Tracking of Nutritional Intake using Artificial Intelligence (TONI-AI)

Binge Eating Disorder (BED) is an eating disorder characterized by recurrent
and persistent episodes of binge eating in the absence of inappropriate weight
control methods that are applied by other eating disorder subtypes, such as
purging. Although the prevalence of BED in the Netherlands is unclear, due to
BED being a relatively *new* diagnostic category in the DSM 5, as well as
methodological shortcomings of existing research, based on Australian research
(Hay, 1998) it is estimated that approximately 160.000 people in the
Netherlands suffer from BED.

BED is strongly associated with obesity. Although most people with obesity
don*t have BED, most people with BED are obese and can have the medical
difficulties associated with this condition. Comorbid problems are not only
physical, but also psychiatric. BED itself is often marked by the use of food
to handle emotional distress (Goldfield et al., 2008), along with dysregulation
of interoceptive awareness, appetite and satiety mechanisms (Sysko et al.,
2007). Treatment for BED therefore needs to address the disordered eating and
associated psychopathology, as well as the excess weight. Cognitive behavioural
therapy (CBT) is the first-choice treatment for BED and has the strongest
empirical support so far (NICE 2004, Wilson et al., 2007), resulting in
approximately 50% binge abstinence after treatment. This shows there is much
room for improvement. Also, and similar to alternative psychological and
behavioural treatments, CBT does not result in meaningful weight loss for most
patients (Wilson et al, 2007). Finding ways to improve binge eating and weight
loss outcomes therefore represents a major research priority.

Adding a lifestyle intervention to CBT may possibly enhance binge eating and
weight loss outcomes. Several existing lifestyle interventions for obesity have
targeted portion and eating awareness, as individuals with obesity and BED
generally report a sense of inadequacy using nutritional guidelines (Kristeller
& Wolever, 2011), as they lack awareness regarding eating. The ability to
accurately estimate and measure food portion sizes is important for preventing
and treating obesity (Ayala, 2006) and lifestyle interventions have targeted
portion control and eating awareness in several ways with successful results.
Examples include mindfulness-based eating awareness training (Kristeller &
Wolever, 2011), Mandometer training aimed at decreasing speed of eating and
total intake (Ford et al. 2011), and the use of portion control plates (Kesman
et al, 2011).

In the present study we aim to pilot an innovative tool for a lifestyle
intervention. More specifically, we propose to use an app tracking calorie
intake via photos taken with a smartphone.

This app could not only be useful as a tool for a lifestyle intervention per
se, but also as a reliable method to assess consumption. An accurate assessment
of dietary consumption is particularly challenging in patients with BED,
because of underreporting that is common among obese and overweight
individuals. The use of mobile phones to track and photograph what they eat may
be a more convenient and reliable way to collect data (see for example
Segovia-Siapco & Sabaté (2016)).

This proposed pilot is part of a bigger project led by 5M ICT. In the bigger
project, 5M ICT will apply a deep learning (DL) method for food recognition on
images taken by the user before and after eating, which enables accurate
tracking of nutrition intake. This information will be built into an app with a
food recognition module. Finally, to validate the value proposition, two trials
with several members of a fitness club (not part of the current proposal) and
several patients (N = 3-4) will be performed.

The Centre for Eating Disorder*s contribution to the project is to determine
user requirements before developing the app and testing the app in a small
number of patients with BED, and exploring its feasibility and acceptability,
including dieting behavior.

Primary Objective: testing the food recognition app developed by 5M ICT in a
small number of patients with BED, determining user requirements, and explore
feasibility and acceptability of the app (from the patients* and clinicians*
view), including dieting behavior.
Please note that we do not aim to base any conclusions on the data with regard
to effectiveness. The app will only be tested in order to be developed further.
The end goal is to develop the app using the patients* feedback.

Case series.

It is possible that the use of the app will increase dieting rules. We will
include several questions concerning dieting behavior in the questionnaire.
Patients are in treatment while using the app and the therapist will be able to
intervene.