Тойм: | In order to address limited access to health care in low-resource parts of the world, our group at MIT, over the past 7 years, has developed a variety of health screening tools that rely on a smartphone with access to a remote server. This mobile health platform, known as "PyMed," consists of a Django server, Postgres data base, and a variety of Bayesian network machine learning and data processing algorithms implemented in Python. While several different server platforms have been demonstrated by our group over the past few years, a great deal of additional development was required to deploy these technologies in a real-world scenario. In addition to the mobile application software and machine learning algorithms, actual deployment of these technologies required the development of transaction sequences and work flows that enable a health worker, or doctor, in the field to collect data from a patient, process the result in real-time on the server, and then receive a complete and usable result on the mobile phone. In this thesis, I discuss the detailed workflow and underlying technology required to perform diagnostic health measurements in a real-world setting. I present the various software modules and server API work that needed to be developed. In addition, I describe how the health results were designed and ultimately presented in a simple and usable format that both the patient and health worker could use and understand. In this work, I describe two disease categories: cardiovascular disease and pulmonary disease. In total, our group has developed two separate servers, 11 mobile apps, and multiple algorithms for signal processing and diagnostic prediction for these two disease categories. The work in this thesis was completed in preparation for several field studies in Bangladesh: (1) a study with coronavirus patients in the NIDCH Hospital in Dhaka, Bangladesh; and (2) an efficacy study with private community health workers in two low-resource areas of Chittagong District and Jamalpur, Bangladesh.
|