Anutosh Maitra

A novel mobile application to assist maternal health workers in rural India

The philosophy and architecture of a novel mobile application to support maternal health are presented in this paper. A deployment study presents evidence on the benefits derived by the use of mobile devices for treatment adherence and compliance and care management. The application is intended for usage by rural caregivers for timely pregnancy related information management and to give relevant pregnancy and child care related advices. Automated advisory to help determine the risk condition of pregnancy and suggested treatment courses are handled by an expert system in the application. Much of the advisory rules run on the edge device right at the point of data collection and the dependency on network availability is negligible. The application also demands minimal medical expertise on the part of the caregivers, so early assistance can be provided even before a trained medical practitioner is made available to the patient.

Rapid mHealth - a mobile healthcare application development framework

A new philosophy and architecture of rapid mobile healthcare application development is presented here. Many monolith applications developed on the Lambda architecture had been successful earlier, but of late, this architecture is found inadequate when deployed for different stakeholders. Todays healthcare applications handle large volume of streaming data obtained from dispersed devices. The architecture is to support multiple use cases as the business models are evolving and the products have to be malleable enough to support different functions. This paper describes a working architecture on which mobile healthcare applications can be rapidly built through ingestion and processing of voluminous streaming data. The use of micro-services keeps this architecture flexible. On the client side, application platform as a service (APaaS) frameworks are employed to assist quick development and validation of solutions. Findings from some successfully deployed mobile healthcare applications based on this architecture are also discussed.

An integrated transport advisory system for commuters, operators and city control centres

This paper describes a solution that provides dynamically updated transport information to the mobile commuters, public transport operators and city control centre in a seamless manner. Successful vehicular traffic management needs more and more people availing public transport. This can be achieved only by optimizing the existing transport systems, managing dynamic demand forecasts and real-time contextual transport information sharing amongst all stakeholders. The integrated framework proposed in the current work benefits commuters both in planning their travel as well as by providing them reactive options to counter unforeseen events. It assists the transport operators to optimally plan vehicle operations for providing improved commuter care. The solution also benefits the city control centres by providing current and perceived near and distant future scenarios. The application is hosted on cloud and scalable to cater to individual commuters and vehicles.

Semantic Question Matching in Data Constrained Environment.

Machine comprehension of various forms of semantically similar questions with same or similar answers has been an ongoing challenge. Especially in many industrial domains with limited set of questions, it is hard to identify proper semantic match for a newly asked question having the same answer but presented in different lexical form. This paper proposes a linguistically motivated taxonomy for English questions and an effective approach for question matching by combining deep learning models for question representations with general taxonomy based features. Experiments performed on short datasets demonstrate the effectiveness of the proposed approach as better matching classification was observed by coupling the standard distributional features with knowledge-based methods.

A Novel Text Analysis Platform for Pharmacovigilance of Clinical Drugs

Abstract Analyzing possible drug safety incidents and generating narratives in pharmacovigilance process have traditionally relied upon manual review of case reports from patients, consumers and healthcare professionals. However, due to the vast quantity and complexity of data to be analyzed and for ensuring timeliness, reduction of cost, consistency of reporting and quality of reporting; role of automated computational systems that can accurately detect adverse drug reactions attached to a suspected drug in a timely fashion have become critical. Pharmaceutical companies have started to realize the need for collaborative and integrative approaches and strategies to allow a faster identification of high-risk interactions between marketed drugs and adverse events, and to enable the automated uncovering of scientific evidence behind them. The fundamental requirement for the automatic processing of biomedical text is the identification of information carrying units such as the concepts or named entities. Additionally, there are regulatory guidance or rules with respect to identifiability of reporters, patients, drugs and interactions in the reports of suspected adverse reactions. Owing to these challenges, the problems of automated unambiguous identification of medical drugs and compounds, detection of adverse drug reactions, and generation of case narratives from the text of the reports are not considered to be adequately solved so far. In this paper, we present a novel text analysis platform that assists in bringing intelligent automation in the process by integrating a medical language processing pipeline and causal reasoning chain, with publicly available large-scale biomedical databases containing structure, bioassay, and genomic information, as well as comprehensive clinical data sets.