Malnutrition in older adults in many high-income countries is an unrecognized condition that is increasing in prevalence as the population ages [1]. The impact of malnutrition on morbidity and mortality in older adults is a problem that necessitates urgent attention. The aging process brings about biological, physiological, social, and psychological changes, coupled with a higher morbidity prevalence, which can negatively affect nutritional status [2].

Malnutrition in older adults, also coined as anorexia of aging, has severe implications, resulting in a significantly poor quality of life [3,4]. The loss of swallowing function, polypharmacy, cachexia, and lack of social support can exacerbate malnutrition [5,6]. Malnutrition occurs because of insufficient food intake, the inability of the body to absorb nutrients, or because of the presence of catabolic diseases such as chronic obstructive pulmonary disease or cancer. This can cause a reduction in fat and muscle mass, further leading to impaired physical and cognitive function [7]. Malnutrition can occur gradually over significant periods without being detected, as the symptoms may not be critical to warrant immediate medical attention. However, the severity of malnutrition may be significantly elevated if it is associated with a disease or inflammation. In geriatric patients with diseases, it is evident that a reduction in food intake causes increased muscle breakdown because of increased resting energy expenditure [8].

Malnutrition has been found to lead to sarcopenia, and sarcopenia may aggravate malnutrition, leading to the aggravation of morbidity and mortality [9,10]. Sarcopenia is defined as an age-related loss of muscle mass, where muscle mass is lost involuntarily over time, leading to a decrease in muscle protein synthesis, strength, and function [11]. In a recent review, both malnutrition and sarcopenia were examined concomitantly, and the findings for the management and interventions of these 2 conditions were found to overlap considerably [12]. Hence, the newly proposed clinical syndrome, called malnutrition-sarcopenia syndrome, should be assessed simultaneously to improve health outcomes [13]. Older adults have an increased risk of malnutrition compared with other adult populations, and it has been estimated that 2% to 16% of community-dwelling older adults are deficient in their protein and energy needs [14]. The effects of poor skeletal muscle mass (SMM) can be detrimental as they reduce the prognosis of many diseases, including diabetes, heart disease, chronic kidney disease, sepsis, and cancer [15-20]. In a community-based, cross-sectional study conducted on 722 stroke- and dementia-free participants aged 50 to 75 years, an increase in SMM was observed to protect against ischemic stroke, especially in men [21]. In another study in which the prevalence of sarcopenia in 414 patients with diabetes and 396 control participants was examined, type 2 diabetes was independently associated with sarcopenia, with a 15.7% prevalence in patients with diabetes compared with 6.9% in the control group [17].

Apart from physiological and metabolic implications, malnutrition and sarcopenia in older adults have been widely established to have financial implications and are burdens to health care resources [22]. Patients who are malnourished are associated with more frequent hospitalizations, increased length of hospital stays, and increased readmissions and hospitalization costs [23]. Approximately 40% of patients are malnourished upon hospital admission, and depending on the severity, they are faced with approximately 31% to 38% increase in hospital costs [24].

However, it is difficult to detect subclinical malnutrition because of the absence of clear indicators. This could lead to a lack of identifying poor nutritional practices among older adults, as malnutrition does not often present with immediate problems. Therefore, better screening options for the diagnosis of subclinical malnutrition must be recognized and identified [25,26]. This could potentially ease the burden on the health care system, reduce the cost of resources, improve the quality of patients’ lives, and extend the healthy living years of older adults.

It is imperative that malnutrition and the prognostic value of treatment options to manage malnutrition be identified. The face is often used by health practitioners to make qualitative assessments of health and nutritional status. Modern advancements in sensor and computing technologies enable features in the face to be accurately digitized and could potentially allow these previously intuitive assessments to be quantified. In this paper, we discuss the feasibility of a novel digital remote method for diagnosing malnutrition using facial morphometrics.

Until 2016, there were no universally accepted screening tools or criteria used as the gold standard to identify malnutrition risk [27]. The Global Leadership Initiative on Malnutrition assembled a set of criteria for identifying and diagnosing malnutrition in older adults in hospitals [27]. This initiative comprises a 2-step method in which a validated screening tool is used to identify patients who are at risk of malnutrition. This is followed by an assessment to diagnose and grade the severity of malnutrition. The Global Leadership Initiative on Malnutrition grouped the diagnosis of malnutrition into two distinct categorical criteria: phenotypic (which included weight loss, low BMI, and reduced SMM) and the other etiologic criteria comprising reduced food intake, disease burden, or inflammation [27].

One of the most widely used traditional methods for assessing malnutrition is the subjective global assessment tool. It is an integrated tool that assesses nutritional status based on patients’ past records and physical examinations, and subsequently classifies patients into three categories: well-nourished, moderately malnourished, or severely malnourished [28]. Anthropometric measurements such as midarm circumference and triceps skinfold thickness have been used as basic methods of correlating the percentage of body fat and lean muscle mass [29]. In addition to the subjective global assessment, bioelectrical impedance analysis and ultrasound techniques have been recognized as accurate methods for measuring lean body tissue, which is a key indicator of nutritional status [30]. Other established methods used to assess SMM include scans using computed tomography (CT), dual-energy x-ray absorptiometry, and magnetic resonance imaging (MRI), which are costly and time consuming [31,32]. Analytic morphomics is an emerging field that uses cross-sectional images to provide a global assessment of a patient beyond the localized specific pathology of interest (Englesbe et al [33]). Through the application of CT-based analytic morphomics, Lee et al [31] developed the Morphomic Malnutrition Score as a means of providing standardization and objective measurements for conventional diagnostics. This method provides an in-depth assessment to distinguish between healthy individuals and individuals who are severely malnourished. Although these methods are highly accurate, conducting them can be cumbersome and resource intensive, as they require a trained practitioner to be present to conduct the assessment, and patients need to travel to specialized facilities to obtain their measurements. In cases where older adults do not frequent clinics or hospitals, malnutrition and sarcopenia could be left undetected for extended periods or until the onset of a disease, leading to several negative implications. If malnutrition monitoring and assessments could be delivered or enhanced through the use of readily accessible and cost-effective imaging technologies, this could greatly reduce the burden on rapidly inflating health care costs. More importantly, the implementation of such methods would significantly improve access to and quality of care for older adult patients who are malnourished.

Digital health care is one such approach that allows for malnutrition and nutritional status to be monitored remotely without requiring the physical presence of practitioners or advanced equipment. As we enter the Fourth Industrial Revolution, the widespread availability and improvements in data science and infocommunication (integration of information and telecommunication technology sectors) technologies have rapidly expanded the vast range of available services and applications in the health care sector [34-39]. Digital health care, in its many forms, is rapidly gaining traction as a scalable means of improving efficiency in the health care system, allowing health care institutions to extend their reach and increase the number of touchpoints they have with patients. Increased levels of connectivity offered by these technologies are changing the ways in which the health care system is structured and administered.

Telemonitoring is the real-time monitoring of patients using mobile technologies to conduct routine medical tests and communicate results to health care workers for evaluation [38]. This capitalizes on the recent progress in Internet of Medical Things devices and widespread internet access, which can range from data gathered from a single device to an entire network of personal digital devices, connected medical devices, implants, and other sensors [38]. The integration of these sensors and collection of accurate real-time data can support clinical decisions by rapidly communicating changes in physiological or biochemical states, improving the overall quality of health care services and leading to more effective management of chronic diseases [34,37,38,40-46]. Digital applications allow conditions to be managed at home or within local communities, offering patients greater levels of independence and autonomy, reducing travel time, and ensuring better equity for patients in rural areas [34,38]. As such, many studies have shown high levels of satisfaction, empowerment, and reassurance expressed by both providers and end users of telemonitoring solutions [34,37,45,47].

Telemonitoring is also considered an efficient use of health care resources [34,37,38,40-44,46,48]. Many conventional tests that require the presence of a health care worker can now be conducted automatically with the use of specialized, portable equipment, reducing the number of face-to-face interactions required of health care workers [34,38]. Patients are also not required to be physically present at health care institutions. The reduction in traffic and congestion allows for the prioritization of more serious cases that require immediate care, a benefit most notably realized during the COVID-19 pandemic [34,49,50]. The improved management of chronic diseases also reduces the number of hospitalization events and the length of hospital stays for many conditions and, consequently, reduces overall health care expenditure.

Given the many benefits that telemonitoring has to offer, it is critical that new applications are developed to broaden the scope of conditions to which telemonitoring can be applied. In both high- and low-income nations, smartphones have been used as platforms for several aspects of life, such as identity verification and personal banking, and are increasingly being adopted for use in the capture, management, and transmission of personal health data [51-54]. These smart devices are outfitted with sensors such as GPS trackers, accelerometers, cameras, and depth sensors and are progressively updated each year with newer and more powerful components. The incorporation of these advanced sensors may offer a scalable solution for population-wide telemonitoring of nutritional status and allow early detection of malnutrition in the community.

A relatively unexplored method that may offer predictive capabilities in the area of malnutrition screening is the application of facial morphometrics. Facial morphometrics is the study of the contours and structures of faces using geometric mapping tools that allow landmarks and features to be identified [55]. Although facial changes during weight loss and undernutrition have been reported since the time of Leonardo Da Vinci, their implications and ability to predict nutritional status have been recognized only recently [56-62].

The newly integrated depth sensors in personal smart devices may present a novel solution that may be beneficial for the remote monitoring of malnutrition. These depth sensors capture 3D depth data of objects, enabling the use of applications such as augmented reality–based applications, as well as the generation of 3D point clouds for 3D modeling [63]. Most recently, a notable application that highlights the current level of sophistication and accuracy of depth sensors is Apple’s 3D face recognition–based encryption system, known as Face ID [64]. The degree of security offered by the Face ID system can be attributed to the level of 3D detail that the True Depth camera can capture. A total of 30,000 infrared dots are projected onto the target object and captured by an infrared-sensitive camera, allowing fine details such as contours and recesses to be detected by the camera [65]. This granularity has enabled its use in a wide range of applications, ranging from the analysis of human facial emotions to the customization of prosthetics and medical equipment requiring precise fit [63,66-71].

With depth cameras becoming the mainstay in smartphones beyond the iPhone, there may be potential for developing telemonitoring applications that involve the extraction and correlation of facial morphological features with health indices. Typical facial landmarks include the eyes, nose, chin, cheekbones, and the overall shape of the face. The extraction and analysis of features such as geodesic distances or ratios between these landmarks can serve as informative discriminatory tools in the quantification of shape and variation [72].

These quantified features could serve as proxy markers for the nutritional status of the human body. This has been demonstrated in several studies, which have observed correlations between superficial tissues that affect the overall morphology of the face such as the zygomatic fat pads and masseter and temporalis muscles with nutritional status and whole-body SMM [73-83]. The quantification of these tissues involves the use of techniques such as ultrasound, MRI, and CT scans. However, these methods can be expensive, time consuming, and stressful for patients. Therefore, the application of facial morphometrics using widely available depth sensors could provide a more accessible and less invasive solution for larger populations.

The use of facial morphometrics in the field of health sciences is not new. However, most of the research has primarily been conducted in the fields of aesthetics, orthodontics, and prosthetics. There have been attempts to explore the correlation between facial features and nutrition-focused indices such as BMI and waist-to-hip ratio [84-91]. Earlier studies used 2D images and relied on planar distances or 2D angularity to identify features that account for BMI variations [87,89,92-99]. Recent studies have explored the use of 3D meshes for more detailed geometric modeling of faces using tools such as 3D scanners or stereophotogrammetry [86,100-102]. However, these technologies require a specialized setup, and individuals must be physically present for the scans to be conducted, thus limiting the number of individuals that can be recruited. The widespread availability of personal smartphones integrated with accurate depth cameras presents an opportunity for more scalable research in this area, allowing the acquisition of larger data sets within a shorter amount of time.

With these capabilities in mind, this paper further discusses the relevance of certain facial features that have the potential to provide diagnostic value in identifying changes in SMM and overall nutritional status.

With imaging tools capable of deriving facial morphology accurately, both direct and indirect tissue losses can potentially be captured as additional features for analysis. Direct measurements of tissue loss can include the quantification of protrusions and depressions that form from tissue atrophy, whereas indirect measurements include features that become more prominent when there is a loss of fat or muscle tissue, such as certain bone processes and ridges, or the angulation of the jawline [73,100,103].

The degree of bone prominence is important in determining nutritional status [104,105]. Physical examinations of malnutrition typically observe the prominence of bone structures in the clavicle, scapular, acromion, and patellar regions to determine the degree of muscle and subcutaneous fat loss [104]. Certain superficial structures that have a lesser degree of covering tissue, such as the zygomatic process or mandible, are immediately distinguishable. Other bone structures such as those in the orbit, zygomatic bone, and other anatomical details of the mandible or jaw may become more prominent in the case of malnourishment, following the loss of muscle and fat tissue [105].

To discuss potential features, we segment the face into three regions: upper face, midface, and lower face (Figure 1).

Undiagnosed malnutrition is becoming a significant problem that can reduce the quality of life of many individuals and inflate the costs of existing health care systems because of the many metabolic complications it can cause. Many high-income countries are faced with a rapidly aging population that may lead to poor food intake, disease etiology, and improper access to food, which may give rise to an increase in the prevalence of malnutrition. Given the need for specialized equipment and trained practitioners for the diagnosis of malnutrition, the development of simpler and more scalable methods for assessing malnutrition has been a long-standing challenge for public health.

Recent advancements in artificial intelligence, computing, data science, and infocommunication technologies present a wealth of opportunities that will shape and revolutionize the landscape of health care. This paper aims to stimulate further discussion and discourse on how adopting this emerging technology could provide real-time access to nutritional status. We explored the feasibility of a novel, noninvasive approach to tackling malnutrition in the community. The use of facial morphometrics extends the use of currently available technology, providing a scalable and easily deployable method for the real-time monitoring of nutritional status. Clinicians and dietitians will be able to keep track of patients remotely and provide the necessary intervention measures as required. Finally, it will furnish health care institutions and policy makers with essential information that can facilitate necessary interventions targeted at affected populations.