Quahog Semi-Supervised Decision Platform is designed to personalize health care to every individual patient, based on their current health data and history. The unique data model design allows the platform to deliver accurate insights and predictions within a short time span.
To provide a better understanding of cellular biology, we have integrated proteomic, transcriptomic, metabolomic and other associated molecular information, in order to draw relationships right down to its unit parameters. This allows us to visualize any condition to its deepest influence and arrive at solutions with the least offset with its associations. The holistic data model allows in understanding individual process patterns of every disease/malfunctions listed by International Classification of Diseases.
Our belief stems from the fact that the processes of the body are interlinked and finite. Monitoring main cellular processes like replication, production, cellular digestion, and cell death would give us enough data to predict scenarios and avert any possible fall-outs. You can refer this post for a deeper analysis on macro-level processes within the body, where addressing different kinds of stress, replication errors and mutation can take care of any premature cell death, ensuring cell cycle optimization.
The below diagram illustrates how human body parameters are unified using a hierarchical semantic graph where associations between subcategories are formed as indicated. Data is arranged by a unique patient which allows the system to analyze every associated parameter true for a single patient, taking our generalization errors out of the picture.
Just to explain the illustration better, you can compare this with a regular doctor visit. Patient lands up in a clinic and shares his symptoms (inputs) to the doctor. The doctor might further examine his body parts (inputs). Using these inputs, the doctor might arrive at possible scenarios with respect to the patient (body)
To further confirm on these possible scenarios, the doctor might ask the patient to run some tests. If the doctor inputs the same information to the system, the system can quickly draw up scenarios that are associated with the inputs. Without neglecting any possibility, the machine can show up as many possibilities. It would show all these possibilities with a probability weight so that the high probability scenarios are ranked higher.
Based on inputs, the association percolates to the deep layers of cellular and molecular levels to identify which process is failing or underperforming and quickly come up with remedy plan that can quickly bring back the process to normalcy causing little or no damage to other true associations.
Every year, out of 55.3 million people who die, more than a million deaths occur from preventable harm in hospitals. This data model can prevent these deaths as the possibility of misdiagnosis is reduced to negligible measures. The platform can prompt doctors to check for missing inputs using pattern prediction algorithm.
Based on the diagram presented above, the platform stacks data by individual patients so the analysis engine incorporates historical data to its computation. The data stack is converted to a string which holds aggregate information using which the analyzer engine decodes and uses necessary information to analyze and output.
The analyzer runs up the Neural Network with pre-classified semantic data model allowing in real-time feature extraction, pattern detection, auto-segmentation, weight calculation and response selection. In order to make response selection in real-time, the network is made dynamic where states are changing continuously until they reach an equilibrium point and remain until the input changes. The final state is achieved through a staged manner based on a feedback received from previous iteration’s output. The output state is updated back to the string for ready consumption.
Based on the request of specific parameters for diet personalization or drug personalization, the string is delivered to the respective touch-points.
The string can be decompiled to get substrings containing raw data (collected from IoT Devices, Wearables, Nanodevices, Medical Equipment or Manual Entry), synthesized data, prescriptive data, observational data, overall historical data and constants for deviation analysis.
The synthesized data can be readily used to serve recommendation for diet personalization or drug personalization.
In future, most of the body activities can be tracked in Vivo using nanodevices. If this tracking data is seamlessly integrated, we can identify any conditions at its infancy and quickly come up with measures to negate malfunctions using simpler diet therapies or off-the-shelf drugs. This will allow us to stay aware of our body at all times and make a quick effort to administer measures to achieve homeostasis.