202 1 this year, it is not difficult to keep track of your heart rate at any time.
On June 25th, Apple's Apple Watch was finally approved by the US Food and Drug Administration for listing ECG and atrial fibrillation monitoring medical equipment. In fact, in 20 18, Apple's Apple Watch Series 4 already had ECG detection function. However, because ECG belongs to the function of medical devices, Apple failed to pass the certification at first and had no choice but to "lock the area" in China.
Apple is not the only company that has entered the field of wearable device health in China. Foreign manufacturers have Fitbit, and domestic manufacturers have Huawei and Huami. The current smart watches and bracelets of these manufacturers can monitor data such as heart rate and blood oxygen, and may even "measure blood sugar without needles" in the future.
Manufacturers can make such efforts because the market in related fields is considerable: according to market statistics, by 2025, the market value of medical wearable devices will exceed 48 billion US dollars.
It sounds like the "health bracelet" market is booming. However, consumers who wear bracelets to monitor their heart rate may not have considered a question: Who is the dominant force of data?
0 1 Same heart rate, different data.
The origin of the problem comes from a statistical study. Responsible for this research is JP Anella, an associate professor at Harvard School of Public Health.
In the academic field, he generally does not introduce consumer products such as Apple Watch, but uses laboratory equipment for research. However, he recently cooperated with a hospital to collect data with Apple Watch.
Therefore, he became interested in the "heart rate bracelet": both manufacturers and researchers knew that there was something wrong with the data collected by the device. He and the team also want to see how big the data problem is.
The research team collected the heart rate data from the end of 20 18 to September 2020, and exported the data through Apple Watch: the first time was September 2020, and the second time was April 20021year. In other words, because the original data has not changed, Apple Watch has no problem in processing the data, and the results of the two data processing should be very close.
However, the result of the experiment is surprising: the coincidence degree of the two data is not so high on the premise that the original heart rate remains unchanged. The yellow curve and the blue curve "play separately", and it is impossible to see the heart rate of the same person.
If a group of data is relatively aggregated in terms of the degree of data dispersion, it is considered to be "closely related". Another set of data is everywhere, "flying yourself." When the two sets of data are put together, there is almost no correlation.
According to Onnela's own blog, "the results of two sets of data may be the most obvious representative of this deviation."
Your heart rate, the algorithm is in charge?
The same set of heart rates, the same Apple Watch, why are the output results so different? The answer is simple, algorithm.
In the traditional heart rate measurement, collecting data is simple: the patient connects the electrodes and the equipment obtains the electrocardiogram. The result of ECG is raw data. Without algorithms, there would be no AI. What results are measured.
But when it comes to the bracelet, the rules change: after the smart bracelet is tested, it will not be exported immediately, but will be analyzed and filtered. Researchers are exposed to "optimized" data, which will be biased compared with the actual heart rate.
Simple "optimization" is not the whole problem. The analyzed algorithm will also be "optimized" into "one every day": in previous research, Onnela said that wearable device algorithm is a "black box": device manufacturers only know the algorithm is updated regularly, but researchers simply don't know how the algorithm counts data. Resulting in the lack of comparability of output results.
The existing results and possible concerns made Onnela give up collecting data with consumer wearable devices in subsequent research. He also implicitly said that the "black box" of the algorithm is a "continuous challenge" for researchers.
Olivia walch of the University of Michigan put it more directly: Although she also studies wearable devices, she lets her research team use raw data directly. Because she studies sleep monitoring, she needs long-term follow-up, and the experiment cost is high. If the results are output by the "smart bracelet" algorithm, the research will start again because of the version change.
From walch's point of view, even if she can accept the algorithm update, she can't know the change in advance: enterprises have no reason to specifically inform researchers that the algorithm has changed, but because of product updates, enterprises often take the initiative to update the algorithm.
For rigorous research, the data obtained by frequently changing the rules itself is not trustworthy. For the application of health monitoring, Apple Watch should also provide continuous and stable medical data, which is obviously not done.
03 Use deviation, "bracelet" is not intelligent.
In fact, the "smart health monitoring" equipment represented by Apple Watch is not reliable from the official recognition.
20/kloc-in September, 2008, Apple announced that the electrocardiogram (EKG) and heart rate monitoring functions of Apple Watch Series 4 had been approved by the US Food and Drug Administration (FDA).
However, the FDA's words are interesting, because the FDA's rating of new devices is divided into three indicators: publicity, license and approval.
Publicized products do not need professional review by FDA, and the standards are the most relaxed. The products that need to be approved have great use risks and need a lot of testing and evaluation, which also makes the Class III products that need to be approved only occupy 10% of the device market.
If you put the "smart watch/bracelet" into this system for evaluation, you will find that. These devices have a technical threshold and need to be inspected. It's definitely not enough to publicize them. However, the function of "monitoring heart rate" does not go deep into the treatment of diseases, so the use risk is very low. Generally speaking, permission is more appropriate.
The positioning of "license plate" also represents the product dilemma of wearable monitoring equipment: it is not simple to produce and unreliable to use.
Take the monitoring function of electrocardiogram and atrial fibrillation promoted by Apple and Huami as an example. The reason why manufacturers promote this specific field is that under the current technical conditions, the "smart bracelet" can only realize single-lead ECG monitoring. Compared with clinical 12 lead, the monitoring method is "sloppy" and cannot give accurate data.
In the promotion of Apple Watch related functions, Apple can only say that "the data is for reference only". Telling consumers to "master health" but actually saying "if you want to master health, please find a doctor yourself" is playing with consumers' expectations.
In addition, wearable devices are still relatively vague in experience. The dual attributes of "digital equipment" and "medical function" make consumers' demand for such products "convenient and accurate".
However, consumers who use "smart bracelets" are not professional doctors and have different habits, which leads to the lack of reasonable standards for product experience: too tight adjustment of bracelets and strenuous exercise when going out will lead to "health warning" of bracelets. The value of "for reference only" is even more difficult to talk about.
Intelligent monitoring, first of all, we must set rules.
Whether it is "data screening" or "for reference only", wearable medical devices represented by smart bracelets are facing the same problem: how to provide consumers with truly convincing products with professional standards under the premise that the industry is still in the early stage of development and technical conditions are still limited.
At present, the wearable medical equipment industry, whether it is a traditional medical equipment manufacturer or a new digital equipment enterprise, wants to grow wildly in the early stage of industry development, thus occupying the market. Last year, 14 wearable devices were approved by the FDA, and 18 devices in China were certified by the Food and Drug Administration, equivalent to the sum of the products approved in the previous three years. This shows the popularity of wearable devices. Huawei, Goer and OPPO have all come to an end, and the industry's heat has soared.
No matter how "digital" wearable devices are, the classification of practical applications is still "medical devices". Since medical health is involved, products should be managed according to the standards in the field of medical health. However, from the market of related products, whether it is data collection or actual use, too many "final interpretation rights" of wearable medical devices are in the hands of enterprises.
In such a market, it is obviously not enough to rely solely on enterprise self-discipline to formulate industry standards. Relevant departments are also required to introduce special industry standards according to the current situation of the industry. In 20 15, FDA classified wearable health devices as "general health" devices, and formulated relevant regulations. In contrast, at the end of 20 17, the Food and Drug Administration issued the Guiding Principles for Technical Review of Mobile Medical Device Registration. However, the specific implementation of these rules still needs to be further refined and clarified.
Consumers who use smart bracelets want to monitor the peace of mind brought by health, not the high-intensity competition in the digital industry. If enterprises are addicted to updating products and algorithms and ignore the real needs, then they need a real "education". Because the data representing the health of consumers, only consumers have the final say.
Source | scientific and technological strength