The system, called the Digital Health Navigator, or DiNa, was designed around three linked functions: monitoring physiological and behavioural data, supporting patient activation and self-management, and giving clinicians or care coordinators a shared view for retrospective review. Its Android application supports goals, medication and symptom records, meal logging, reminders and health trends. A web dashboard provides multi-patient monitoring and patient-specific records, while a modular backend connects the mobile interface, dashboard and data-processing services through shared storage and event records. The authors describe this as a system-level attempt to reduce fragmentation, in which a single patient-generated event can inform coaching and remain available for clinical review.

DiNa also includes an embodied conversational agent that speaks with users in voice or text through a rendered digital human. The agent is separated from the software that presents it and can invoke tools to log meals, retrieve history, manage goals, create reminders or query health data. Its context includes active goals, recent logs, wearable summaries and durable memory entries. Educational responses are grounded through retrieval-augmented generation using a curated collection of 147 topics from Healthify and Diabetes New Zealand, with retrieved passages linked to named sources. A lightweight Llama Guard 4 classifier screens incoming requests and generated responses, while additional rules prohibit recommendations to adjust medication or insulin and require escalation for emergencies, severe glycaemia and self-harm. The researchers did not clinically evaluate these guardrails or formally measure fidelity to motivational interviewing and cognitive behavioural therapy techniques.

Ten university-community volunteers—five women and five men with a mean age of 27.6 years—tested the platform in moderated sessions lasting approximately 45 to 60 minutes. None had a diagnosed chronic disease, although each regularly used at least one health or fitness tracking application. Eight of the 10 did not regularly log meals, while most tracked physical activity at least several times per week; mean confidence in learning new health applications was 3.7 out of 5. Participants first explored the application, then completed tasks covering goal setting, meal logging with a photograph, meal-history review, goal-related coaching and access to health trends. Researchers measured task completion, completion time, observed errors and moderator interventions, then collected usability, workload and three five-point acceptability ratings and conducted exit interviews. Reflexive thematic analysis and a second coding pass focused on touchpoints spanning two or more platform components.

Most individually administered tasks reached or approached completion. All 10 participants logged a meal, all completed the meal-history task when it was administered, and all completed the health-trends task when reached. Goal setting was completed by nine participants who received the task. The coaching task was completed by nine of 10, with one participant requiring the facilitator to take over. Some tasks were not administered when sessions exceeded the allotted time: goal setting and meal-history review were each omitted once, while health trends were omitted three times. Although overall workload was modest, mental demand averaged 41.7 and effort 32.7, exceeding frustration at 22.8, temporal demand at 14.3 and physical demand at 9.9. Perceived performance on the NASA-TLX was 38.8. Acceptability was more mixed than the usability score: mean willingness to use the system over the following month was 3.4 out of 5, mean willingness to recommend it was 3.8, and perceived usefulness for supporting health and fitness goals was 4.2.

Video observations revealed what the questionnaire ratings could not. Eight of the 10 coaching sessions required an unmute prompt or facilitator rescue because participants initially did not realize that the microphone was off. Four participants encountered meal-logging mismatches involving food misclassification, incorrect portions or a failed state update; one photographed kiwi was recorded as a mandarin, one cake slice appeared as two meals, and a deletion confirmation was not reflected in the visible log. Across 152 cross-component interaction points, 109 worked or were recovered during the session and 43 remained unresolved. Coach-to-goal interactions worked or were recovered in all 27 cases, while only 14 of 30 touchpoints connecting goals with logging, monitoring or review were successfully completed or recovered. The protocol also did not test a complete chain from logging a meal and setting a goal, through coaching that used both, to review, because meal review occurred before the coaching task.

Five themes emerged from the interviews. High-frequency logging created friction: participants liked photograph-based capture because it avoided manual nutritional calculation, whereas lengthy forms led some to bypass manual entry or delegate logging to the conversational agent. Participants with advanced tracking habits instead wanted ingredient-level precision and calorie visibility. Goal setting was also poorly matched to expectations. Users wanted goals that could vary by day, contain multiple criteria, run for a defined period and be reviewed later, while the prototype often imposed fixed structures. In one session, a request to use exercise for stress was reframed as mindfulness, which the participant rejected. Ambiguous labels such as “Commit,” “Confirm” and “Describe,” along with unclear information locations, further reduced confidence. Subtitle timing, synchronisation and overlap with interface elements also interrupted some conversations.

The conversational coach was approachable for lightweight tasks but often seemed generic, repetitive or insufficiently corrective. Participants valued the ability to describe a meal conversationally instead of completing a form, and they expected advice to draw visibly on their goals, logs and previous conversations. Transcript coding found that about half of coach turns enacted a behaviour-change technique and about one-third drew on the participant’s own data or statements. Goal setting, action planning, goal review, self-monitoring and feedback were accepted more readily than generic information and instruction; generic health content accounted for half of all observed rejections. Participants also expected the coach to disagree when appropriate and explain its reasoning, rather than remaining uniformly affirmative. The findings indicate in-session acceptability of these techniques among volunteers, not evidence that the coaching changed behaviour or improved clinical outcomes.

Technical measurements reinforced the interaction findings. Across five sequential single-user repeats, ordinary REST operations for creating goals, retrieving meal history and trends had mean response times of 34.1 to 47.3 milliseconds. Meal persistence averaged one second because a 4.9-second response raised the mean above its 42.2-millisecond minimum. Text coaching over WebSocket was the dominant interactive delay, taking 3.61 to 11.08 seconds and averaging 6.46 seconds. This latency was consistent with the coaching task having the longest completion time and with participants being unsure whether the agent was processing their speech. A separate burst of requests triggered an HTTP 429 rate-limit response, revealing a condition that clients must handle. The wearable pathway was not exercised in the usability sessions, and the Dexcom continuous-glucose-monitor integration carries an approximately three-hour vendor delay and is therefore used for retrospective pattern analysis rather than real-time alerts. DiNa does not duplicate the regulated Dexcom application’s safety-critical glucose warnings.

These findings position integrated care technology as technically promising but not yet ready for inference about clinical usefulness. The platform’s central strength is that one event can be reused across self-tracking, coaching and coordinator review without manual transfer. Yet that same coupling means a local error can weaken confidence in the entire system. The researchers identify several design priorities: make meal-recognition outputs easy to correct, represent goals as flexible and evolving, expose the user data underlying conversational responses, and display clear listening, processing and saved states. Future testing should evaluate chained workflows, clinician dashboards, wearable synchronisation and concurrent system load with adults who have chronic conditions and varied levels of health or digital literacy. The present mean SUS score should be treated as an early reference point from a favourable volunteer sample, not as an estimate of usability, acceptance or effectiveness in clinical care.

Subject of Research: Formative usability evaluation of an integrated digital health platform for chronic care support

Article Title: Design and formative usability evaluation of a prototype integrated digital health platform for chronic care support

Article References: Nanthakumar, K., Gupta, K., Nia, A. F., Dempsey, S., Saffaryazdi, N., Nassani, A., Broadbent, E., Tawhai, M., & Maso Talou, G. (2026). Design and formative usability evaluation of a prototype integrated digital health platform for chronic care support. Frontiers in Digital Health, 8, 1780895. https://doi.org/10.3389/fdgth.2026.1780895

Image Credits: AI Generated

DOI: 10.3389/fdgth.2026.1780895

Keywords: digital health, chronic care management, usability evaluation, conversational agents, wearable sensors, large language models, health coaching, patient monitoring, care coordination, system integration, Design, formative

The appeal of contactless monitoring is easy to understand. Traditional hospital instrumentation, from finger-clip pulse oximeters to adhesive electrode patches and inflatable blood pressure cuffs, works well enough in controlled clinical settings, but it is uncomfortable for long-term wear, can irritate fragile skin in elderly and neonatal patients, and requires constant replacement and maintenance. Non-contact systems sidestep all of this by capturing physiological information at a distance. Cameras read the subtle color shifts in facial skin caused by pulsing blood flow. Radar transmitters bounce millimeter-wave signals off the chest and detect the millimeter-scale movements of breathing and heartbeat. Wi-Fi routers measure how the human body perturbs ambient wireless channels. Acoustic sensors and infrared detectors add further modalities, and modern machine learning pipelines fuse these raw signals into usable clinical estimates.

What distinguishes the new review is its system-level perspective. Previous surveys tended to examine a single sensing technology, such as radar or computer vision, or focused narrowly on signal processing algorithms. The authors instead characterized every included study across five dimensions: sensing modality, system architecture, the physiological parameter being targeted, the validation setting, and reported performance. They also evaluated methodological quality and, critically, asked how often privacy-preserving mechanisms had actually been built into the systems, an unusually honest question for a field whose primary instruments are cameras and radio transmitters pointed at people in intimate moments of illness.

The results reveal a field of striking imbalance. Radar and vision-based systems dominate the literature, together accounting for roughly three-quarters of all studies analyzed. These two technologies have matured fastest because their underlying physics and computer vision toolchains are comparatively well developed, and because both can plausibly reach the accuracy thresholds needed for clinical triage. By contrast, approaches based on sound, Wi-Fi sensing, embedded sensors integrated into furniture, and mixed-modality combinations remain at an earlier stage of validation, with fewer studies, smaller cohorts, and less standardized evaluation.

Even more striking is the concentration on two vital signs. Respiration rate appears in 83 percent of the reviewed studies, and heart rate in 67 percent, a significant overlap given that many systems extract both from the same chest-motion or skin-color signal. No other physiological parameter exceeds 11 percent of the corpus. Blood pressure, oxygen saturation, and body temperature, all staples of routine clinical assessment, remain largely unaddressed by contactless methods. The reason is physics as much as preference: breathing and heartbeat produce measurable mechanical displacement and periodic optical change that sensors can resolve with modest hardware, whereas blood pressure and SpO₂ estimation from a distance requires subtle modeling of vascular properties that current remote techniques capture only indirectly and with limited reliability.

Perhaps the most sobering finding concerns measurement itself. The review found that heterogeneous reporting metrics across studies prevent direct statistical comparison of performance between modalities. One radar study might report mean absolute error on healthy volunteers at one meter; a vision study might report correlation coefficients on hospitalized patients at three meters under different lighting. Without standardized datasets, common evaluation protocols, and agreed-upon error metrics, clinicians and hospital administrators have no rigorous way to compare a radar system against a camera-based alternative, and the field’s headline accuracy numbers remain effectively incommensurable. This absence of benchmarking, the authors argue, is now a primary brake on clinical translation.

When it comes to deployment, the review’s central message is that context should drive technology selection, not laboratory accuracy rankings. Radar excels in hospital wards and intensive care units, where patients may be covered by blankets, moving restlessly, or positioned unpredictably, conditions that defeat optical systems but not penetrating radio waves that measure motion through fabric and darkness. Vision and acoustic systems, meanwhile, suit structured clinical spaces with stable lighting and predictable geometry, such as examination rooms and sleep laboratories, where a camera’s rich spatiotemporal information can be fully exploited. Embedded sensor designs, built into eldercare beds and chairs, offer a middle path: technically not remote, but requiring no wearable and no user cooperation, making them practical for dementia care and nursing homes where continuous compliance cannot be assumed.

For settings where the stakes are highest, the review points to mixed radar and vision systems as the emerging precision tool. By fusing the robustness of radar with the richness of optical sensing, hybrid architectures can provide redundant, mutually validating measurements suitable for neonatal monitoring and postoperative care, where a single missed apnea event or unrecognized desaturation can be catastrophic. The trade-offs are real: greater hardware complexity, harder signal fusion, and higher cost. But as both sensing strands mature, the fusion approach increasingly looks like the endpoint toward which precision-critical contactless monitoring is converging.

The review does not shy away from the field’s persistent failures. Signal reliability in uncontrolled environments remains the fundamental technical challenge: ambient light changes, multipath radio reflections, other people moving through the scene, pets, fans, and vibrating beds all contaminate the weak physiological signals these systems hunt for. Privacy and ethical concerns compound the problem, since cameras and radar implicitly surveil patients, families, and sometimes entire rooms, raising questions about consent, data ownership, and secondary use that few current systems adequately address. Finally, poor integration with existing healthcare infrastructure, electronic health records, nursing alarm systems, and clinical workflows, means that even accurate prototypes often die at the pilot stage, unable to slot into the operational reality of hospitals.

As a forward-looking contribution, the authors propose a conceptual federated, edge-based learning framework, explicitly presented as a design proposal rather than a validated system. The core idea is compelling: instead of shipping raw camera and radar data to central servers, processing would occur at the edge device itself, with only model updates, not identifiable physiological data, shared across institutions through federated learning. Such an architecture would simultaneously address privacy, reduce bandwidth, and allow hospitals to pool learning without pooling patient data. Combined with the review’s other prescriptions, standardized open datasets, unified evaluation metrics, and co-design with clinical end users, this roadmap suggests that contactless vital-sign monitoring may finally be approaching the inflection point where it leaves the demonstration stage and becomes an invisible, ambient layer of everyday clinical care.

Subject of Research: Non-contact medical IoT systems for monitoring physiological biomarkers

Article Title: A systematic review of non-contact medical IoT systems for physiological biomarker monitoring

Article References: Banik, B., Ghose, S., Nahiun, A. H., Lin, Y., Shalaginov, A., & Ghose, D. (2026). A systematic review of non-contact medical IoT systems for physiological biomarker monitoring. Frontiers in Medical Technology, 8, Article 1917332. https://doi.org/10.3389/fmedt.2026.1917332

Image Credits: AI Generated

DOI: 10.3389/fmedt.2026.1917332

Keywords: non-contact monitoring, medical IoT, radar sensing, vision-based sensing, respiration rate, heart rate, vital signs, federated learning, edge computing, patient monitoring, healthcare technology, PRISMA review

The clinical motivation is straightforward. Cough after pulmonary resection is not a trivial nuisance; persistent or intolerable coughing is associated with air leaks, sore throat, sleep disruption, and delayed recovery, and it is one of the more common complaints in the weeks following lobectomy and pneumonectomy. Yet routine assessment in hospital wards remains intermittent and subjective, relying on patient recall and nursing observation. Objective, continuous measurement of cough frequency and cough type could give clinicians a far richer signal about recovery trajectories, but until now there has been no practical tool to collect it. The new study set out to test whether such a tool is technically feasible inside a real hospital environment.

To build one, the researchers turned to Tongji Hospital in Wuhan, where they collected more than 7,000 raw audio recordings from 80 patients undergoing lung surgery between June 26 and August 13, 2024, under an approved ethics protocol with written informed consent from every participant. Raw ward audio is messy: it contains speech, footsteps, monitor alarms, and the ambient noise of a busy surgical floor. After segmentation of the recordings and a screening process based on expert consensus, the team assembled a cough-type cohort of 1,908 unique clips, of which 1,500 were labeled dry and 408 productive. That imbalance between the two cough classes is itself a realistic feature of the dataset, since productive coughs are less frequent but clinically important markers of airway secretions.

A critical methodological decision shaped the entire evaluation. Rather than splitting individual clips randomly into training and test sets, which would let near-duplicate coughs from the same patient inflate performance, the researchers partitioned patients themselves into training, validation, and test subsets of 56, 12, and 12 patients respectively. This patient-level split means the models were evaluated on coughs from people they had never heard before, a much stricter test that better approximates how the system would behave on new admissions. Each one-second recording was resampled to 16 kHz and converted into a 128-coefficient representation of mel-frequency cepstral coefficients, or MFCCs, a compact spectral fingerprint that captures the timbre of a cough while discarding irrelevant detail such as absolute loudness.

The architecture is deliberately split into two stages with different jobs. The first stage, a network the authors call K-DenseNet, answers a binary question: is this one-second snippet a cough or not? DenseNet-style architectures pass feature maps forward through dense connections, allowing each layer to reuse the outputs of all preceding layers, which strengthens feature propagation and keeps parameter counts modest. The second stage, dubbed RUNet, takes only the snippets that passed the first filter and classifies them as dry or productive. Its name echoes the U-Net family of convolutional networks originally developed for biomedical image segmentation, here adapted to treat the time-frequency structure of a cough clip as an image-like grid of features.

The headline numbers are striking. Under deterministic frozen-checkpoint inference, K-DenseNet achieved 99.16 percent accuracy and 99.16 percent cough recall, with a Matthews correlation coefficient of 0.983 across 239 held-out recordings. The MCC is a particularly demanding metric because it accounts for all four cells of the confusion matrix and penalizes models that succeed only on the majority class. RUNet, evaluated on 308 held-out cough clips, reached 99.03 percent accuracy, and for the harder task of spotting productive coughs it posted a precision of 96.61 percent, a recall of 98.28 percent, and an F1-score of 97.44 percent, with an MCC of 0.968 and an area under the precision-recall curve of 0.9967. In plain terms, the system almost never missed a cough and almost never confused a wet cough for a dry one on clean audio.

But the study is equally notable for what it admits. When the team deliberately degraded the audio with noise at a signal-to-noise ratio of 5 decibels, a level that mimics a genuinely noisy ward, detection accuracy fell to 71.97 percent and cough-type classification to 86.69 percent. The authors are explicit that this identifies the first-stage cough detector as the principal bottleneck for real-world deployment. It is a candid assessment that distinguishes the work from much of the cough-AI literature, where headline accuracies are reported on curated clips and robustness is left unexamined. A missed cough at the detection stage cannot be recovered downstream, so hardening that first stage against ambient noise becomes the clear engineering priority before the system can leave the laboratory.

The framework also extends beyond classification into structured reporting. By counting cough events and tallying dry versus productive episodes over time, the system can generate follow-up reports that quantify a patient’s recovery in a way that intermittent nursing notes cannot. Cough sound itself has established clinical meaning: prior work has shown that acoustic measures correlate with cough intensity and even predict extubation outcomes in ventilated patients, and the distinction between dry and productive cough informs decisions about airway clearance therapy and secretion management. A continuous, automated monitor could therefore turn a symptom that patients struggle to describe into a measurable physiological signal tracked hour by hour.

The researchers are careful about scope. They describe the results as internal feasibility, meaning the models were developed and tested on data from a single center, and they state plainly that external, multicenter, and prospective validation is required before any clinical use. The cough recordings cannot be made public because they contain clinical audio from human participants, though de-identified data may be shared on reasonable request subject to ethics approval, and the code is available from the corresponding author. Those caveats matter, because hospital acoustics, microphone placement, and patient demographics vary enormously across institutions, and a model tuned to one ward’s noise floor may falter in another.

Even with those limits, the study sketches a plausible near future in which a small microphone at the bedside, or perhaps a wearable device, runs a lightweight neural network around the clock and hands the surgical team an objective cough diary. The two-stage design is well suited to that vision: an efficient detector filters the torrent of ambient audio, and a specialist classifier examines only the small fraction of clips that matter, keeping computational cost low enough for edge hardware. The remaining challenge is noise, and the authors have framed it precisely. If subsequent work can push detection accuracy back up under realistic acoustic conditions, the humble cough, long the most under-measured vital sign on the surgical ward, may finally get the continuous monitoring it deserves.

Subject of Research: Audio-based deep learning detection and classification of postoperative cough after lung surgery

Article Title: Audio-based postoperative cough monitoring after lung surgery using a two-stage deep learning framework

Article References: Lin, W., Song, E., Hung, C.-C., Cai, C., Xu, P., Jin, T., Li, Y., Zhou, L., & Qian, Y. (2026). Audio-based postoperative cough monitoring after lung surgery using a two-stage deep learning framework. Biomedical Engineering Letters. https://doi.org/10.1007/s13534-026-00623-7

Image Credits: AI Generated

DOI: 10.1007/s13534-026-00623-7

Keywords: cough monitoring, lung surgery, deep learning, postoperative care, audio signal processing, MFCC, convolutional neural networks, biomedical engineering, clinical AI, cough classification, patient monitoring, Tongji Hospital

The motivation is rooted in a familiar limitation of critical care monitoring. Conventional ICU devices excel at point measurements—heart rate, blood pressure, oxygen saturation—but they attach to the patient and capture little of the spatial information written on the body’s surface. Patterns of peripheral perfusion, localized inflammation, or uneven temperature distribution can carry early signals of deterioration, yet standard multiparameter monitors are largely blind to them. Infrared thermography offers a way in: a radiometric thermal camera can map surface temperature remotely, without touching the patient, and previous studies have used thermal imaging to estimate respiratory rate and other vital signs. The catch, the authors note, is that most of that evidence comes from controlled laboratory prototypes, small cohorts, or algorithm-centered studies that never faced the realities of a hospital ward.

Those realities are formidable. Any system deployed in a modern hospital must pass through institutional information technology governance: network segmentation, cybersecurity policy, bandwidth limits, authentication requirements, and data protection law. Public cloud architectures, common in commercial Internet of Medical Things products, are often difficult or impossible to approve in European hospitals that must comply with the General Data Protection Regulation and maintain local control over patient-related data. The Madrid team therefore built their system entirely on-premise. Seven acquisition nodes—each a Raspberry Pi 4B single-board computer paired with a low-cost radiometric thermal camera of 256 by 192 pixel resolution—were installed at ICU beds, with the architecture designed to scale to all thirteen beds in the unit. Each node captured one thermal image per second, tagged it with timestamps and temperature statistics, and streamed it over the hospital’s existing Wi-Fi network to a local server for storage, indexing, and near real-time visualization. No parallel network was created, and no external cloud service touched the data.

Privacy was engineered into the hardware rather than bolted on afterward. Unlike clinical video systems, which typically require post-hoc de-identification steps such as blurring faces before footage can be released, the low-resolution radiometric thermal sensor cannot resolve the optical facial features needed to identify a person in the first place. Non-identifiability is a structural property of the acquisition modality itself, not a downstream processing step that could be omitted, misconfigured, or reversed. The database stored only bed identifiers and technical metadata, image acquisition was conditional on informed consent and activation by authorized personnel, and the study was approved by the relevant research ethics committee under approval number CEIm 23/63, conducted in accordance with the Declaration of Helsinki.

The first deployment phase delivered a humbling lesson in systems engineering. In its initial configuration, with synchronous image transmission and accumulating queues, the system buckled under sustained load. Analysis of 127,618 latency samples revealed a mean end-to-end delay—the time between image capture at the bedside and availability on the dashboard—of 26,344 milliseconds, with a median of 20,616 milliseconds and a 95th percentile reaching nearly 84 seconds. Images were piling up in transmission queues faster than they could be processed, turning a nominally real-time system into a delayed archive. The fix came not from replacing hardware but from rethinking the data flow: the team decoupled image reception from persistence, prioritized the most recent frame for visualization, improved queue management, and adjusted capture and transmission parameters, allowing older frames to be retransmitted asynchronously when bandwidth allowed.

The results of that architectural overhaul were dramatic. Across 6,917,464 subsequent latency samples, mean end-to-end latency fell to 126 milliseconds—a 99.52 percent reduction—with a median of 117 milliseconds and a 95th percentile of just 207 milliseconds. Crucially, the improvement extended to the upper tail of the distribution, confirming that prolonged queue accumulation had been eliminated rather than merely reduced. The system had been transformed, in the authors’ framing, from a delayed acquisition platform into a genuinely near real-time monitoring infrastructure, achieved purely through software and data-flow design on the same inexpensive sensing hardware.

With the optimized architecture in place, the team ran a structured ten-day evaluation with all seven nodes active. The system generated 5,763,312 thermal images, roughly 1.47 terabytes of data at an average of 275 kilobytes per image and about 147 gigabytes per day. The aggregate acquisition rate of 6.67 images per second came remarkably close to the nominal target of seven, indicating that the system held nearly full capture capacity under real operating conditions. Latency remained stable throughout the evaluation, showing no progressive deterioration as the stored dataset grew into the millions of images—a sign that the hybrid persistence strategy, with raw images in a hierarchical file system and metadata indexed in a PostgreSQL database, could sustain continuous long-term operation. Server resource use stayed almost trivially low, with mean CPU utilization of just 0.66 percent and mean RAM usage of 6.93 percent, leaving ample headroom for the planned expansion to thirteen beds.

The system was not flawless, and the failure modes it exhibited are themselves informative. Effective availability during the evaluation was 95.29 percent, with the shortfall traced almost entirely to transient wireless connectivity interruptions and local node issues rather than any failure of the central server or storage infrastructure. Telemetry recorded 44 Wi-Fi interface drop or reconnection events, cumulatively lasting 106,218 seconds, with individual incidents ranging from the ten-second minimum observable interval to a maximum of 40,807 seconds. Some nodes, labeled BOX7 and BOX8, showed larger temporal discontinuities, while BOX9 and BOX11 behaved most consistently. The team also contended with clock synchronization drift between nodes and server, addressed through network time synchronization and a dual timestamp strategy, and with power supply instability in some devices, mitigated by stabilizing the supply. The pattern is a familiar one in hospital deployments: the bottleneck is rarely computation or storage, but the shared, policy-constrained wireless environment that research systems must coexist with.

The authors are careful to frame this as foundational infrastructure rather than a clinical tool. The dashboard at this stage provides no alarms, no diagnostic interpretation, and no automated decision support, and the reported temperature metrics were computed over the entire thermal frame rather than a patient-isolated region, so they cannot be read as validated measurements of body temperature—bedding and nearby equipment could fall within the field of view. This deployment is explicitly the first of two phases in the approved research protocol, establishing technical feasibility, reliability, and governance compliance before a planned second phase evaluates clinical utility, including automated detection of patient falls, self-extubation, pressure ulcer development, and cardiac arrest, a role in differentiating septic, cardiogenic, and hypovolemic shock, and monitoring of postoperative recovery of normothermia. The team also acknowledges honest limitations: transport between nodes and server used HTTP without TLS encryption, a risk mitigated by network-layer isolation but flagged as a security-hardening priority; no formal penetration testing was performed; the study was single-center with one hardware configuration; and scalability to the full thirteen-bed configuration was projected from operational trends rather than directly tested. The system remains a pre-product research infrastructure, not a certified medical device.

Even with those caveats, the study offers a rare and valuable data point: a complete, quantified account of what it actually takes to run a distributed medical sensing network inside a live ICU. The practical lessons the authors distill—deploy progressively rather than all at once, coordinate closely with hospital IT teams from day one, build local buffering and asynchronous transmission into the architecture for resilience, and evaluate operational metrics like latency, availability, and storage growth rather than sensor accuracy alone—read as a reference model for anyone attempting similar deployments. As hospitals edge toward AI-assisted, contactless monitoring, the hardest problems may not be in the algorithms at all, but in the unglamorous plumbing of queues, clocks, tokens, and Wi-Fi. In Madrid, at least, that plumbing now works, one thermal frame at a time.

Subject of Research: On-premise thermal Internet of Medical Things monitoring in an intensive care unit

Article Title: Real-World Deployment and Operational Evaluation of an On-Premise Thermal IoMT System in an Intensive Care Unit

Article References: Avilés, E., Lafuente, J.-L., González, S., & Beunza, J.-J. (2026). Real-World Deployment and Operational Evaluation of an On-Premise Thermal IoMT System in an Intensive Care Unit. Journal of Medical Systems, 50(1), Article 144. https://doi.org/10.1007/s10916-026-02455-5

Image Credits: AI Generated

DOI: 10.1007/s10916-026-02455-5

Keywords: Internet of Medical Things, thermal imaging, intensive care unit, patient monitoring, on-premise architecture, hospital IT governance, Raspberry Pi, infrared thermography, latency optimization, data privacy, edge computing, critical care