donor or match photo have to be? Astonishingly, most driver’s license photos are good enough, according to Ms. Bachman. Remember facial recognition software was developed for entities like Homeland Security who use driver’s licenses routinely for identification purposes. Another interesting fact is that the company founders tried to use this software years ago in 1999 but by the time they were ready to go, the terrorist attacks of 9/11 resulted in the government labeling the software  as “Classified”, removing it from public use and delaying the project for eight years. Two and a half  years ago the software was declassified and the software company that developed it was able to license it to Precision Donor for their patented use in donor matching.

How does this work for patients? Precision Donor charges patients $200 for 30 days access to their web portal which delivers a list of the top ten Eigen map matches to their submitted photograph.  Patients can access this service through either their physician office or directly through one of the sperm banks. Ms. Bachman informed me that to date, Precision Donor has signed up about 50% of the domestic sperm banks as partners in this service and is fielding inquiries from international banks as well.

Physician offices can also sign up with Precision Donor so that their patients can access the service through their doctor. If the patient accesses the Precision Donor site through their physician’s office, they are more likely to have more donors to choose from since most offices already give patients multiple banks to choose from. Of course, your physician office may specifically exclude some banks that they have found to be sub-optimal to deal with or promote other banks that they have a good history with. Alternatively,  you can go directly through the sperm bank’s website, but you will, not surprisingly, only be able to find donor matches from their own inventory of donors.

In either case, the patient/couple provides a photo they want to match to either their physician office or the sperm bank.  The image is uploaded to Precision Donor and converted into an Eigen Map.  Pictures of donors or of partners are not stored at Precision Donor, just the mathematical equations- Eigen maps- that describe the face so both patients’ and donors’ identity are not revealed to Precision Donor. The patient is given an anonymous Precision donor ID and password which they can use to log into Precision Donor’s site to see which donors most closely matched the photo they submitted for matching.  On the precision donor site , the patient sees a page with a list of donor icons that have additional info that was provided by the sperm bank for each. For example, you could find out height, blood type, hair and eye color, race, religious affiliation if any, and level of education. You can delete or “heart” your choice and can access the rest of the sperm donor information (and order a specimen) by clicking on the donor icon via a url link directly to the sperm bank’s web page.

When is this going to be available to patients? According to Ms. Bachman, the target date for the role-out for sperm donor access is Jan 1, 2013.

So what about egg donors? Interestingly, this project was first envisioned by the Precision Donor founders as only something to help find better sperm donor matches. One of the founders is a female urologist dealing with infertile males who had run out of all other options for parenthood except sperm donation and she saw first-hand the difficulty patients had in choosing donors that would resemble the male partner/husband based on the limited donor information available at the time. At the ASRM roll-out, Ms. Bachman says they were astounded by the number of egg banks who were also interested in this facial-recognition approach as a tool to help them match donors with recipients. Ms. Bachman estimates that the role-out for egg donor matching may take at least six months longer  (maybe by April 2013?) than the debut of Precision Donor for sperm donor matching.

Is this a fad or revolutionary? I think it could be a game changer because if we are honest about it,  most of us really appreciate seeing a likeness to ourselves and our partners in the children we have together. Now, that doesn’t mean that we can’t grow terribly attached to and desperately love kids that don’t look anything like us –hello, adoption works and works well- but it doesn’t seem to hurt and may help bonding if the kids look like the parents.  If we are looking to match  a donor based on facial appearance, this is the first tool that provides an objective- not subjective-  facial match while preserving donor’s anonymity.  It’s a little odd to think that patients can benefit from the same facial recognition software previously used to identify terrorists but there you have it.

I suppose if better donor matching based on facial recognition makes patients/couples feel better about selecting a donor that looks like one of them, and if that  similarity further encourages their emotional connection to and acceptance of the child they have then the most important thing is achieved through better donor matching – that every child should be a wanted child. Ummm, that is “wanted” as in desired, not, ahem, as in Top Ten Most …  you know what I mean. 😉



© 2012, Carole. All rights reserved.

8 Responses to this entry

  • jen Says:

    How cool is that! Thanks for sharing!

  • marilynn Says:

    OMG. They’re trying to make it easier to lie to the outside world if not to the born individual who must shoulder this incredible burden. This is pretty sick stuff. Look like the child your buying or buy one that looks like you? Pay a parent that looks like you to abandon their child. Pay someone that looks like you to get your wife pregnant and then pay him to go away and never come back or pay him to be willing to be known 18 years after you’ve isolated his child away from all their paternal relatives under a false identity.

    Has anyone heard of step parenthood? Its pretty cool. The husband gets to raise his wife’s child but the child does not loose support or contact with or their identity as part of their actual paternal line. Ultimately it is the fault of the child’s father who is the donor for creating this mess. If he does not want to raise any offspring all he has to do is not have any offspring. He should not be going around mating with strangers making babies he has no interest in raising. He’s tearing his family apart none of his relatives can avoid inbreeding after he donates in fact even he can no longer avoid inbreeding due to donation. Not only will he forever be avoiding children and grandchildren but he has no earthly idea who he is making babies with – could be as close as a 3rd cousin or closer which is close enough to cause birth defects. So foolish and cruel. And his children how are they suppose to feel being told that they were so wanted by these people that went out and hijacked someone else’s kid? Of course they will feel wanted they were bought! They’ll feel just like a fashion accessory. And they’ll feel rejected by their father. I know they think that calling him a donor will make having him give them up not feel like abandonment but how could it not? Everyone has bio parents and theirs signed an agreement that he would not seek contact or custody of his offspring when they were born how is that like just giving up some sperm. Facial recognition my eye. If you were just selling sperm and not your children why would anyone care what you looked like? Why would they build wire frame models of your face if they were not trying to buy the donor’s body so his offspring would be born in captivity under the thumb of those that want them. Like little birds in gilded cages wire frame model cages.

  • Carole Says:

    Hi Marilyn,
    Transparency regarding the donor and use of facial recognition software to find a donor are not mutually exclusive. Intended parents can choose donors open to a future connection. Intended parents more frequently than not, tell their children about their biological origins. Using the donor number alone, kids can find their half-siblings using the Donor Sibling Registry. Your condemnation is very harsh and unkind. Phrases like “born in captivity” are not helpful for dialogue. You’ll be taken more seriously in your efforts to improve the lot of donor-conceived children if you remove the vitriol from your speech.

  • Fertility Lab Insider - Lessons learned from over fifteen years of working inside fertility labs. » Blog Archive » Creation of “virtual” children identifies risky sperm donor genomes for recipients Says:

    […] based on a shopping list of characteristics is not a new phenomenon. Recently, I blogged about a company which uses facial recognition software to find look-alike matches to a preferred male face submitted by the recipient.  All of this information is useful for […]

  • Audrey Bailey Says:

    Fantastic Piece of writing

  • Isaac Morgan Says:

    I hardly ever comment on these articles, but I thought this on deserved a thumb up

  • ios news Says:

    Good post! We are linking to this great article on our site.

    Keep up the great writing.

  • pdf Says:

    Ӏt’ѕ going tߋ be еnd of mine daƴ, except before end I am reading this greeat piece οf
    writing tօ improve my experience.

Join the discussion