Navigation auf uzh.ch
Navigation auf uzh.ch
MSc thesis: Passive sensing data in suicidality
Study background: MULTICAST is an SNSF-funded project involving a consortium of researchers from key disciplines and institutions relevant to investigating suicidality. It aims at developing models to predict suicidal ideation (SI) at 1 month after discharge from inpatient treatment in a sample of up to 200 psychiatric patients admitted in the context of a suicide attempt or acute suicidality. Predictors include (1) baseline variables (structured interview, demographics, clinical, psychological, behavioral (speech) and neurobiological characteristics), and (2) daily diary samplings collected via a smartphone application (i.e., speech, phone sensors, clinical and behavioral variables). Predictions are also made considering varying proportions of daily samplings collected.
Smartphone passive sensing refers to the usage of sensor data collected unobtrusively from phones of the participants to investigate behavioral patterns and derive so-called digital biomarkers that could inform about symptom severity and the course of disease. This is a novel field of research, combined with actively recorded self-report measures sampled throughout the day, to gain novel insight into what characterizes psychopathology.
The proposed work consists of two parts to be divided into an internship, and upon further agreement, into a master’s thesis.
At first, the candidate will perform a literature review to extensively investigate the available custom-built passive sensing platforms. These are software frameworks that allow the collection of passive sensing data that can then be analyzed. Specifically, this will look into custom software developed by research groups for their research purposes into the capabilities and shortcomings of different frameworks. See: link.
This first part will allow to gain a critical understanding of the peculiarities of passive sensing data that will be leveraged in the second part of the process, which can constitute work towards a master’s thesis. In the second phase, passive sensing data collected in MULTICAST will be imported and analyzed. In particular, the candidate will take care of performing ETL operations (extract, transform, load) on passive sensing and EMA data, developing a processing pipeline importing raw data, populating a SQL instance, cleaning it, deriving descriptive analytics, and storing it in its final state. Potentially, in addition to this, it will be investigated whether passive sensing readouts allow to cluster participants into different subgroups and whether it is feasible to predict suicidal ideation and related outcomes.
Requirements (relevant to the second part of the project only):
If interested, please reach out to jacopo.mocellin@uzh.ch.