How is self improvement a doable act

How body-hugging and implanted systems are changing medicine and healthcare

Interaction between natural and artificial intelligence pp 75-88 | Cite as

  • Michael Marschollek
  • Klaus-Hendrik Wolf
Chapter
First Online:

Summary

Sensors and actuators in the personal living environment enable new health-related services. Due to the sheer volume of the data, it cannot be interpreted by humans alone. The interaction of patients and doctors with systems that influence people's health based on their analysis and the corresponding feedback poses new challenges for medicine and society. The present article presents several current examples from the fields of rehabilitation, nursing and clinical medicine and shows the possibilities and challenges of the interaction of such assistance systems in the context of the socio-technical systems in which they are embedded. In summary, he discusses the possible consequences of the interaction of the various unequal actors.

keywords

Assistive Health Technologies Telemedicine Health Care Ethics Artificial Intelligence Health Care

1 Introduction

There is now a wide range of sensors and actuators in the personal living environment. These are available in both portable and spatial (ambience) Systems integrate and enable new health-related services. Such technical systems often work as embedded systems, continuously automated and without human interaction in the background - for example to measure physical activity or to recognize medical emergency situations. At the same time, new types of dialog interfaces increasingly allow technically less experienced users to interact with technical systems (Chatbot1Approach) and also enable new services, e.g. B. the support of cognitive performance through context-dependent presentation of information or the preclinical diagnosis support for patients. More and more methods of Artificial intelligence2,3 are used, so that there is increasing talk of intelligent machines.

The intelligent machines described represent new actors in the field of health services and supply medicine. The cooperation of these new actors with one another and with those already involved in the multidisciplinary work in the health sector leads to changes. Medicine and health care have been and are subject to constant change, not least due to the influence of new technology. Technologies such as the microscope, the electrocardiogram and examinations with X-rays and ultrasound have expanded the spectrum of medical diagnostics. The automated analysis of ECG signals has long been routine and relieves the doctor of time-consuming routine activities. Currently, the increasingly routine use of genome and proteome analyzes continues this development. In addition to the tools mentioned to support diagnostics, there are therapeutic tools such as B. pacemakers, hearing aids, cochlear implants and surgical robots, which significantly expand the spectrum of what is possible in medicine and, in the case of implants, become part of the body. The new tools compensate for functional deficits (e.g. pacemaker) and expand sensory (e.g. X-ray diagnostics), cognitive (e.g. EKG analysis) and motor (e.g. surgical robotics) capabilities. It is difficult to foresee how the continuously increasing possibilities for the interaction of increasingly ubiquitous existing and increasingly omniscient and intelligent-appearing and partially autonomous tools will change health care and medicine in the present and will profoundly influence it in the future.

To discuss this topic, this article first presents several current examples from the fields of rehabilitation, nursing and clinical medicine and shows the possibilities and challenges of the interaction of such assistance systems in the context of the socio-technical systems in which they are embedded.

The article first gives an overview of the health-related parameters that can be measured with sensors as well as common application scenarios in the field of non-clinical and clinical care. Starting with the application in diagnostics, through the therapeutic and combined use to knowledge-based systems, the arc of the scenarios spans. Using specific examples, the article shows the new forms of interaction between people and the new technologies, as well as the resulting changes. The article closes with a summarizing discussion of these aspects and gives an outlook on emerging possibilities and risks of the interaction of artificial and human intelligence in medicine.

2 Close-to-body and spatial sensors

Within the last decade there has been sales of portable devices (Wearables) increased sharply (estimate: 2 million in 2018), and in the last five years in particular more and more inexpensive devices (consumer market) have come onto the market that can not only perform simple activity measurements (e.g. step counts), but increasingly also medical parameters such as B. can record the heart rate permanently and without additional electrode belts. The sustainability of the use of such devices, which are currently often used in the wellness sector, has not been well studied. Hermsen et al. (2017) report on a study with 711 participants who were equipped with FitBit activity trackers (Hermsen et al. 2017). After 320 days, only 16% of the participants were still using the portable device. Such portable systems can often be connected to smartphones, so that with them or after the data has been transferred to the server, even large amounts of data can be processed quickly using complex algorithms.

Modern smartphones are ubiquitous and usually have more than 20 integrated sensors for position determination, motion detection, image capture, etc. At the same time, the storage capacity and computing power are so high that there are practically no restrictions in the usability for the evaluation of data. Mobile internet usage in Germany was 68% in 2018 and is increasing steadily (Initiative D21 2019). In 2018, more than 100,000 apps in the areas of health and wellness were already available in just one app store (Albrecht et al. 2018).

Based on their mobility, sensor systems can be roughly divided into mobile systems and spatial, stationary (ambient) systems (Koch et al. 2009). The mobile sensors are further divided into implanted and non-implanted systems. The latter are further subdivided into sensors that require direct contact with the body (primarily the skin) and those that are close to the body, e.g. B. carried in a pocket (e.g. smartphones). Using integrated electrodes or electrodes connected to portable devices, parameters such as electrical signals from the heart (EKG) or skin conductance can be recorded. Furthermore, the temperature, heat convection, body wall movements (ballisto- / seismocardiography), acoustic signals and also chemical / biochemical parameters can be recorded (example: continuous blood sugar measurement with a semi-invasive sensor system). In the broadest sense, mobile sensor devices also include the increasingly widespread devices for professional laboratory diagnostics (mobile point-of-care measuring devices and kits), which range from laboratory parameters (e.g. hemoglobin, inflammation parameters, liver values, electrolytes, kidney function values, etc.) to genetic tests can be used in a variety of ways.

Above all, spatial sensors allow the use of space and the movement of people within individual rooms, apartments, buildings and cities, but also vehicles. In addition, networked room-related sensors can record a wide range of health-related parameters, such as: B. the EKG via sensors integrated in seating or the heart rate via optical sensors.

In the following we show several application examples of how portable and space-related sensors record health-related variables and how they can contribute to solving health or medical problems through intelligent processing of the data.

2.1 Application example: motion detection in medical research

One of the longest established fields of application for sensors in medical research is the objective detection of body movements. This ranges from simple portable step counters, which can only provide aggregated data, to multi-sensory portable devices, which allow detailed gait and movement analyzes, to stationary multi-camera systems in gait laboratories. The researched questions are as broad as the technical approaches. With the widespread acceleration sensors, which are now also integrated in smart watches and phones, the number of steps can be recorded (at error rates (cf. Marschollek et al. 2008)) and the active energy consumption of the person examined can be estimated. By using several, synchronized sensor devices with gyroscopes to record the angular acceleration, the measurement of joint movements under everyday conditions, e.g. B. at work or during sports, possible. In the study Partial Knee Clinics Calliess et al. (2014) used a multi-sensor approach to show that differences in the movement sequence in patients with different knee prostheses occur primarily when descending stairs (Calliess et al. 2014). In addition, the effects of fatigue could be measured. Such measurements cannot be carried out under stationary laboratory conditions.

A major advantage of sensor-based movement measurements is their objectivity in contrast to the widespread survey method (RecallQuestionnaires) on individual physical activity. One’s own activity is often overestimated. For this reason, epidemiological studies have been using sensor devices for some time, e.g. B. already in National Health and Nutrition Examination Survey 2005/2006 and the UK biobank study with 100,000 subjects (Willetts et al. 2018). The availability of such extensive data on everyday activity now allows studies to be carried out on associations with other, e.g. B. genetic data on (Ferguson et al. 2018). Our own studies have shown that using a cluster method with movement data recorded exclusively by sensors, movement types can be identified that have significant differences in metabolic parameters or risk factors with regard to the risk of falling (Marschollek 2016).

The previous examples show how technical diagnostic systems penetrate everyday life and precisely for this reason can provide new and diagnostically relevant information. Diagnostics is not limited to institutions such as hospitals and medical practices. Citizens initially use systems from the expanded health market for convenience functions or for self-measurement with the aim of improving themselves in the area of ​​physical fitness. They use the automated evaluations to control their training or to change their lifestyle towards more activity. The systems support this with appropriate motivation, among other things through playful elements, such as the achievement of set goals or the comparison with other people via social networks. In this application, the technical systems acquire a status that is comparable to that of a personal trainer and companion.

Increasingly more intelligent algorithms allow very individual control of the behavior of people. The value of the information collected for medical care cannot yet be specifically estimated. An automated early detection of health-relevant situations that can only be managed with the help of medical experts would be interesting and important. An exchange of the available data with them should then be easily possible and the experts should present the data that appear relevant to the situation at hand in a form that is easy to interpret. The event-related, individual configuration of existing technical devices and the addition of other diagnostically necessary components create a diagnostically effective living environment.

The interaction of the technical systems with their users and medical experts leads to completely new forms of care. Only when the data collected from the patients are combined with their health information and data from further diagnostic tests and are available to the experts for evaluation and assessment with the support of expert systems can the technical living environment be ideally instrumentalized or designed for recording health.

2.2 Application example: TeleReha

In addition to the previously shown use in diagnostics, the new technical systems can also be used in therapy. One field of application is the motor rehabilitation of patients with musculoskeletal disorders. The possibilities of technical support in this area range from providing assistance to patients and therapists, e.g. B. in the selection of personal exercises through catalogs that allow intelligent searches, through the motivation and guidance of the patients in performing the prescribed exercises to robots that help patients with reduced mobility through precise training loads in relearning movement sequences.

The AGT rehab project (AGT =A.ssisting GhealthTTechnologies) has shown in recent years that equipping patients in their homes with appropriate hardware software systems enables better control of the independent exercise execution (Wolf et al. 2013). The participants in the voluntary studies use a computer system with a depth camera4that can recognize the implementation of physical therapy self-exercises. The technically guided training as well as the automated evaluation and feedback of the quantity and quality of the exercise execution by the computer allow the trainees to carry out all the therapeutically effective exercises more precisely and thus to avoid incorrect stress. The trainees report that the guided training with the system is helpful, not forgetting any exercises and performing the exercises correctly and completely. The participants perceive the feedback of the training successes to the supervising therapist as an additional motivation, and it gives them the secure feeling of being supervised during their home exercises. The aftercare with AGT-Reha is flexible in terms of time and can be carried out in your own premises, which means that AGT-Reha enables some patients to undertake the important, regular post-inpatient rehabilitation in the first place.

The therapeutic systems, of which AGT rehab is just one example, allow new forms of care in addition to the novel diagnostics described above. Through their interaction between patients and therapists, they expand and complement the otherwise direct relationship between human actors. They expand the care of the therapists as their teletherapeutic eyes and arms and thereby also give the trainees a good feeling of being cared for. The largely automated control of the quality and quantity of training should lead to more regular and correct training, which should increase its effectiveness. At the same time, they enable therapists to create new, more flexible working time models through asynchrony. The automated control of the training releases the therapists from routine parts of their work and enables them to concentrate on those aspects that promise increased effectiveness. The information gathered about the actual implementation of the training allows new, more objective insights into the reality of independent home therapy.

One risk of use is that the systems penetrate deeply into the privacy of the users. AGT Reha, for example, interprets video recordings of users in their private living environment. It also collects objective information about adherence5 of users. Even if AGT-Reha does not record the images and thus they never leave the user's apartment, it is easy to imagine misuse of such systems. This would permanently impair the relationship of trust between therapist and patient, which is essential for the treatment. It is also easy to imagine a use of the information about training quality and quantity, which could be disadvantageous for the patient. For example, the prescription of further therapies or the reimbursement of costs could be made dependent on the information.

2.3 Example of use: combination of modalities

While AGT rehab is a delimited system that patients bring into their private environment, the home environment itself, as mentioned above, can serve as a diagnostic and therapeutic space. If apartments are equipped with ambient sensors, these can be combined with other systems (including actuators) to e.g.B. to collect diagnostic information, to evaluate it algorithmically and to make decisions if necessary.

An example of the use of different modalities in real living environments is the detection of falls in apartments. Camera systems and microphones installed in the homes of a total of 28 elderly people at risk of falling recorded for eight weeks each as part of the work on the research network Design of age-appropriate living environments the activities of the people (Feldwieser et al. 2014). At the same time, portable sensor devices recorded the subjects' movements. Using specifically developed algorithms, a computer system installed in the apartment merged the sensor data, evaluated it autonomously and made the decision as to whether there was a fall or not.

This example shows how the personal living environment is changing into an active companion and actor in health care. The use of machine intelligence is particularly useful in the field of emergency detection in the event of such rare, often late recognized events that can have serious consequences for those affected. With built-in sensors, actuators and embedded systems, the apartment itself is becoming an actor in the healthcare sector. For example, it can recognize changes in the health of its residents, suggest suitable countermeasures, offer and arrange support services, recommend a visit to a doctor and, if desired, compile and transmit the data that led to the assessment. In an emergency, the intelligent apartment can recognize this, call for help and open the apartment door for those who come to help.

2.4 Example of use: Knowledge-based systems that are easy to use

Another field in which technical entities change the interface between humans and the healthcare system are dialog-based applications that allow users easy access to medical knowledge that is relevant to them in their current context. A prominent example is the smartphone app Ada, which interactively asks its users about symptoms in a text-based dialog, limits the number of possible diseases and uses the answers to calculate the probabilities of possible diseases. A knowledge-based system created and continuously maintained by experts works in the background. In addition to the available medical knowledge, feedback from users regarding the correctness of the assessment is also incorporated. A quote from the description of the app by the manufacturer on the Internet: "Ada was developed by more than 100 doctors and scientists and already knows over a thousand diseases with several billion combinations of symptoms - from a simple cold to rare diseases." (Appgefahren 2018) .

These types of applications, which enable users to do a kind of self-service and self-care in a medical context, represent a new form of health care. For a number of years now, the proportion of patients who use the Internet to find out about possible illnesses before visiting a doctor has increased. Compared to this form of preparation for (or decision-making about) a doctor's visit, intelligent applications such as Ada promise a much more realistic assessment of the current state of health and thus even more targeted, timely care. Ideally, such intelligent systems relieve both doctors and patients by helping to avoid unnecessary visits to the doctor and thus enabling more efficient and more precisely tailored care for patients who require the assistance of a doctor.

The knowledge-based system on which the application is based also offers AdaDX, an interface specifically developed for medical experts. Ronicke et al. (2019) report on the use for the diagnosis of rare diseases that are often overlooked and therefore often treated late. The authors argue that there are around 7000 rare diseases that are difficult to assess, even for experts with all combinations of symptoms. The application analyzes the case data and offers doctors diagnostic support by calculating the accuracy of the data for different diseases on the one hand, and their probabilities on the other.

Machine intelligence offers cognitive support in a very knowledge and data-intensive environment. The care of patients often takes place under enormous time and cost pressure, which leads to rare diseases being recognized too late. The final decision on the diagnosis remains with the doctor, especially with regard to the therapeutic consequences in each individual case. Here man and machine act synergistically. The doctors play an essential mediating role between the patients on the one hand and the human and machine experts, decision-makers and agents on the other.

In the MoCAB (Mobile Care Backup) project, machine intelligence takes on an advisory and support role in a very similar way for the relatives of people in need of care. A smartphone-based dialogue system works here (Chat bot) with family carers in order to record their specific information and support needs. An algorithm then decides which specific information relatives need in a certain situation and provides them with the corresponding knowledge modules (Wolff et al. 2018). Such systems can help close knowledge gaps and break down barriers between experts and laypeople.

3 Summary and Outlook

Several examples from the field of body-friendly and ambient medical technology systems have illustrated the possibilities and challenges of the interaction of living and non-living entities in the field of health care and medicine. The use of technical systems is also increasingly taking place outside the immediate medical environment and regardless of existing or suspected medical problems. They are increasingly penetrating the private environment of citizens and enabling new forms of health care. The increased capacity for self-care and self-sufficiency made possible and the targeted use of health services has the potential to make the health care system more effective and efficient in the future. As a result of their interaction with technology, patients receive greater responsibility and a better and more informed participation in their health status.

3.1 Ethical issues

In addition to the promising positive possibilities of the emerging technologies, in addition to the new forms of health care, there are also ethically questionable incentives and applications of the initially value-neutral technology. Information about the state of health is one of the most private and therefore most valuable information about an individual. The data collected about a person's behavior also allows deep insights into their private life. These can be used in the interests of the patient to improve the state of health, but improper use is also conceivable. In less free societies, this information can be used comparatively easily to monitor citizens.

Even with the intended uses of the technology, ethical questions arise with regard to the restriction of what is technically feasible. Some of the new technologies can also be used by healthy people. With regard to self-improvement or self-optimization, similar questions arise here as they exist in the context of improving physical and mental performance with drugs (doping). An emerging risk is that the technology shifts the norms that an individual can, should or should meet, thus increasing social expectations and the pressure on the individual. This standardization is contrary to individuality and autonomy that are fundamentally worth preserving. A social discourse about the acceptable limits of what is feasible is necessary here.

3.2 Changes in the interaction in medicine

Wherever intelligent systems are used, they will change the roles of the people who interact with them directly or indirectly. They deepen and expand the objective information about patients and enable them to better understand their health through automated and individualized feedback as well as direct dialogue and thus make them more informed participants in the health care system who can access health services more precisely. For the medical practitioners, this ideally makes their work easier. The health status of informed patients, about whom technical systems have collected, condensed and evaluated a lot of information, can be diagnosed more easily, more specifically and hopefully faster. Technical systems can, in turn, support physicians in this knowledge work, for example by providing context-related knowledge and pointing out the possible presence of rare diseases. Another question that arises from the construction of the technical systems and their decision-making is the question of the traceability of these decisions. The exact function of some systems, which, for example, achieve better results than human experts in the image-based diagnosis of skin cancer, can hardly be understood. Decision-making thus represents a black box. The question to be answered here is whether we are satisfied with the fact that the results of the black box are better, or whether it makes sense to demand that every decision is understood in detail by a person. The so-called responsible data science goes in that direction. On the other hand, it can be argued that experts are also a black box for laypeople and can often hardly explain their decision-making basis to him or her. After all, that plays Gut feeling, which is based on many years of experience and hardly explicable knowledge, also plays a role in decisions made by experts.

Given the current technical development, it is foreseeable that in certain areas of medicine, computer-based intelligent systems make a diagnosis more reliably than the majority of medical professionals. In this situation the question must be asked whether it is not immoral for these doctors to make a diagnosis without technical support.

Footnotes

  1. 1.

    A chatbot is a text-based dialogue system that allows natural language interaction with a technical system.

  2. 2.

    "[...] General term for applications in which machines acquire human-like abilities such as learning, judging or problem-solving." (Deutscher Bundestag 2018, p. 5).

  3. 3.

    From the proposal for the Dartmouth Conference 1956: “The study is to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it. An attempt will be made to find how to make machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans, and improve themselves. "(Wikipedia 2019).

  4. 4.

    In addition to the two-dimensional image, depth cameras record the distance between objects and the camera, thereby facilitating their automatic detection.

  5. 5.

    Adherence describes the extent to which a person's behavior is in accordance with the recommendations agreed with their therapist.

literature

  1. Albrecht, U. V., Hasenfuss, G., & von Jan, U. (2018). Description of cardiological apps from the German app store: Semiautomated retrospective app store analysis. JMIR Mhealth Uhealth,6(11), e11753. 10.2196 / 11753CrossRefGoogle Scholar
  2. Calliess, T., Bocklage, R., Karkosch, R., Marschollek, M., Windhagen, H., & Schulze, M. (2014). Clinical evaluation of a mobile sensor-based gait analysis method for outcome measurement after knee arthroplasty. Sensors (Basel),14(9), 15953-15964. 10.3390 / s140915953CrossRefGoogle Scholar
  3. German Bundestag. (2018). Scientific Services Documents - Artificial intelligence and machine learning. https://www.bundestag.de/resource/blob/592106/74cd41f0bd7bc5684f6defaade176515/WD-10-067-18-pdf-data.pdf. Accessed Jan 4, 2020.
  4. Feldwieser, F., Gietzelt, M., Goevercin, M., Marschollek, M., Meis, M., Winkelbach, S., et al. (2014). Multimodal sensor-based fall detection within the domestic environment of elderly people. Journal of Gerontology and Geriatrics. 10.1007 / s00391-014-0805-8CrossRefGoogle Scholar
  5. Ferguson, A., Lyall, L.M., Ward, J., Strawbridge, R.J., Cullen, B., Graham, N., et al. (2018). Genome-wide association study of circadian rhythmicity in 71,500 UK biobank participants and polygenic association with mood instability. EBioMedicine,35, 279-287. 10.1016 / j.ebiom.2018.08.004CrossRefGoogle Scholar
  6. Hermsen, S., Moons, J., Kerkhof, P., Wiekens, C., & De Groot, M. (2017). Determinants for sustained use of an activity tracker: Observational study. JMIR Mhealth Uhealth,5(10), e164. 10.2196 / mhealth.7311CrossRefGoogle Scholar
  7. Initiative D21 e. V. (2019). D21 digital index 2018/2019. https://initiatived21.de/publikationen/d21-digital-index-2018-2019. Accessed Jan 4, 2020.
  8. Koch, S., Marschollek, M., Wolf, K. H., Plischke, M., & Haux, R. (2009). On health-enabling and ambient-assistive technologies. What has been achieved and where do we have to go? Methods of Information in Medicine,48(1), 29-37. Google Scholar
  9. Marschollek, M. (2016). Associations between sensor-based physical activity behavior features and health-related parameters. Human Movement Science,45, 1-6. 10.1016 / j.humov.2015.10.003CrossRefGoogle Scholar
  10. Marschollek, M., Goevercin, M., Wolf, K. H., Song, B., Gietzelt, M., Haux, R., & Steinhagen-Thiessen, E. (2008). A performance comparison of accelerometry-based step detection algorithms on a large, non-laboratory sample of healthy and mobility-impaired persons. Conf Proc IEEE Eng Med Biol Soc,2008, 1319-1322. Google Scholar
  11. Ronicke, S., Hirsch, M. C., Turk, E., Larionov, K., Tientcheu, D., & Wagner, A. D. (2019). Can a decision support system accelerate rare disease diagnosis? Evaluating the potential impact of Ada DX in a retrospective study. Orphanet Journal of Rare Diseases,14(1), 69.10.1186 / s13023-019-1040-6CrossRefGoogle Scholar
  12. Wikipedia - The Free Encyclopedia. (no year). Dartmouth Conference. https://de.wikipedia.org/wiki/Dartmouth_Conference. Accessed 24 Aug 2019.
  13. Willetts, M., Hollowell, S., Aslett, L., Holmes, C., & Doherty, A. (2018). Statistical machine learning of sleep and physical activity phenotypes from sensor data in 96,220 UK biobank participants. Scientific Reports,8(1), 7961. 10.1038 / s41598-018-26174-1CrossRefGoogle Scholar
  14. Wolf, K.-H., Franz, S., Schwartze, J., Kobelt, A., Borrmann, H. P., Kasprowski, D., & Haux, R. (2013). AGT Reha: Assistive health technologies for medical tele-rehabilitation training. In H. Handels & J. Ingenerf (eds.), Proceedings of the 58th gmds annual conference (Pp. 428-429). IOS Press, Lübeck. Google Scholar
  15. Wolff, D., Behrends, M., Gerlach, M., Kupka, T., & Marschollek, M. (2018). Personalized knowledge transfer for caregiving relatives. Studies in Health Technology and Informatics,247, 780–784.Google Scholar

Copyright information

Open Access This chapter is published under the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/deed.de), which permits use, copying, editing, distribution and reproduction in any medium and format. Provided you properly identify the original author (s) and source, include a link to the Creative Commons license and indicate whether changes have been made.

The images and other third-party material contained in this chapter are also subject to the named Creative Commons license, unless otherwise stated in the legend. If the material in question is not under the named Creative Commons license and the action in question is not permitted under statutory provisions, the consent of the respective rights holder must be obtained for the further uses of the material listed above.

Authors and Affiliations

  1. 1.Peter L. Reichertz Institute for Medical Informatics at the TU BraunschweigMedical Hochschule HannoverHannoverGermany