Telemedicine and AI

#ai #ml #artificialintelligence

Telemedicine name started to be used in the 1970s [1]. It refers to treatments and medical directions being provided from a distance. These include the use of ICT, medical images, and medical data sharing with secure communications. Use of radiology from a distance, oncological services, gynae, respiratory illness treatment, and consultations all from a distance. With the advent of technology, all these are possible now, but the question is privacy, security, correctness, and assurance of recovery. This includes not just diagnosis but also delivering the required treatments at a distance.

These include medical call centers, e-meetings, e-consultations, e-treatments, and e-follow-up to mention a few. These things are possible but what is needed is the means to deliver the e-solutions to the end users which are accurate, secure, and timely. It makes it mandatory to use robots to accomplish some or all of the tasks and subtasks, and secure communication for the privacy of patients’ data, robust AI for autonomous and semi-autonomous decision-making. Telemedicine does not include Electronic Health Records and concepts such as e-prescriptions.

Telemedicine can be used for the comfort of treatments from the medical provider of choice at the place of choice. However, there are various concerns that revolve around telemedicine, some include accuracy, others such as privacy, and some such as new disease treatments, and newer variants of some viruses or mosquitoes.

To make it possible it is required to make huge data centers so as to infer a treatment or suggestion. The treatments provided by telemedicine can be overlooked by a medical practitioner in a way or may be recommended per se. The use and security of medical data centers are of utmost importance. The key issue is the medical data carries with it the specifics of patients and hence each such record is linked to the description of the patient. And if the patient is not willing to let his data be used for research and development of medical reasons or for telemedicine advances then it is a privacy concern in which patient data have to be protected and debarred from further use.

However, these technologies are being developed in a fast way but neither the patients fully use them to maximum nor are the doctors fully embracing the upcoming change in the distance treatment frameworks. These systems require a lot of data in a dynamic environment, where the environment itself changes with time. With the change in inputs from the environment and the patients themselves, the inputs to AI decision-making may stumble and need to update its parameters as well as need criteria to produce outputs. For example, a patient is undergoing treatment for a kind of virus and another variant comes in, this makes decision-making for an AI system confusing. There are ways to solve this problem too. But these lead to blame being put on either the introducer, the manufacturer, or the software engineer. While none needs to be blamed as this is the case change of input data at a later stage of treatment, while it must be flexible enough to handle such cases.

But in case of some wrong treatment provided at this stage, someone needs to be held responsible. Hence, future systems need to accommodate changes in patients’ requirements, making it a dynamic system. These systems must be guided by medical professionals in the nascent stages. This can make businesses reluctant to do research in this area as blame is put on them. This can be considered as an informed negligence of the telemedicine provider, which should have taken the issue to the nearby medical provider.

Then there are AI-enabled gadgets to guide this process such as bands, wristwatches, and some other invasive techniques to get the measurement to use the round-the-clock e-treatments with the help of AI. However, due to a lot of data being involved in these processes, some of these data collection techniques are put under strict restrictions and are considered as high-risk technologies. But the question is then how would telemedicine grow then? Another issue of using Machine Learning is overfitting where data is tested on training set performs well, but on test set or new data the algorithm may not perform well, this is a issue in many areas of AI.

When the number of classes, and inputs change this may lead to change in the structure of the deep learning model that was employed. In this case this can produce wrong outputs as previous ones were erased out. This needs medical professionals who can understand in-depth technological relations. And here comes the need of continuous learning. This means post-sales and post-installation on patient care to be provided. This means ongoing learning, feedbacks and maintenance.

To be continued…

References

[1] Gallese Nobile, C. (2023). Regulating Smart Robots and Artificial Intelligence in the European Union. Journal of Digital Technologies and Law, 1(1), 33–61. https://doi.org/10.21202/jdtl.2023.2