Validation of the Photography Method for Nutritional Intake Assessment in Hospitalized Elderly Subjects

A Cross-Sectional Reproducibility Study of a Standard Camera Sensor Using Artificial Intelligence to Assess Food Items: The FoodIntech Project

Having a system to measure food consumption is important to establish whether individual nutritional needs are being met in order to act quickly and to minimize the risk of undernutrition. Here, we tested a smartphone-based food consumption assessment system named FoodIntech. FoodIntech, which is based on AI using deep neural networks (DNN), automatically recognizes food items and dishes and calculates food leftovers using an image-based approach, i.e., it does not require human intervention to assess food consumption. This method uses one-input and one-output images by means of the detection and synchronization of a QRcode located on the meal tray. The DNN are then used to process the images and implement food detection, segmentation and recognition. Overall, 22,544 situations analyzed from 149 dishes were used to test the reliability of this method. The reliability of the AI results, based on the central intra-class correlation coefficient values, appeared to be excellent for 39% of the dishes (n = 58 dishes) and good for 19% (n = 28). The implementation of this method is an effective way to improve the recognition of dishes and it is possible, with a sufficient number of photos, to extend the capabilities of the tool to new dishes and foods.

Evaluation of a Novel Artificial Intelligence System to Monitor and Assess Energy and Macronutrient Intake in Hospitalised Older Patients

Malnutrition is common, especially among older, hospitalised patients, and is associated with higher mortality, longer hospitalisation stays, infections, and loss of muscle mass. It is therefore of utmost importance to employ a proper method for dietary assessment that can be used for the identification and management of malnourished hospitalised patients. In this study, we propose an automated Artificial Intelligence (AI)-based system that receives input images of the meals before and after their consumption and is able to estimate the patient's energy, carbohydrate, protein, fat, and fatty acids intake. The system jointly segments the images into the different food components and plate types, estimates the volume of each component before and after consumption, and calculates the energy and macronutrient intake for every meal, based on the kitchen's menu database. Data acquired from an acute geriatric hospital as well as from our previous study were used for the fine-tuning and evaluation of the system. The results from both our system and the hospital's standard procedure were compared to the estimations of experts. Agreement was better with the system, suggesting that it has the potential to replace standard clinical procedures with a positive impact on time spent directly with the patients.

Visual Modeling: A Socialization-Based Intervention to Improve Nutritional Intake Among Nursing Home Residents

Purpose Malnutrition is a widespread, dangerous, and costly condition among institutionalized older adults and can be both a contributor to and consequence of dysphagia for individuals with cognitive impairment. However, interventions to maximize intake in individuals with dementia are limited and frequently problematic, with negative implications for independence and quality of life. The goal of this study was to examine a novel, socialization-grounded intervention based on visual modeling, utilizing the theoretical underpinnings of motor resonance and mimicry. Method To examine the impact of environment on intake, data were collected from four nursing home residents (M _age = 83.5 years, SD = 4.2; three women) with dementia. Weight of food and liquid intake was measured across 15 meals and three different mealtime conditions: the "baseline condition" in which the individual ate alone, the "watch condition" in which the individual ate in the company of a "mealtime buddy," and the "eat" condition in which the individual consumed a meal while the "mealtime buddy" did the same. Results Data visualization supported a weak functional relation between eating environment and amount of intake consumed across participants. Log response ratio estimates suggested a trend for increased weight of food consumed during the eat condition as compared to baseline and the eat condition as compared to the watch condition for some participants. Conclusions These results preliminarily support the benefit of a visual model for increased consumption in some individuals with dementia. The presence and magnitude of the effect across conditions varied based on individual-level factors, such as cognitive status, which has implications for implementation. Overall, this study provides initial proof of concept regarding the use of visual modeling as an intervention approach, laying the foundation for larger scale future studies.

Is Weight Loss More Severe in Older People with Dementia?

Weight loss, a hallmark feature of dementia, is associated with higher mortality in older people. However, there is a lack of consensus in the literature as to whether the weight loss commonly observed in older people with dementia results from reduced energy intake and/or increased energy expenditure. Understanding the cause of energy imbalance in older people with dementia would allow more targeted interventions to avoid detrimental health effects in this vulnerable group. In this paper, we review studies that have considered weight change, energy intake, and energy expenditure in older people with and without dementia. We critically assess the studies' methodology and outline the various factors which may decrease and increase energy intake and expenditure respectively in older people with and without dementia. Current available literature does not support the view that there is a lower energy intake and/or a higher energy expenditure in older people with dementia when compared to those without dementia. The need for more high-quality studies is also highlighted in order to shed more light towards this issue which continues to elude researchers and clinicians alike.

Assessment of Dietary Intake Using Food Photography and Video Recording in Free-Living Young Adults: A Comparative Study

BACKGROUND

Conventional methods of dietary assessment are prone to recall bias and place burden on participants.

OBJECTIVE

Our aim was to compare the performance of image-based dietary assessment (IBDA), including food photography (FP) and video recording (VR), with the criterion of weighed food records (WFR).

DESIGN

In this comparative study, participants captured meals using FP and VR before and after consumption, over 2 days. Food type and portion size were assessed using the images and videos. Energy and nutrient intakes (mean of 2 days) were compared against WFR.

PARTICIPANTS/SETTINGS

Eighty-four healthy adults (mean [standard deviation] age = 29 [8] years), recruited through advertisement in Glasgow, UK, between January and August 2016 were enrolled in the study. Eighty participants (95%) (mean [standard deviation] age = 28 [7] years) completed the study and were included in the analysis.

MAIN OUTCOME MEASURES

Agreement in estimated energy and nutrient intake between WFR and IBDA. The IBDA method feasibility was evaluated using a questionnaire. Inter-rater and intra-rater reliability were assessed.

STATISTICAL ANALYSIS PERFORMED

The performance of the IBDA methods against WFR and their inter and intra-rater reliability were tested with Bland-Altman plots and Spearman correlations. Intra-class agreement between methods was assessed using κ statistics.

RESULTS

Inter-rater reliability was strong for both IBDA methods in estimating energy intake (ρ-coefficients: FP = 0.80; VR = 0.81). There was no difference in the agreement between the 2 assessors. Intra-rater reliability was high. FP and VR underestimated energy intake by a mean (95% agreement limits) of -13.3% (-56.4% and 29.7%) and -4.5% (-45.5% and 36.4%), respectively. IBDA demonstrated moderate-to-strong correlations in nutrient intake ranking, median ρ-coefficients for all nutrients: FP = 0.73 (interquartile range, 0.09) and VR = 0.82 (interquartile range, 0.02). Inter-class agreement of IBDA methods was moderate compared with the WFR in energy intake estimation. IBDA was more practical and enjoyable than WFR.

CONCLUSIONS

IBDA and VR in particular demonstrated a moderate-to-strong ability to rank participants' dietary intake, and considerable group and inter-class agreement compared with the WFR. However, IBDA was found to be unsuitable for assessment in individuals.

Development and validation of nutrient estimates based on a food-photographic record in Japan

Background

Previous studies have reported that estimates of portion size, energy, and macronutrients such as carbohydrates, proteins, and fat based on the food-photographic record closely correlate with directly weighed values. However, the correlation based on a large sample of the test meal with the evidence of many nutrients is yet to be determined. We conducted this study to assess the correlation and difference between the food-photographic record and weighed results for 44 nutrients based on a larger number of test meals than those in previous studies.

Methods

We assessed the nutrients of test meals using a food-photographic record and direct weighing and compared the results of the two methods. Twenty participants prepared a total of 1163 test meals. Each participant cooked 28–29 different kinds of dishes. Five participants cooked the same dish with their own recipes. For the most commonly consumed 41 dishes, 20 participants served a meal with their usual portion size. For the remaining 73 dishes, five participants served a meal with their usual portion size. An independent researcher weighed each ingredient and calculated the nutrients of the test meals. The participants took photographs of the test meals using a digital camera. Two independent, trained analysts measured the longitudinal and transverse diameters of the food area on the photographs of the test meals, compared the portion size with the reference photographs, and calculated the nutrients based on a database that contained reference photographs.

Results

Rank correlation coefficients between estimates from the food-photographic record of each test meal and weighed results were high for portion size (r = 0.93), energy (r = 0.93), protein (r = 0.90), fat (r = 0.92), and carbohydrate (r = 0.94), and those for the 44 nutrients ranged from 0.78 to 0.94. We found high reproducibility between the two analysts for all the nutrients (r > 0.90).

Conclusions

We found a high correlation and small difference between the food-photographic record method and weighed results of a large number of nutrients in many test meals.

An Artificial Intelligence-Based System for Nutrient Intake Assessment of Hospitalised Patients

Regular nutrient intake monitoring in hospitalised patients plays a critical role in reducing the risk of disease-related malnutrition (DRM). Although several methods to estimate nutrient intake have been developed, there is still a clear demand for a more reliable and fully automated technique, as this could improve the data accuracy and reduce both the participant burden and the health costs. In this paper, we propose a novel system based on artificial intelligence to accurately estimate nutrient intake, by simply processing RGB depth image pairs captured before and after a meal consumption. For the development and evaluation of the system, a dedicated and new database of images and recipes of 322 meals was assembled, coupled to data annotation using innovative strategies. With this database, a system was developed that employed a novel multi-task neural network and an algorithm for 3D surface construction. This allowed sequential semantic food segmentation and estimation of the volume of the consumed food, and permitted fully automatic estimation of nutrient intake for each food type with a 15% estimation error.

Validation of a novel image-weighed technique for monitoring food intake and estimation of portion size in hospital settings: a pilot study

OBJECTIVE

Optimal nutrition for hospital patients is crucial and routine monitoring of patients' nutrient intake is imperative. However, personalised monitoring and customised intervention using traditional methods is challenging and labour-intensive, consequently it is often neglected in hospital settings. The present pilot study aimed to examine the reliability and validity of the Dietary Intake Monitoring System (DIMS) against the weighed food method (WFM).

DESIGN

The DIMS 2.0 is composed of an integrated digital camera, weighing scale, radio-frequency identification sensor and WIFI connection for real-time image and weight dietary data acquisition and analysis. The DIMS equipment was used to collect data for a paired set of meals both before and after meal consumption at lunchtime.

SETTING

Odense University Hospital, Denmark.

SUBJECTS

Photos and weights of seventeen patient meals were captured.

RESULTS

The results showed a significant correlation between DIMS and WFM for energy (r=0·99, P<0·01) and protein intake (r=0·98, P<0·01). Intraclass correlation coefficients (ICC) revealed a high degree of agreement among the four non-trained assessors for estimates of portion size of each food item before (0·88, P<0·01) and after consumption (0·99, P<0·01). The ICC for energy and protein intake were 0·99 (P<0·01) and 0·99 (P<0·01), respectively. Bland-Altman plots revealed no systematic bias.

CONCLUSIONS

Considering the huge benefits associated with routine monitoring, technological advances have made it possible to develop a novel, easy-to-use DIMS that, according to the findings, is a valid alternative for use in hospital settings.