Page 147 - Handbook of Deep Learning in Biomedical Engineering Techniques and Applications

136 Chapter 5 Depression discovery in cancer communities using deep learning

Online support groups (OSGs) and cancer communities on so-
cial media platforms such as Twitter comprise both the mindful
and unconscious sentiments of cancer patients. A more precise
and stronger conceptualization of the role of social media plat-
forms in detecting health-related sustenance is very important.
Authors in Ref. [60] investigated 660 tweets of cancer survivors
who enumerated their circumstances on the social media plat-
form Twitter. They automated the method of enumerating these
self-reported diagnostic pointers by using ML for SA. Their anal-
ysis led to several findings worth noting about OSGs. They found
out that a fraction of the users on the OSG were responsible for
majority of the posts, reinforcing the finding that posts on OSGs
are largely controlled by a handful of users. In particular, the SA
was toward Tamoxifen, a substance that blocks the female hor-
mone estrogen in the body.
Authors in Ref. [61] analyzed the comments posted by breast
cancer patients on OSGs and concluded the more likelihood of
context which can provide the support for like-minded patients.
Twitter as a social media platform has contributed potential sup-
port for monitoring public health trends as reported by authors in
Ref. [62,63]. In Ref. [62], the authors analyzed the Twitter postings
of over 204 people and identified depressed and healthy individ-
uals out of these using a number of different features for training
and building supervised learning models, namely, affect, linguistic
style, and so on. They concluded that the onset of depression is
indeed detectable from users' Twitter posts many months before
clinical diagnosis of depression is possible. In Ref. [63], the au-
thors also conducted a feasibility analysis to gauge whether
Twitter could be used as a medium to carry out depression detec-
tion and be an alternative for clinical depression diagnosis. The
authors do this by calculating semantic similarities among tweets
in their training set and test data set using WordNet. The experi-
mental results obtained by them indicate the suitability of online
platforms for depression detection and in particular the use of
Twitter as a medium for depression detection. Rodrigues et al.
[64] have developed an SA health tool named as “Senti-Health-
cancer (SCH-pt).” It is tailored specifically to detect sentimental
messages of patient's online, i.e., positive, negative or neutral
messages over online platform. In this, the authors identify senti-
ment of cancer patients in Brazil analyzing their posts written in
Portuguese language making use of tools to analyze sentiment
in Portuguese or converting them to English and analyzing senti-
ment making use of tools that carry out SA in English. The posts
obtained for SA were extracted from the social networking web-
site, Facebook.