Automatic Gestational Age Estimation Tool using Baby Feet Photographs

Singhal, Gautam (2016) Automatic Gestational Age Estimation Tool using Baby Feet Photographs. [Dissertation (University of Nottingham only)]

[thumbnail of Gautam Singhal 4245504.pdf] PDF - Registered users only - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Download (3MB)

Abstract

Medically the newborns are classified into five classes namely, extremely premature, very premature, moderately premature, term and late. Both premature and post-mature babies have poor survival rates because of their vulnerability to microbial infections and low immunity to many life threatening diseases. The more premature or post-mature the newborn is, the less are the chances of survival of the newborn. These lives can be saved by identifying the newborn’s gestational age which can be utilised to find the best treatment available for that age. There are some traditional methods to find the gestational age of the newborns which can be used both prenatally or postnatally. But usually these methods either require expensive machinery and/or well trained personnel who are not always easily available in developing countries. This is why neonatal mortality rate is high among developing nations. This project aims to create a model which can automatically estimate the gestational age of infants from the photographs of their foot. This model would be useful pertinently for the countries with poor or inadequate medical facilities as it voids the requirements for any expensive machinery or trained personnel. This project analysed the subject by making extensive use of machine learning and deep learning. The image classification shows promising results that is based on Artificial Neural Network and Convolutional Neural Network architectures. In order to find the best fit model for this real world issue, different architectures were trained and tested to classify the newborns according to their gestational age. ANN architectures have showed an encouraging performance for the classification. However, CNN architectures showed poor results and would require more data for better results of classification.

Item Type: Dissertation (University of Nottingham only)
Keywords: Gestational age, feature extraction, Artificial Neural Networks, Convolutional Neural Networks.
Depositing User: Gonzalez-Orbegoso, Mrs Carolina
Date Deposited: 18 Jan 2017 12:34
Last Modified: 19 Oct 2017 17:36
URI: https://eprints.nottingham.ac.uk/id/eprint/39163

Actions (Archive Staff Only)

Edit View Edit View