Governing AI-Assisted Diabetes-Related Foot Ulcer Monitoring in Viet Nam: Wound Image Review, Limb-Threat Assessment and Healing Follow-Up

Governing AI-Assisted Diabetes-Related Foot Ulcer Monitoring in Viet Nam: Wound Image Review, Limb-Threat Assessment and Healing Follow-Up

A patient with type 2 diabetes has been receiving treatment for a foot ulcer for several weeks. Between clinic appointments, the patient uses a smartphone to photograph the wound and upload the images to a hospital monitoring platform.

An artificial intelligence (AI) system compares the latest photograph with previous images. It estimates that the visible wound area has decreased and reports a favorable healing trend.

At the next consultation, however, the clinician identifies a more serious problem. Blood supply to the affected foot is inadequate, and new tissue changes raise concern about the viability of part of the foot.

The wound appears smaller, but the patient's risk of a serious complication has not necessarily decreased.

This situation illustrates an important limitation of image-based wound monitoring. A photograph can provide useful information about visible changes, but it cannot independently establish whether circulation is adequate, exclude infection involving deeper structures, or determine whether the limb is threatened.

For healthcare organizations in Viet Nam, the question is not simply whether AI can measure a wound accurately. It is whether image analysis contributes information that helps clinicians recognize complications, deliver appropriate treatment, and arrange timely follow-up.

The objective is not to produce more photographs, automated wound scores, or reassuring progress reports. It is to help patients achieve healing, avoid preventable complications, and preserve mobility and limb function.

Why Diabetes-Related Foot Ulcers Require More Than Wound Observation

Diabetes-related foot ulcers are serious complications associated with peripheral neuropathy, peripheral artery disease, altered foot biomechanics, and other clinical factors.

Loss of protective sensation can allow repeated pressure or minor trauma to damage tissue without producing the pain that would ordinarily alert someone to an injury. Foot deformities may concentrate pressure over particular areas, while impaired circulation can reduce tissue oxygenation and delay healing.

Infection may develop and extend into deeper structures, including bone. These problems frequently coexist, making the patient's condition more complicated than the appearance of the ulcer alone might suggest.

A person may have a relatively small ulcer but substantial peripheral artery disease. Another may have a larger superficial wound with adequate circulation and no clinical infection. Although their wounds look different, size alone does not establish which patient faces the greater risk of serious complications.

Persistent ulcers can lead to infection, hospitalization, reduced mobility, impaired quality of life, and lower-extremity amputation. Even after an ulcer heals, recurrence remains a concern because neuropathy, abnormal mechanical pressure, foot deformity, or vascular disease may persist.

Diabetes-related foot care is therefore a continuing clinical responsibility rather than a single episode of wound treatment.

Digital photography can help clinicians document wound appearance, compare visible changes over time, and communicate information between appointments. AI may improve some of these functions by identifying wound boundaries or measuring visible changes more consistently.

However, several important determinants of healing cannot be established through photography alone. A meaningful monitoring program must connect image findings with assessment of the patient, the affected limb, the treatment being delivered, and the need for further clinical review.

Viet Nam's National Guidance Provides the Clinical Foundation

Ministry of Health Decision No. 1530/QD-BYT, dated 24 March 2023, issued the professional guidance Diagnosis and Treatment of Diabetes-Related Foot Ulcers (Hướng dẫn chẩn đoán và điều trị loét bàn chân do đái tháo đường).

The guidance applies to healthcare examination and treatment facilities throughout Viet Nam and provides a national clinical foundation for evaluating and managing diabetes-related foot ulcers.

It addresses wound assessment, peripheral neuropathy, vascular examination, infection, foot deformities, diagnosis, treatment, and prevention.

The guidance recognizes the interactions between neuropathy, impaired circulation, abnormal mechanical pressure, and infection. It also addresses debridement, wound care, pressure redistribution, treatment of vascular problems, and management of the underlying disease.

These elements are relevant to AI-assisted monitoring because successful care depends on more than observing changes in wound appearance.

Before introducing an image-analysis system, healthcare organizations should determine whether their existing clinical processes adequately assess neurological function, peripheral circulation, wound characteristics, infection, and treatment needs.

International recommendations provide additional clinical context.

The International Working Group on the Diabetic Foot (IWGDF) published its latest comprehensive guideline update in 2023. As of November 2026, this remains the current comprehensive version, with the next update planned for 2027.

The guidance covers prevention, ulcer classification, infection, peripheral artery disease, pressure offloading, wound healing interventions, and other aspects of diabetes-related foot disease.

The joint IWGDF and Infectious Diseases Society of America (IDSA) guideline addresses the diagnosis and treatment of diabetes-related foot infections.

The intersocietal guideline developed by IWGDF, the European Society for Vascular Surgery (ESVS), and the Society for Vascular Surgery (SVS) provides recommendations concerning peripheral artery disease in people with diabetes and foot ulcers.

The American Diabetes Association (ADA) Standards of Care in Diabetes—2026 also emphasizes comprehensive foot examination, recognition of neuropathy and peripheral artery disease, appropriate referral, and interprofessional management.

Together, these resources provide a clinical foundation against which the intended contribution of AI should be evaluated.

International guidance may inform Vietnamese hospital protocols, but it does not automatically become binding Vietnamese law.

What AI-Assisted Wound Image Review Can Actually Do

AI-assisted wound assessment includes several different technological functions.

Some systems identify whether a photograph contains an ulcer. Others locate the lesion, segment its visible boundary, estimate surface area, or classify selected image features.

More advanced applications may compare photographs obtained at different times and calculate changes in visible wound dimensions. Some models attempt to predict whether an ulcer is likely to heal within a defined period using images, structured clinical information, or both.

Thermal imaging and other sensor-based approaches are also being investigated for detecting temperature patterns and physiological changes associated with foot complications.

These functions should be distinguished.

Wound detection establishes whether an image appears to contain a relevant lesion. Segmentation identifies the visible boundary of that lesion. Measurement estimates a particular dimension or area, while longitudinal monitoring compares findings over time.

A healing prediction model attempts to estimate a future outcome rather than simply describe the current wound.

Each function requires evidence appropriate to its intended use.

A system that accurately segments wound boundaries may not reliably characterize tissue viability. An algorithm that measures visible surface area may provide no information about wound depth, infection involving underlying structures, or impaired circulation.

Likewise, a model that predicts healing probability does not necessarily explain why a particular wound is failing to improve.

Healthcare organizations should therefore examine the specific claims made for each product.

A system evaluated for wound photography and measurement should not be assumed capable of independently diagnosing peripheral artery disease, deep infection, or osteomyelitis.

An apparent improvement in predicted healing should also not become a reason to continue treatment unchanged when concerning clinical findings emerge.

The intended clinical function must be defined before implementation, including what information the system produces, what evidence supports its performance, and which decisions require professional assessment.

Image Quality and Measurement Reliability Matter

The quality of wound photographs can substantially influence AI performance.

An ulcer photographed under bright clinical lighting may appear different when photographed at home under warm or uneven light. Changes in camera angle, distance, focus, shadows, and background can affect the apparent wound boundary. Different cameras may also reproduce skin and tissue colors differently.

Without appropriate controls, the system may interpret photographic variation as a change in the wound itself.

An image taken from a slightly different angle may make an irregular ulcer appear smaller. A photograph captured after dressing removal or wound cleaning may show tissue characteristics that differ from those visible before the procedure.

Exudate, debris, surrounding callus, and partially obscured wound edges can further affect measurement.

These differences become particularly important when an algorithm compares serial images and produces a healing trend.

Reliable monitoring requires standardized image capture procedures appropriate to the intended clinical use.

Healthcare organizations should define how photographs are obtained, which anatomical views are required, whether a suitable measurement reference is used, and how image quality is assessed.

When the system claims to measure wound area, it must demonstrate appropriate calibration and account for relevant imaging geometry.

A ruler appearing in a photograph does not automatically establish that the calculated area is accurate, particularly when the ulcer is located on a curved or angled part of the foot.

Image quality requirements should also consider differences in skin tone and tissue appearance. A model developed using a narrow range of patient images may perform differently when introduced into another population.

Blurred, poorly illuminated, incomplete, or otherwise unsuitable photographs should not be converted into apparently definitive measurements.

A well-designed system should identify inadequate image quality and direct the user toward repeat photography or an alternative clinical assessment.

Clinicians also need to understand what a measurement actually represents. Estimated visible surface area is not interchangeable with wound depth, volume, tissue viability, or overall severity.

Changes following debridement provide an example.

Removing callus or nonviable tissue may make the visible wound appear larger immediately afterward, even when the procedure was clinically appropriate.

An algorithm that interprets every increase in measured area as deterioration may therefore generate misleading conclusions.

Wound measurements should be documented and interpreted alongside relevant treatment events and clinical findings.

A Smaller Wound Does Not Necessarily Mean a Safer Limb

One of the most important limitations of image-based monitoring is that surface appearance does not fully represent limb-threatening disease.

An ulcer may become smaller while blood supply remains inadequate. A wound may appear relatively clean despite infection involving deeper tissues. A patient with severe neuropathy may experience little pain even when significant tissue damage is developing.

These situations require assessment beyond photography.

The IWGDF guidelines emphasize evaluating peripheral artery disease in people with diabetes-related foot ulcers.

Clinical assessment may include relevant symptoms, pedal pulses, capillary refill, skin findings, and objective vascular investigations.

Doppler waveform assessment, ankle-brachial index, toe-brachial index, and toe pressure measurements can provide useful information.

However, ankle-brachial index measurements may be misleading in people with diabetes because arterial calcification can make vessels poorly compressible.

An apparently normal or elevated ankle-brachial index should therefore not automatically exclude clinically important peripheral artery disease.

Toe pressure measurements and other appropriate vascular investigations may be needed.

The intersocietal IWGDF–ESVS–SVS guideline recommends considering the Wound, Ischemia and foot Infection (WIfI) classification system in patients with diabetes, peripheral artery disease, and foot ulcers or gangrene to help estimate healing likelihood and amputation risk.

The classification incorporates tissue loss, the severity of ischemia, and infection.

An image-only AI model cannot establish a complete WIfI classification because important components require clinical information that a conventional photograph cannot provide.

Even when a system combines images with structured clinical data, the accuracy and intended use of the resulting classification require validation.

For healthcare organizations, the practical safeguard is to ensure that vascular assessment and evaluation of limb threat remain part of the clinical pathway.

AI-generated measurements may contribute useful information, but they should not independently determine whether circulation is adequate or specialist referral is necessary.

Infection Must Be Assessed Clinically

Diabetes-related foot infection presents another challenge for image-based AI.

A photograph may show redness, swelling, discoloration, drainage, or other visible abnormalities. These findings may raise concern, but their appearance alone does not establish infection.

The IWGDF/IDSA 2023 guideline recommends diagnosing soft-tissue diabetes-related foot infection clinically, based on relevant local or systemic signs and symptoms of inflammation.

It also recommends assessing infection severity using the IWGDF/IDSA classification system.

This matters because noninfectious inflammation may resemble infection, while some patients with diabetes and neuropathy may have less obvious inflammatory responses despite clinically significant disease.

An image classifier relying heavily on visible redness may incorrectly identify an uninfected wound as infected. Conversely, an infected wound may not display the features that the system has learned to recognize.

Temperature-based monitoring requires similar caution.

Although foot temperature monitoring has potential applications in preventing recurrent ulceration, the IWGDF/IDSA infection guideline suggests not using foot temperature measurements to diagnose diabetes-related soft-tissue infection.

Monitoring temperature changes to identify possible inflammation is therefore a different clinical function from establishing an infection diagnosis.

Suspected infection involving deeper structures may require additional assessment.

When osteomyelitis is suspected, IWGDF/IDSA recommends considering the probe-to-bone test, plain radiographs, and relevant inflammatory markers as initial investigations.

Magnetic resonance imaging may be appropriate when the diagnosis remains uncertain after clinical, radiographic, and laboratory assessment.

These investigations must be selected according to the patient's condition and the clinical questions requiring resolution.

An AI model detecting an abnormal wound surface pattern cannot independently confirm or exclude osteomyelitis.

Antimicrobial treatment also requires appropriate indications.

IWGDF/IDSA advises against using systemic or local antibiotics for clinically uninfected foot ulcers solely to prevent infection or promote healing.

This recommendation is relevant to AI governance because a false-positive infection classification could encourage unnecessary antibiotic treatment. Conversely, a false-negative result must not delay treatment of clinically significant infection.

Both risks need to be considered when determining how image analysis contributes to clinical decisions.

Recognizing When Urgent Escalation Is Necessary

Diabetes-related foot ulcer monitoring must distinguish routine healing assessment from clinical conditions requiring urgent attention.

Severe infection, rapidly progressive tissue damage, gangrene, or significant limb ischemia may require immediate or accelerated clinical evaluation.

Other concerning findings may include spreading infection, suspected deep abscess, necrotizing infection, substantial discoloration, or systemic illness.

The urgency depends on the clinical presentation and the patient's overall condition.

The IWGDF/IDSA guideline recommends urgent surgical consultation for severe diabetes-related foot infection and selected moderate infections complicated by extensive gangrene, necrotizing infection, suspected deep abscess, compartment syndrome, or severe lower-limb ischemia.

Where peripheral artery disease coexists with infection and ulceration or gangrene, urgent consultation involving surgical and vascular specialists may be necessary to determine whether drainage or revascularization is indicated and when those procedures should occur.

These decisions should not depend on an AI classifier reaching a particular probability threshold.

If concerning clinical findings are present, appropriate assessment and escalation should proceed regardless of the automated result.

This principle is equally important during remote monitoring.

Patients submitting wound photographs between appointments need clear information about changes requiring urgent medical attention.

A platform that accepts photographs but provides no dependable route to clinical assessment may create a false sense of security.

Healthcare organizations should define which reports require immediate escalation, which warrant earlier clinic review, and which can reasonably be addressed through scheduled follow-up.

The objective is to recognize significant deterioration without treating every minor change in wound appearance as an emergency.

Offloading Is Central to Healing

Continued mechanical pressure is an important reason why diabetes-related foot ulcers may fail to heal.

Patients with peripheral neuropathy may repeatedly load the affected area without experiencing enough pain to limit their activity.

A wound-monitoring platform may document slow healing without identifying the underlying cause.

The 2023 IWGDF offloading guideline recommends a non-removable knee-high offloading device as the first-choice intervention for a neuropathic plantar forefoot or midfoot ulcer when clinically appropriate.

This is a strong recommendation supported by moderate-certainty evidence.

However, it does not apply indiscriminately to every ulcer.

Ulcer location, infection severity, ischemia, contraindications, patient tolerance, mobility, and available resources influence the choice of intervention.

For selected patients, removable knee-high or ankle-high devices may be more suitable, particularly when a non-removable device is contraindicated or not tolerated.

The guideline also addresses how offloading should be adapted when infection or ischemia is present.

The intervention must be practical for the patient.

A removable device may be appropriate in principle but provide limited benefit if the patient does not wear it during weight-bearing activities.

Patients may encounter difficulties related to work, transportation, walking, balance, or the ability to manage the equipment independently. These barriers require discussion and appropriate support.

Offloading devices may also affect mobility and balance in some patients, making individualized assessment important.

Viet Nam's Ministry of Health guidance recognizes the role of pressure redistribution in diabetes-related foot ulcer treatment.

For healthcare organizations, the important question is whether the selected intervention is appropriate, available, tolerated, and actually used.

An AI system that repeatedly reports poor healing while the patient continues to place excessive pressure on the wound may add little clinical value unless the monitoring process prompts reassessment of the treatment plan.

The purpose of measuring progress is to support better care, not simply generate another numerical result.

Healing Trends Need Clinical Interpretation

Treatment of diabetes-related foot ulcers often extends over weeks or months.

Serial wound measurements may help clinicians understand whether visible changes are progressing in a favorable direction.

An ulcer that becomes progressively smaller may be moving toward healing. However, reduction in visible area does not establish adequate circulation, infection control, or tissue viability.

Similarly, a wound that remains unchanged or becomes larger may require reassessment of several contributing factors, including mechanical pressure, peripheral perfusion, infection, wound care, glucose management, and the patient's ability to follow the treatment plan.

The intersocietal IWGDF–ESVS–SVS peripheral artery disease guideline provides a useful example of how wound progression can guide reassessment.

It recommends reassessing vascular status and consulting a vascular specialist when a diabetes-related foot ulcer deteriorates or fails to improve significantly despite appropriate management.

A reduction in wound area of less than 50% within four weeks is provided as an example of insufficient improvement.

This is a Best Practice Statement based on indirect evidence and expert opinion.

It is not an automatic diagnosis of ischemia, nor is it an instruction to perform revascularization solely because a numerical threshold has been reached.

The recommendation also does not mean clinicians should wait four weeks when severe infection or limb-threatening ischemia is already suspected. Clinical findings may require much earlier action.

For AI-assisted monitoring, the recommendation illustrates how longitudinal wound measurements might contribute to a structured reassessment process.

A system could identify that the measured wound area has not improved as expected. The responsible clinician would then verify the measurements, assess the patient, and determine whether further investigations or changes in treatment are necessary.

The system should not independently conclude that vascular intervention, antibiotics, or an advanced wound treatment is required.

Monitoring is most valuable when it helps the care team recognize that the existing plan needs professional reconsideration.

What the Latest Evidence Shows About AI Performance

Research on AI-assisted diabetes-related foot ulcer assessment has expanded, but the available evidence remains uneven.

A scoping review by Wartakusumah and colleagues, published in Diabetes Research and Clinical Practice in September 2026, examined clinical validation studies of AI systems used for diabetes-related foot ulcer detection and assessment.

The review included 11 studies published between 2020 and 2026 from eight countries.

Reported sensitivity ranged from 91% to 100%, while specificity varied considerably, from 20% to 96.8%.

For wound measurement reliability, reported intraclass correlation coefficients ranged from 0.825 to 0.998.

Some included comparisons found that AI-assisted methods reduced manual overestimation of wound area by approximately 13.4%–25.2%.

These findings suggest that selected AI systems may improve the consistency of visible wound detection and measurement.

However, the estimates came from different systems, clinical tasks, datasets, and study designs.

They should not be interpreted as pooled measures of overall AI accuracy or as evidence that every AI wound assessment system performs within those ranges.

High sensitivity for identifying visible ulcers does not establish equivalent accuracy for recognizing deep infection, ischemia, or osteomyelitis.

The review also identified barriers affecting real-world implementation, including image quality, algorithmic misclassification, limitations in digital literacy, unreliable infrastructure, and data privacy concerns.

The authors emphasized that current image-based approaches remain limited in their ability to assess deeper wound features adequately.

They called for larger prospective multicenter studies with standardized evaluation methods before broad routine clinical adoption can be recommended.

These findings support further investigation, but they do not establish that AI-assisted wound monitoring improves patient outcomes in routine practice.

Recent Reviews Highlight the Gap Between Model Development and Clinical Use

Another scoping review by Zulfiqar and colleagues, published in JMIR Diabetes on 8 July 2026, examined AI-based models for diabetic foot ulcer assessment.

The researchers identified 654 records, of which 46 studies met the inclusion criteria.

The included studies predominantly addressed image segmentation, diagnostic classification, risk prediction, or monitoring.

Convolutional neural networks were commonly used, although the review included several other machine-learning approaches.

Many studies reported promising technical performance using measures such as accuracy, Dice similarity coefficient, and area under the receiver operating characteristic curve.

However, most relied on retrospective datasets or single-center study designs. External validation was limited, and relatively few studies evaluated implementation in routine clinical practice.

A more recent systematic review by Aldhubiea and colleagues, published in Cluster Computing in October 2026, examined deep learning methods for diabetic foot ulcer classification, segmentation, and multimodal integration.

The review examined 48 eligible studies or study configurations and assessed methodological rigor, reproducibility, and practical deployment considerations.

Only 13 of the 48 studies, approximately 27.1%, reported external validation. Four reported publicly available code or model artifacts, while eight described mobile or web deployment constraints.

These findings are important because a model may perform well on a carefully prepared dataset but less reliably when introduced into a different hospital, used with another smartphone, or applied to wounds that differ from those represented in its development data.

Multimodal systems combining ordinary photographs with thermal or other imaging may offer additional information, but they introduce requirements for calibration, alignment, and consistent acquisition.

A more complex model does not remove the need for reliable data and a clearly defined clinical purpose.

Taken together, the 2026 reviews indicate that AI-assisted wound assessment is technically promising.

They do not establish that routine deployment reduces lower-extremity amputations, improves healing rates, or decreases mortality.

Those clinical benefits require separate evaluation.

For Vietnamese hospitals, strong technical performance should be considered one part of the evidence, not proof that an AI product is ready for widespread clinical use.

Connecting Remote Images With Actual Clinical Care

For healthcare organizations, the practical challenge is determining how AI-generated information fits into an established wound care service.

Consider a patient who attends a diabetic foot clinic every two weeks and submits photographs between appointments.

The hospital needs to establish who reviews those images, when the monitoring service operates, and what happens when information suggests deterioration.

A nurse or wound care professional may initially review submissions under an approved protocol. When an image or accompanying symptom report raises concern, the case may require assessment by the responsible clinician.

The clinician can then determine whether the patient needs an earlier appointment, additional investigation, or specialist referral.

The pathway must also recognize the limitations of remote observation.

A patient may report increasing swelling, new discoloration, fever, or another concerning change even when the photograph appears relatively unchanged.

Such findings may require direct assessment regardless of the AI result.

Patients need clear information about what the digital monitoring service can and cannot provide.

A platform accepting photographs at any time is not necessarily a continuously monitored clinical service.

The organization should explain when submissions are reviewed, how urgent concerns are handled, and what patients should do when they need medical attention outside those hours.

An automated acknowledgment does not mean that a clinician has assessed the wound.

The organization also needs an approach to missing submissions.

A patient who stops uploading photographs may be experiencing technical difficulties, reduced mobility, worsening illness, or problems using the application.

The absence of a new photograph should not be interpreted as evidence that the wound is stable.

The response to missed reports should reflect the patient's clinical risk and the agreed monitoring plan.

Remote monitoring is most useful when it supports continuity of care without creating inappropriate reassurance.

Multidisciplinary Responsibility and Referral Follow-Up

Diabetes-related foot ulcer management often requires coordinated care from several professional groups.

Clinicians involved in diabetes management may oversee the underlying condition and coordinate broader treatment.

Nurses and wound care professionals contribute clinical observation, dressing care, patient education, and follow-up.

Vascular specialists assess significant peripheral artery disease and determine whether further vascular investigation or revascularization may be appropriate.

Surgical teams may become involved when deep infection, necrotic tissue, or other complications require operative management.

Infectious diseases specialists, rehabilitation professionals, orthotic services, microbiology teams, and other healthcare professionals may contribute according to the patient's needs.

The organization should establish clear responsibility for each part of the clinical pathway.

An AI-generated improvement in wound area should not automatically close a case when the patient still requires vascular assessment.

Similarly, issuing an electronic referral does not establish that the patient received the intended specialist consultation.

Where clinically important referrals are made, the care pathway should include reasonable arrangements for determining whether assessment occurred and whether further action is required.

These responsibilities become particularly important when patients receive care at multiple facilities or return home between specialist appointments.

A reliable system should preserve relevant information, communicate outstanding concerns, and support coordination among the professionals responsible for care.

Technology may help organize this process, but it does not replace professional accountability.

Measuring Healing and Limb Preservation

The success of AI-assisted wound monitoring should not be measured only by segmentation accuracy or the number of photographs uploaded.

A meaningful evaluation should examine whether the overall care pathway becomes more reliable and whether clinically important outcomes improve.

Image quality and measurement reliability are useful starting points.

Hospitals may assess whether photographs meet defined quality standards, whether measurements are reproducible, and how frequently clinicians need to correct AI-generated findings.

The next concern is clinical assessment and decision-making.

The organization should determine whether patients receive appropriate evaluation of infection, peripheral circulation, wound characteristics, and other factors affecting healing.

Where specialist referral is indicated, timeliness and completion of the assessment should also be examined.

Treatment implementation matters.

A prescription for an offloading device does not mean that the patient obtained it or can use it appropriately. Likewise, a wound care plan recorded in the medical record does not establish that every intended intervention was delivered.

Clinical outcomes may include complete wound healing, time to healing, recurrence, infection-related hospitalization, surgical intervention, and lower-extremity amputation.

Healing should be defined consistently.

Complete epithelialization, rather than simply a sufficiently small AI-estimated wound area, is an important component of confirming wound closure. Where appropriate, the durability of closure should also be evaluated.

Amputation outcomes should distinguish minor and major amputations because their clinical and functional implications differ.

The patient's mobility, functional ability, quality of life, pain, and ability to continue daily activities may also be relevant.

These outcomes require careful interpretation.

Patients with extensive tissue loss, severe infection, or significant peripheral artery disease may have worse outcomes despite appropriate care.

Comparing healing or amputation rates between hospitals without accounting for differences in patient risk and clinical severity can therefore be misleading.

The evaluation should also include unintended consequences.

Excessive alerts, unnecessary referrals, duplicate investigations, inappropriate treatment changes, additional staff workload, and patient costs may offset potential benefits.

Improved documentation is useful, but the ultimate objective is safer and more effective clinical care.

Monitoring Must Continue After the Ulcer Heals

The disappearance of an open ulcer does not mean that the underlying risk has disappeared.

Peripheral neuropathy, foot deformities, abnormal pressure, and peripheral artery disease may persist after wound closure.

Patients with a history of diabetes-related foot ulceration remain at increased risk of recurrence.

The ADA Standards of Care in Diabetes—2026 recommends ongoing preventive foot assessment, including examination at every visit for people with sensory loss or a history of ulceration or amputation.

Individuals with previous ulcers and other high-risk conditions may need continuing specialist foot care and surveillance.

Footwear, skin inspection, management of callus, pressure redistribution, and patient education may remain important.

The frequency and nature of follow-up should reflect individual risk and clinical circumstances.

For digital monitoring programs, this creates a distinction between observing an active wound and monitoring for recurrence after healing.

A system designed to segment open ulcers may no longer be suitable once the skin has healed.

Post-healing surveillance may require different information, including new skin lesions, changes in pressure points, footwear problems, or symptoms suggesting recurrent disease.

Patients and caregivers need to understand how to recognize new concerns and when to seek assessment.

Digital tools may support some of these activities, but their effectiveness must be established for the intended function.

An AI-generated healed-wound status should not automatically terminate appropriate long-term preventive care.

Practical Challenges in Vietnamese Healthcare Settings

Healthcare organizations in Viet Nam differ in their access to specialized diabetic foot care, vascular investigations, wound care professionals, and digital infrastructure.

A tertiary hospital may have established multidisciplinary services and access to vascular interventions.

Another facility may provide routine diabetes and wound care while referring patients to a higher-level center for specialized assessment or treatment.

These differences influence what an AI-assisted monitoring system can realistically achieve.

A digital application may identify a wound that is failing to improve, but it cannot ensure that vascular assessment or revascularization is readily available.

Likewise, an automated recommendation concerning pressure offloading may offer limited benefit if appropriate equipment is unavailable or the patient cannot use it.

Clinical and operational capacity must therefore be considered alongside technology.

Patient circumstances also matter.

Some people may have difficulty photographing the soles of their feet because of limited mobility, impaired vision, obesity, or difficulty reaching the affected area. Others may not have a suitable smartphone or reliable internet access.

Patients with peripheral neuropathy may feel little pain even when serious complications develop, making education and accessible follow-up especially important.

A monitoring program should not assume that every patient can obtain reliable photographs independently.

Caregiver assistance may be helpful, but it introduces additional considerations involving training, privacy, and communication.

The organization should also examine whether digital monitoring creates additional travel, equipment, internet, or service costs for patients.

An accessible alternative should remain available when someone cannot use the technology.

The purpose of remote monitoring is to improve continuity and access, not create another obstacle to appropriate care.

Legal and Professional Responsibilities in Viet Nam

AI-assisted diabetes-related foot ulcer monitoring must be considered within Viet Nam's applicable healthcare, artificial intelligence, medical device, and personal data protection frameworks.

The Law on Medical Examination and Treatment No. 15/2023/QH15 provides the broader legal foundation for professional healthcare practice, patient rights, medical records, and the responsibilities of healthcare organizations.

Ministry of Health Decision No. 1530/QD-BYT, dated 24 March 2023, provides directly relevant national professional guidance for diagnosing and treating diabetes-related foot ulcers.

Healthcare organizations should ensure that locally approved clinical procedures reflect applicable Vietnamese requirements and the clinical circumstances of patients receiving care.

The introduction of AI does not transfer professional responsibility for diagnosing infection, assessing circulation, evaluating wound severity, or selecting treatment from appropriately qualified healthcare professionals to a software system.

Viet Nam's Law on Artificial Intelligence No. 134/2025/QH15 took effect on 1 March 2026.

Decree No. 142/2026/ND-CP, effective from 1 May 2026, provides detailed implementing provisions, including requirements concerning risk classification.

Decision No. 33/2026/QD-TTg, effective from 15 August 2026, establishes the current list of high-risk AI systems.

AI-assisted diabetes-related foot ulcer monitoring is not separately identified by name among the healthcare entries concerning specified AI-assisted surgical and robotic surgical systems.

However, the absence of a specifically named entry does not automatically establish that every such system is low-risk or exempt from applicable AI obligations.

A product-specific regulatory assessment must consider its intended purpose, actual functions, degree of automation, deployment arrangements, and relevant legal criteria.

A system that stores photographs and calculates measurements for professional review may raise different regulatory considerations from software marketed as independently diagnosing infection or recommending treatment.

Formal regulatory classification and clinical risk are also distinct matters.

A system that incorrectly reassures clinicians about a deteriorating wound or contributes to delayed assessment of limb-threatening disease may create significant patient-safety consequences regardless of its formal classification.

Healthcare organizations should also determine whether a particular product falls within the applicable medical device framework.

Decree No. 98/2021/ND-CP on medical device management, as amended, and Ministry of Health Circular No. 24/2026/TT-BYT may apply depending on the intended medical purpose and characteristics of the product.

The presence of artificial intelligence alone does not determine medical device status.

Conversely, describing software as a documentation platform or administrative dashboard does not automatically exclude it from relevant medical device requirements.

The product's intended medical function and actual capabilities require assessment before clinical deployment.

Protecting Wound Images and Patient Information

Wound photographs may contain identifiable personal information even when the patient's face is not visible.

Images may include distinctive anatomical features, scars, skin markings, or other characteristics that contribute to identification when combined with clinical records.

Digital photographs may also contain metadata relating to when, where, or how the image was captured.

When photographs are linked to a patient's medical record, they should be managed within an appropriately protected clinical information process.

The Law on Personal Data Protection No. 91/2025/QH15 and Decree No. 356/2025/ND-CP, both effective from 1 January 2026, establish relevant requirements for personal data processing and protection.

Decree No. 102/2025/ND-CP on health data management, effective from 1 July 2025, is also relevant to the governance, processing, protection, and use of health data.

Healthcare organizations should establish the applicable legal grounds and purposes for processing patient information and implement safeguards appropriate to the actual activities involved.

These may include access restrictions, information security controls, data retention arrangements, processing documentation, incident response procedures, and other applicable compliance requirements.

The organization should determine whether wound photographs are processed locally, transmitted to external services, or stored by technology providers.

Different arrangements may create different responsibilities concerning confidentiality, data access, security, and incident management.

Images collected for clinical care should not automatically be reused for unrelated model development, commercial activities, or research without assessing the applicable legal grounds and required safeguards.

Where external technology suppliers are involved, contractual arrangements should address data processing, storage, permitted uses, access controls, security, and any applicable cross-border processing requirements.

Hospitals should also avoid uncontrolled use of consumer AI services for identifiable wound images without appropriate organizational authorization and protection.

The aim is to allow patients to benefit from digital monitoring without unnecessary exposure of sensitive clinical information.

International guidelines issued by IWGDF, IDSA, ADA, and other professional organizations may inform local practice, but they do not automatically become binding Vietnamese legal requirements.

Similarly, organizational safeguards recommended for AI implementation should be distinguished from specific statutory obligations unless an applicable legal provision makes them mandatory.

A Practical Starting Point for Vietnamese Healthcare Organizations

A hospital or diabetic foot service considering AI-assisted wound monitoring should begin by examining how its existing clinical pathway functions.

A focused review of recent patients with diabetes-related foot ulcers can identify how wounds are documented, how infection and vascular status are assessed, and whether patients receive appropriate treatment and follow-up.

The review should also examine what happens when healing does not progress.

Are patients reassessed promptly? Are concerns about infection or circulation investigated? Is offloading reviewed? Are referrals completed when necessary?

These questions may reveal opportunities for improvement that do not require AI.

If wound measurement is inconsistent, standardized photography and documentation may be useful initial interventions.

If the main problem is delayed vascular referral, strengthening communication and referral arrangements may offer greater clinical benefit than introducing an image-analysis model.

If patients have difficulty attending frequent appointments, a carefully designed remote monitoring service may support continuity, provided that appropriate clinical review and escalation remain available.

Once the actual problem has been defined, the organization can determine whether AI is likely to add value.

One initial use case might involve comparing AI-assisted wound measurements with measurements performed by trained healthcare professionals.

Evaluation should examine accuracy, reproducibility, image quality requirements, and the frequency and clinical significance of errors.

Testing should reflect realistic conditions rather than relying only on carefully selected development images.

Patients with different skin tones, wound locations, tissue characteristics, and photographic environments should be represented where relevant to the intended use.

During initial evaluation, AI outputs should not independently determine treatment or replace established clinical assessment.

A supervised or silent evaluation may help the organization identify errors and assess workflow feasibility while existing care continues.

If performance is acceptable and appropriate safeguards are established, a limited clinical pilot may be considered.

The pilot should examine whether AI contributes useful information, whether important deterioration is recognized, and whether patients receive appropriate follow-up.

It should also identify additional workload, unnecessary alerts, patient difficulties, and other unintended consequences.

Clinical leaders should review missed cases, misleading measurements, and situations in which image analysis conflicts with bedside findings.

Criteria for modifying, restricting, or discontinuing the system should be established before wider deployment.

Expansion to additional patient groups or facilities should depend on evidence relevant to those settings.

A system validated for visible wound measurement should not automatically be extended to infection diagnosis, ischemia assessment, or treatment recommendation without additional evidence and appropriate regulatory review.

The objective is to strengthen an established diabetic foot care service rather than build the clinical process around the technology.

From Wound Images to Safer Limb Care

AI-assisted diabetes-related foot ulcer monitoring offers a potentially useful way to improve wound documentation, support longitudinal assessment, and identify patients whose healing progress requires attention.

Research published in 2026 suggests that selected AI systems can perform promising wound detection, segmentation, and measurement tasks.

However, accurate image analysis does not establish that a system can assess blood flow, diagnose deep infection, or prevent lower-extremity amputation.

Those questions require broader assessment of the patient and the affected limb.

For healthcare organizations in Viet Nam, responsible implementation should be grounded in the Ministry of Health's clinical guidance and informed by relevant international recommendations.

Wound images need to be interpreted alongside neurological findings, vascular assessment, infection severity, mechanical pressure, treatment implementation, and the patient's overall condition.

When healing does not progress, the care team must investigate why. When serious infection or ischemia is suspected, appropriate assessment and escalation should occur without waiting for AI confirmation.

Healthcare organizations must also ensure that recommendations lead to actual care, that patients can access the services they need, and that follow-up continues after wound closure when recurrence risk remains high.

A well-designed monitoring system should make clinically meaningful changes easier to recognize without presenting image measurements as a complete assessment of foot health.

The measure of success is not how many photographs patients upload or how accurately an algorithm identifies wound boundaries.

It is whether patients receive timely assessment, appropriate treatment, and reliable follow-up—and whether healthcare organizations can demonstrate better healing, preservation of limb function, and fewer avoidable complications.

That is where AI-assisted diabetes-related foot ulcer monitoring may make a meaningful contribution to responsible digital medicine, healthcare quality, and patient safety.

Selected References and Regulatory Sources

Vietnamese Clinical Guidance

Ministry of Health of Viet Nam. Decision No. 1530/QD-BYT, dated 24 March 2023, issuing the professional guidance Diagnosis and Treatment of Diabetes-Related Foot Ulcers (Hướng dẫn chẩn đoán và điều trị loét bàn chân do đái tháo đường).

International Clinical Guidelines

International Working Group on the Diabetic Foot. IWGDF Guidelines on the Prevention and Management of Diabetes-Related Foot Disease. 2023 update.

Senneville E, Albalawi Z, van Asten SA, et al. IWGDF/IDSA Guidelines on the Diagnosis and Treatment of Diabetes-Related Foot Infections. Clinical Infectious Diseases. 2023. DOI: 10.1093/cid/ciad527.

Fitridge R, Chuter V, Mills JL, et al. The Intersocietal IWGDF, ESVS, SVS Guidelines on Peripheral Artery Disease in People With Diabetes and a Foot Ulcer. Diabetes/Metabolism Research and Reviews. 2024;40(3). DOI: 10.1002/dmrr.3686.

Bus SA, Armstrong DG, Crews RT, et al. Guidelines on Offloading Foot Ulcers in Persons With Diabetes (IWGDF 2023 Update). Diabetes/Metabolism Research and Reviews. 2024;40(3). DOI: 10.1002/dmrr.3647.

American Diabetes Association Professional Practice Committee for Diabetes. 12. Retinopathy, Neuropathy, and Foot Care: Standards of Care in Diabetes—2026. Diabetes Care. 2026;49(Suppl 1)–S276. DOI: 10.2337/dc26-S012.

Clinical AI and Implementation Evidence

Wartakusumah R, Mukai K, Noguchi H, Oe M. Performance of Artificial Intelligence in Diabetes-Related Foot Ulcer Detection and Assessment: A Scoping Review of Clinical Validation Studies. Diabetes Research and Clinical Practice. 2026;239:113485. DOI: 10.1016/j.diabres.2026.113485.

Zulfiqar M, Yusuf S, Taming MJ, Burhan H. AI-Based Models for Diabetic Foot Ulcer Assessment: Scoping Review. JMIR Diabetes. 2026;11. DOI: 10.2196/77925.

Aldhubiea H, Fetais N, Newaz M, et al. Deep Learning for Diabetic Foot Ulcer: A Systematic Review of Segmentation, Classification and Multimodal Integration. Cluster Computing. 2026;29:826. DOI: 10.1007/s10586-026-06619-3.

Relevant Vietnamese Legal and Regulatory Framework

National Assembly of Viet Nam. Law on Medical Examination and Treatment No. 15/2023/QH15.

Law on Artificial Intelligence No. 134/2025/QH15; Decree No. 142/2026/ND-CP; Decision No. 33/2026/QD-TTg.

Law on Personal Data Protection No. 91/2025/QH15; Decree No. 356/2025/ND-CP.

Decree No. 102/2025/ND-CP on health data management.

Decree No. 98/2021/ND-CP on medical device management, as amended; Ministry of Health Circular No. 24/2026/TT-BYT.

Editorial and Clinical Notice

This article is intended for professional education, responsible healthcare AI governance, and healthcare quality-improvement discussion. It draws on Vietnamese professional guidance, international clinical recommendations, published research, and relevant legal and regulatory sources.

It does not establish a mandatory AI implementation standard, provide patient-specific medical advice, authorize autonomous diagnosis or treatment decisions, or constitute a legal determination of any particular AI product's regulatory status.

The clinical examples and proposed implementation approaches are illustrative. Assessment of diabetes-related foot ulcers, diagnosis of infection or ischemia, treatment selection, specialist referral, wound management, and follow-up must remain under the responsibility of appropriately qualified healthcare professionals, consistent with applicable Vietnamese requirements, current clinical guidance, and locally approved procedures.

AI-generated wound measurements and image classifications must not independently determine whether a wound is infected, whether circulation is adequate, or whether a patient requires urgent or specialist care.

The research discussed does not establish that AI-assisted diabetes-related foot ulcer monitoring necessarily improves healing rates, reduces lower-extremity amputations, prevents hospitalization, or improves survival in Vietnamese healthcare settings.

Claims of clinical effectiveness require appropriate evidence relevant to the intended patient population, system function, and deployment setting, together with local evaluation, professional oversight, and ongoing monitoring.

International clinical guidelines should be interpreted within their intended populations and clinical contexts. Proposed organizational safeguards should not be presented as mandatory Vietnamese legal requirements unless supported by applicable legal provisions.

Digital Medicine Vietnam (DMVN) | Evidence and regulatory review: 1 November 2026.