Page 62 - AIH-1-2

Artificial Intelligence in Health AI in AD – Diagnosis and monitoring

actively researched to improve acoustic monitoring. These
techniques analyze sound data to detect movement and
quantify scratching activity. Despite the advantages of
acoustic recording over visual surveillance, it has only been
tested in a restricted capacity in the brief trial described
earlier. Therefore, more extensive and comprehensive
research is needed to confirm this strategy across a wider
range of patient demographics. 68

8. Smart devices
In the past 10 years, medical research has gained insights
into a range of medical conditions, and novel therapies have
been introduced subsequently, which include monitoring
dietary intake and sending reminders for medication
adherence, facilitated by the portability and processing
power of smart devices. 69,70 Researchers investigating
pruritus have modified wrist actigraphy for smartwatches
while maintaining its core principles and keeping up with
technological advancements. In a pilot study conducted
71
Figure 5. Diverse data modalities (such as images, genetic data, and by Lee et al., an accelerometer-equipped wristwatch
biomarkers) employed across distinct AI studies. was used with three subjects to identify scratching
Abbreviations: AD: Atopic dermatitis; AI: Artificial intelligence; ANN: tendencies. Remarkably, when compared to infrared video
Artificial neural network; ML: Machine learning; MPT: Multiphoton
tomography. surveillance, the wristwatch demonstrated remarkable
accuracy, with detection rates ranging from 98.5 – 99.0%
for right-hand scratching motions and 93.3 – 97.6% for
automated and significantly more rapid analysis of scratch left-hand scratching.
data, akin to video surveillance but without the need for
direct observation. By contrasting the results of the acoustic In 2017, the Itch Tracker device was developed
counting method with those from video surveillance in the through a joint effort between dermatologists, Nestle
mouse model, the efficacy of the method was confirmed, Skin Health, and Apple Inc. This invention consists of an
revealing a close similarity between the two methods. 67 application (software program) designed for well-known
smartwatches, aimed at tracking nocturnal scratching and
68
In 2014, Noro et al. expanded on the achievements
of the mouse model by developing a sound detector addressing the need for improved objective techniques in
assessing itching. As stated in the study, the application
integrated with a similar acoustic surveillance system, incorporates an algorithm that evaluates acceleration
designed to be worn on the wrist to monitor scratching data from smartwatches, using unique wrist motions to
behavior. This groundbreaking technology distinguished
a particular bone-conducted sound produced by motions distinguish scratching from other types of movement. In
through the finger and wrist bones, rather than relying on addition, the application features a smartphone interface
air-conducted noises. Over a 6-h sleep period, sound data that enables users to respond to surveys linked to itching,
were gathered from both AD patients and healthy controls, thus combining subjective patient comments with
71
while infrared video captured individuals’ scratching objective data from wearables. These results highlight the
motions. The algorithm swiftly analyzed the audio data, effectiveness of Itch Tracker as a smart gadget for tracking
taking only a few minutes compared to the several hours scratching, offering an objective and indirect measure of
required for human observers to score video recordings. pruritus severity in patients with AD. With its subtle design
According to the study, the scratching time recorded by the and user-friendly interface, Itch Tracker is appropriate for
sound detector was later discovered to be almost identical to general use. However, further studies are necessary to
the outcomes from video surveillance, which is considered evaluate the application in various clinical settings and
the gold standard for comparison. As a result, the sound patient demographics. 72
detector greatly accelerated and improved the objectivity 9. Neurological imaging
of the study while also reducing the invasive need for
human behavioral observation. Similar to advances in A unique method for objectively identifying structural
video surveillance, machine-learning algorithms are being and functional changes in both acute and chronic pruritus

Volume 1 Issue 2 (2024) 56 doi: 10.36922/aih.2775