Special Edition: A Toilet And A Neural Network
You didn't know you needed this. And you probably don't.
If you’re enjoying the newsletter, do me a solid and shoot this over to a friend or healthcare slack channel and tell them to sign up. The line between unemployment and founder of a startup is traction and whether your parents believe you have a job.
The last couple of posts were really long, but I promise they won’t all be like that. Sometimes your boy feels the need to go off.
As a palate cleanser, I thought I’d write a short post about this paper. If you’re uncomfortable with poop, I suggest you stop reading now.
I am also sorry if you just signed up for this newsletter and it’s the first one you’re reading. This is what happens when I have to stay in the house for a month.
This is the kind of paper we need in the middle of the pandemic. A paper about the smartest toilet that’s ever lived, with a twist.
It starts pretty much like every other paper about a new health monitoring device. With a comparison to a jet engine.
Instead of precision medicine, the rapidly evolving field of precision health has expanded efforts in the prevention and early detection of disease through risk-tailored longitudinal monitoring. A similar idea is already realized in the aircraft industry, which implements continuous monitoring of jet engines with hundreds of sensors to prevent engine failure. Unfortunately, in contrast to the aircraft industry, the fact that a normal adult in the United States visits a healthcare provider fewer than four times per year implies that surveillance of the human body is limited, infrequent and possibly insufficient.
Very normal comparison, definitely a necessary sentence.
It continues on, talking about the clinical utility of urinalysis and stool samples. Of course it would be great if we could get more regular samples. Of course it would be awesome if we could intervene earlier using urinalysis data. Of course it would be better if we didn’t have to rely on patients remembering the characteristics of their log and map it on to this insane diagnostic scale that the British made and named after their least favorite city.
Now you obviously have to talk about why this hasn’t been explored yet or why previous attempts at smart toilets have not achieved their goals. And of course, the main barrier has been EHR integration. Because that’s what it always comes down to. Classic Epic, preventing innovation.
A Japanese company, has pursued the smart-toilet system since the mid-1980s and has published several patents on smart-toilet implementation. One of its latest products, announced in 2008, however, lacks clinical utility and is not readily integrated with a user’s electronic health record (EHR) system. Moreover, the major measurements are simple health data, such as urine temperature, diet, body fat and weight, which, unfortunately, rarely provide clinically actionable information. Moreover, with a listed price of US$6,100 per unit (with complete installation), it is probably not affordable for common household use.
The Identification System
Now we get to the true gem of this paper. The prototype itself.
I want you to look at this and tell me what you think and if you see anything interesting.
I’m legitimately floored. How on EARTH can this toilet do that.
A F***ing Cloud. Based. Health. Portal. By jove, they’ve pulled it off.
No I’m obviously talking about the analprint scan. Someone in a lab decided that the best use of a convolutional neural network was to identify a person based on their butthole. There’s actually an entire section dedicated to how accurate this method is.
I’m very glad they decided to put sample images. This was definitely necessary.
But let’s take a look at why specifically they decided to do this.
User identification in the toilet system is crucial as the system is expected to be shared with housemates. Over time, it will eventually be integrated into the EHR, so that every toileting event will be recorded and associated with the user. To accomplish this, we used two methods of biometric identification. First, we designed a fingerprint scanner embedded in the flush lever….the ROC curve was generated, and AUC was 0.95 (Fig. 5a).
So we’ve established that fingerprint scanning works very well, but I guess we need a second level of identification to really ensure that it’s the right person?
To ensure EHR compatibility, an additional method of biometric identification was used. Although the fingerprint scanner installed on the tank lever provides reasonable user identification, some scenarios may be misleading in an identification process. For example, if someone other than the original user flushes the toilet, it causes confounding errors in identification. Furthermore, some industry level smart toilets are already equipped with automatic flushing without the flush lever and, therefore, require different identification methods. Another identification method—utilizing the analprint, which is user-specific—was therefore designed and implemented into the toilet system. A scanner was installed to record a short video clip of the user’s anus.
Ah yes, we need something else other than a fingerprint to identify them - there’s only one possible way to do that. Not voice recognition, facial recognition, or any of the other sane methods. We’re going to do it the RIGHT way. The EMR will accept nothing but the truest form of identity authentication. The anus.
This team is clearly very worried about the possibility of someone masquerading as another person while they’re pooping and decided to flex their TensorFlow skills to solve this issue.
Like this is actually in the paper’s supplement. Look at all the math, statistics, and computing power that’s being used to compare buttholes.
…In order to use all the algorithms to determine the identity of a user via their analprint, we must compare a user’s analprint at the time of toileting to a reference set of analprints for all possible users. When the current user’s print is compared to the references, the lowest [Mean Squared Error aka. MSE]-scoring reference analprint will identify the user.
In our case we calculate the MSE scores for each frame in the video taken, and then take the average.…As can be seen in Supplementary Table 4, in most cases, all algorithms provide a reasonable distinction of the research participant by analysing the pattern of the anus.
All data including images, videos, and user information generated by the toilet system were stored in an online, shared folder. This folder was shared with the host server performing machine learning and user annotations on the generated dataset. After uploading, remaining datasets in a local machine were deleted to protect user’s privacy.
An online…shared…folder.
Hopefully malicious actors don’t find a backdoor entry.
Finally it’s worth noting that to make sure the fingerprinting was accurate, they tested other fingers and assigned impostor scores to the wrong fingers.
To validate the fingerprint identification module, we performed ‘genuine’ and ‘impostor’ judge studies; genuine scores indicate the level of agreement between multiple samples of the one’s fingerprint and impostor scores are obtained by matching samples from different fingers
Did…they also do this…with…
The answer is yes, they did. They took a bunch of pictures of the different anuses, compared them and tried to see if they were unique enough to be mapped to the right person. When put through a convolutional neural network to compare each picture against a picture of participant #1, here’s what you get.
Who is participant #3? Should they know this connection to participant #1? We may have finally found the first non-derogatory definition of “butt buddies”.
Even more interesting in the notes of this chart…
We performed 10,201 comparisons among video-extracted frames for each participant except for participant 4, for whom we performed 10,908 comparisons
Who is participant 4? Why did they need an extra 707 comparisons? I have a lot of questions about this paper.
The participants
I need to know more about the people that participated in this study. All it says is that there were 11 participants. How did they find them? Who are these people? What are their motivations?
I ask this, because I need to know how this is possible.
Only two research participants graded their own stool, whereas the other nine participants declined to review/grade their own stool.
And yet on the same page.
To test the feasibility of this approach, ani from 11 research participants were analysed and the anus morphologies were compared.
Did not know that “ani” was the plural of anus and had to open an incognito window to confirm that.
Either way…why did they decline to review their own stool, but approve the collection of data from their butts for examination and comparison purposes?
Who are these people?
The obstacles
As with any paper, this eventually moves on to the study limitations and obstacles with a prototype like this. There are two major ones, but they’ve addressed them head on.
The first is around people being uncomfortable with this idea. They conducted a survey.
…the overall acceptance level of the toilet system (Supplementary Fig. 8) was within an acceptable range as the majority of responses were either ‘somewhat comfortable’ (37.33%) or ‘very comfortable’ (15.33%) to use the toilet system
<52% of people being comfortable with this device makes me think it’s a stretch to call it an acceptable range, but whatever that’s fine.
Then we get to this nice nugget (pun intended):
The vast majority of the concerns of participants were regarding privacy protection and data security enactment in the toilet system. Interestingly, we observed a statistically significant preference of non-camera-based modules (the urinalysis and fingerprint modules over uroflowmetry, stool analysis and analprint modules combined; P<0.0001, two-sample t-test). The most accepted module is urinalysis, whereas the least favoured module is analprint.
INTERESTINGLY. The authors find it very interesting that their asshole identifying algorithm raises concerns of privacy and is viewed as generally suspect.
The second limitation is around the fact that many people don’t have access to toilets and could probably benefit from this system. This is a real paragraph.
Nonetheless, squatting toilets often coincide with use by populations in which the most urgent need is for sanitation and the prevention of infectious diseases. Health monitoring remains important; however, it is beyond the scope of what is urgently needed by some populations. Globally, approximately 1 billion people have no access to a toilet at all and are forced to defecate in the open. Diseases transmitted through the faecal–oral route or by water, such as cholera and typhoid fever, can be spread by open defecation. They can also be spread by unkempt toilets, which cause pollution of surface water or groundwater. In this case, rather than a health monitoring functionality, sanitation will be a key functionality in squatting toilet systems. Nonetheless, we have developed a mobile phone application to minimize any disparity caused by not having a toilet system for health monitoring (Supplementary Fig. 6).
I legitimately don’t think I could write better satire if I tried. The best part is that in the supplement they describe the app and it turns out…it’s only made for iOS.
Very glad they addressed the issue of Apple users without access to toilets.
And just for good measure.
I’m out.
Lol
I’m going to stop here. I didn’t talk about the fact that the design of the toilet itself makes no sense because you’re going to poop right on top of every instrument. I didn’t even get into the characters used in the diagrams.
But even though it’s funny, a lot of the pain points that this paper describes are real and smarter toilets will genuinely be really helpful to help manage our health in the future. This device follows a bunch of themes I've been hammering for a long time.
Bring preventive healthcare and screening to where patients already are instead of requiring them to come into clinics.
Standardize data capture especially in measurements that are highly subjective with high variance.
Capture new types of data not previously measurable in clinics (e.g. period between bathroom uses, time till first defecation from sitting, etc.).
Capture data more consistently to make it easier to find anomalies.
Build butthole classifiers for neural nets. Okay maybe not that one.
Hope you enjoyed this breakdown, and if you ever come across funny papers definitely send them to me. The next issue will be less absurd please don’t unsubscribe thank you.
Think boi out,
Nikhil aka. Participant #4
If this was forwarded to you, you should sign up! It’ll be marginally less cr(ass) and more analytical in future newsletters.
Amazing
LMAO