inCognita
continuous monitoring of mental disorders and progressive neurological diseases
inCognita helps detect signals related to mental disorders and progressive neurological diseases. inCognita brings real-time monitoring that can facilitate earlier diagnosis and more effective treatments.
inCognita is a unique data science framework that Volv have developed.
“the dementias are responsible for the greatest burden of disease, with Alzheimer's disease and related disorders affecting over 7 million people in Europe, and this figure is expected to double every 20 years”[1].
With uptake of smartphones, if we can harness the sensor data effectively, we can produce solutions that address this nexus of disease growth and technology ubiquity, cost-effectively.
We have applied our data science in a repeatable way to be able to detect and predict many different types of cognitive and neurodegenerative diseases passively, with consent, from smartphone devices.
And it is a cohort of 1
Each person's experience is unique
Continuous monitoring, with consent, is now possible unobtrusively
The ubiquity of smartphones in modern life suggests the possibility to use them to continuously monitor patients, for instance to detect undiagnosed diseases or track treatment progress. Such data collection and analysis may be especially beneficial to patients with
as real-time monitoring could facilitate earlier diagnosis and more effective treatment.
Volv has developed a new method of leveraging passively-collected smartphone data and machine learning to detect and monitor brain disorders such as depression and Parkinson’s disease. Crucially, the Volv inCognita framework is able learn accurate, interpretable models from small numbers of labeled examples: i.e., smartphone users for whom sensor data has been gathered and disease status has been determined.
Predictive modeling is achieved by learning from both real patient data and ‘synthetic’ patients constructed via adversarial learning. The inCognita framework is shown to outperform state-of-the-art techniques in experiments involving disparate brain disorders and multiple patient datasets.
Crucial to management of mental health is the ability to both detect appropriate intervention points as well as determine context sensitive interventions. Importantly, Conginitive Behavioural Therapy (CBT) has been shown to have significant impact on reducing disease progression and is a digital intervention that can be triggered in a context sensitive way through inCognita.
As previously mentioned, the increasing prevalence of smartphone usage globally permits patients to be continuously monitored [1-3], suggesting the possibility to track their health status, detect undiagnosed diseases, evaluate treatment progress, and predict the emergence of new conditions.
Such data collection and analysis may be particularly beneficial to individuals suffering from
inCognita delivers the possibility of using passively-collected smartphone data and machine learning to detect and monitor brain disorders in individuals. To increase generality, with inCognita we assumed patients’ smartphones are equipped with only standard sensors: GPS (latitude-longitude coordinates), accelerometers, microphone, ambient light sensor, and screen usage detection.
We focus on depression and Parkinson’s disease as important and illustrative example disorders [e.g. 1-15], but emphasize that the proposed analytic methodology is expected to have broader utility.
While promising, proposed smartphone-based methods for detecting and monitoring brain disorders possess several limitations [1-15]. Perhaps most importantly, they suppose requisite models can be induced via conventional statistical or machine learning algorithms [e.g. 20] and thus experience difficulties in the usual situation where only a small amount of labeled training data is obtainable (i.e., there are few individuals for whom target disease status has been determined and smartphone data has been collected). Additionally, existing detection/monitoring systems:
Volv has developed algorithms for learning smartphone-based brain-disorder detection models which overcome the limitations of previous schemes and shows that the learned models outperform state-of-the-art tools. More specifically, we
See inCognita Data Science for a brief overview.
This represents a new method for leveraging passively-collected smartphone data and machine learning to detect and monitor brain disorders such as depression and Parkinson’s disease. Crucially, the algorithm is able learn accurate, clinically-compatible models from small numbers of labeled examples (i.e., smartphone users for whom sensor data has been gathered and disease status has been determined). Predictive modeling is achieved by learning from augmented training sets composed of both real and ‘synthetic’ patients.
The inCognita framework is shown to outperform state-of-the-art techniques in experiments involving disparate brain disorders and multiple patient datasets. Future work Volv plans to undertake will include exploring the efficacy of the approach for other disorders.
inCognita example business case
inCognita helps detect signals related to mental disorders and progressive neurological diseases. inCognita brings real-time monitoring that can facilitate earlier diagnosis and more effective treatments. Here we present a business case relevant to Switzerland
Two areas that we have already applied the framework to, are outlined below:
Major depressive disorder
Depression is a leading cause of disability worldwide, estimated to affect more than 350M people [16]. The DSM [17] lists nine symptoms commonly associated with major depressive disorder: depressed mood, diminished interest and pleasure in activities, fatigue, restlessness, sleep change, weight change, diminished ability to concentrate, feelings of worthlessness, and thoughts of death and suicide. Several of these symptoms appear to have proxies which can be continuously and conveniently inferred from the outputs of sensors available in (standard) smartphones. This observation indicates it may be possible to recognize depression through computational analysis of passively-gathered smartphone data, and indeed this possibility has attracted the attention of clinicians and researchers (see, for example, [1,2,4-9] and references there-in).
Parkinson's disease
Parkinson’s disease (PD) is the second most prevalent neurodegenerative disorder worldwide, and incidence is rising [18]. While current treatment strategies aim to improve symptoms, increasing effort is being devoted to developing therapies which can slow or prevent PD progression. Such proactive treatment methods are likely to be most effective when used early in the disease process, before substantial neuronal loss has occurred [19]. However, existing methods for detecting the onset of PD are invasive and expensive, impeding early diagnosis [e.g. 13,18,19]. Symptoms of PD such as motor disfunction and speech impairment may have proxies which can be conveniently and in-expensively measured with smartphone sensors, suggesting the potential to efficiently recognize PD through analysis of smartphone data [10-15].
