System and method for early detection of mild cognitive impairment in subjects

A system (100), comprising:

• a memory storing instructions(102) and a centralized database (110);
• one or more communication interfaces(106);
• and
• one or more hardware processors(104) coupled to the memory via the one or more communication interfaces, wherein the one or more hardware processors are configured by instructions to:
• receive sensor data from a plurality of sensors, wherein the received sensor data indicates Activities of Daily Living (ADL) pertaining to a subject for a specific time interval;
• substitute, by using Long Short-Term Memory (LSTM), corrupt sensor data in the received sensor data with correct sensor data to generate corrected sensor data in pre-processing module 202, wherein the corrupt sensor data is substituted with the correct sensor data based on a learning pattern and historical data stored in a database;
• generate a semantic vector representation of the corrected sensor data, wherein the semantic vector representation comprises a plurality of sub-vectors, wherein each of the plurality of sub-vectors comprise of a subset of the corrected sensor data of a pre-determined time interval;
• optimize, size of the semantic vector representation of the corrected sensor data to generate an optimized vector data;
• estimate a dynamic threshold parameter (ø) based on one or more similarity values obtained from a comparison of (i) the optimized vector data for the time interval with (ii) an adjacent optimized vector data comprising corresponding corrected sensor data stored in the database in optimizing module 204;
• perform a comparison of the one or more similarities values with the dynamic threshold parameter;
• detect, based on the comparison, one or more abnormalities indicative of behavior deviation of the subject;
• and
• detect one or more abnormalities in the deviated compressed vector data by comparing the deviated compressed vector data with one or more reference vectors.

The non-transitory computer-readable medium having embodied thereon a computer program for detection of Mild Cognitive Impairment (MCI) in a subject, the method comprising:

• receiving sensor data from a plurality of sensors, wherein the received sensor data indicates Activities of Daily Living (ADL) pertaining to a subject for a specific time interval (302);
• substituting, by using Long Short-Term Memory (LSTM), corrupt sensor data in the received sensor data with correct sensor data to generate corrected sensor data, wherein the corrupt sensor data is substituted with the correct sensor data based on a learning pattern and historical data stored in a database (304);
• generating a semantic vector representation of the corrected sensor data, wherein the semantic vector representation comprises a plurality of sub-vectors, wherein each of the plurality of sub-vectors comprise of a subset of the corrected sensor data of a pre-determined time interval (306);
• optimizing, by using an auto encoder, size of the semantic vector representation of the corrected sensor data to generate an optimized vector data (308);
• estimating a dynamic threshold parameter (ø) based on one or more similarity values obtained from a comparison of (i) the optimized vector data for the time interval with (ii) an adjacent optimized vector data comprising corresponding corrected sensor data stored in a database (310);
• performing a comparison of the one or more similarities values with the dynamic threshold parameter (312);
• and
• detecting, based on the comparison, one or more abnormalities indicative of behavior deviation of the subject (314).