Search

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder that involves difficulties in social communication. Previous research has demonstrated that these difficulties are apparent in the way ASD children speak, indicating that it may be possible to estimate ASD severity using quantitative features of speech. Here, we extracted a variety of prosodic, acoustic, and conversational features from speech recordings of Hebrew speaking children who completed an Autism Diagnostic Observation Schedule (ADOS) assessment. Sixty features were extracted from the recordings of 72 children and 21 of the features were significantly correlated with the children's ADOS scores. Positive correlations were found with pitch variability and Zero Crossing Rate (ZCR), while negative correlations were found with the speed and number of vocal responses to the clinician, and the overall number of vocalizations. Using these features, we built several Deep Neural Network (DNN) algorithms to estimate ADOS scores and compared their performance with Linear Regression and Support Vector Regression (SVR) models. We found that a Convolutional Neural Network (CNN) yielded the best results. This algorithm predicted ADOS scores with a mean RMSE of 4.65 and a mean correlation of 0.72 with the true ADOS scores when trained and tested on different sub-samples of the available data. Automated algorithms with the ability to predict ASD severity in a reliable and sensitive manner have the potential of revolutionizing early ASD identification, quantification of symptom severity, and assessment of treatment efficacy.

In January 2016 the U.S. National Library of Medicine announced a challenge competition calling for the development and discovery of high-quality algorithms and software that rank how well consumer images of prescription pills match reference images of pills in its authoritative RxIMAGE collection. This challenge was motivated by the need to easily identify unknown prescription pills both by healthcare personnel and the general public. Potential benefits of this capability include confirmation of the pill in settings where the documentation and medication have been separated, such as in a disaster or emergency; and confirmation of a pill when the prescribed medication changes from brand to generic, or for any other reason the shape and color of the pill change. The data for the competition consisted of two types of images, high quality macro photographs, reference images, and consumer quality photographs of the quality we expect users of a proposed application to acquire. A training dataset consisting of 2000 reference images and 5000 corresponding consumer quality images acquired from 1000 pills was provided to challenge participants. A second dataset acquired from 1000 pills with similar distributions of shape and color was reserved as a segregated testing set. Challenge submissions were required to produce a ranking of the reference images, given a consumer quality image as input. Determination of the winning teams was done using the mean average precision quality metric, with the three winners obtaining mean average precision scores of 0.27, 0.09, and 0.08. In the retrieval results, the correct image was amongst the top five ranked images 43%, 12%, and 11% of the time, out of 5000 query/consumer images. This is an initial promising step towards development of an NLM software system and application-programming interface facilitating pill identification. The training dataset will continue to be freely available online at: http://pir.nlm.nih.gov/challenge/submission.html.