Flashback's proprietary CipherSensor computational engine integrates many algorithms derived from a variety of fields, including machine learning, statistics and robotics. CipherSensor is a coherent integration of techniques that include classification, regression, clustering, semi-supervised learning, reinforcement learning and active-ongoing learning, with feature extraction techniques derived from robotics and computer vision. This integration enables efficient discovery and development of models for estimation, prediction and control, which CipherSensor derives autonomously from extremely large (many terabytes of) data. As data increases or changes, CipherSensor is able to efficiently integrate this new knowledge into existing models without the need for relearning the model from scratch. By design CipherSensor is ideally suited to address the analytics challenges of “Big Data” as it exists now, and as data increase in complexity and size in the future.
Rapid advances in electronics, sensing and imaging technologies are leading to an enormous surge in the amount of data available to clinicians who care for critically ill or injured patients. Optimal patient outcomes generally depend on the speed and accuracy of sorting through this data, determining the underlying problems (internal bleeding for example), anticipating the needs of the patient and deciding on a course of action. Maintaining constant awareness of how a seriously ill patient is responding to resuscitation efforts is also critically important in the delivery and coordination of care. The current generation of physiological sensors are designed to generate raw sensed data, rather than to generate interpreted information from these raw data. Healthcare providers are physically unable to continuously monitor multiple vital signs from multiple patients 24 hrs/day. Rather, only short periods of time and discrete data points can be examined, typically on an intermittent basis. Decisions regarding ongoing care and the application of life-saving interventions are made from qualitative “snap-shot” observations, without the benefit of observing trends and the dynamic nature of the evolving pathophysiology of illness or injury. Flashback, utilizing CipherSensor technology, is building a set of algorithmic solutions to address these unmet needs.
Flashback is also taking steps to identify data analysis challenges afflicting other industries in which the CipherSensor technology platform can play a beneficial role. Increasing numbers of challenges across a variety of industries are associated with real time analysis of very large volumes of continuous, complex data streams generated from multiple, industry-specific sources. Using traditional, mainstream mathematical analyses to deal with these growing data streams leads to, inefficiencies and inaccuracies in data analysis, compromised data interpretation, delayed decision making, and suboptimal outcomes.