First Workshop on Artificial Intelligence for Function, Disability, and Health

Transplantation outcomes focus has shifted beyond increasing survival to decreasing the negative effects of liver disease, focusing on outcomes related to physical and social health. These measures have been studied as isolated variables, but they have not been examined as a cluster of recipient characteristics, representing their wellbeing. This paper aims to compare liver transplantation recipient’s general health status pre- and 2-years post-liver transplant, and to examine whether age, gender, race, and comorbidities are associated with better health status post-transplant. We used data derived from electronic health records of recipients 18 years or older who underwent liver transplantation between 01/01/2008 and 3/31/2017. We excluded recipients who died within 2 years from transplant or did not have follow-up data. A Cox proportional hazard model was used to build severity scores for health status pre- and 2 years post-transplant. Age, gender, race, and comorbidities were also examined. A t-test and ANCOVA were used to examine differences pre- and post-LT. Results showed that better health status pre-transplant was not statistically significant associated with better health status post-transplant. However, health status posttransplant was less variable than pre-transplant. There was a statistically significant association between female gender and kidney severity with worse health status post-transplant; thus, gender and kidney disease may be associated with liver transplant recipients’ wellbeing and play an important role in health status post-transplant.

Paper PDF

We present preliminary work-in-progress results of a project focused on developing a conversational agent system to help with training certified assessors in conducting assessments of functioning in activities of daily living. To date, we have designed a modular task-based conversational agent system and collected hypothetical dialogue data required for training system components as well as a knowledge base needed to generate a wide variety of synthetic profiles of "individuals" being assessed. One of the key components of the system is the topic tracking module that determines the current topic of the conversation. We report the results of experiments with several machine learning approaches to topic/domain classification. The highest accuracy of 83% was achieved with a bidirectional long short-term memory (BiLSTM) model with pre-trained GloVe embeddings. In addition to these results, we also discuss some of the other challenges that we have encountered so far and potential solutions that we are currently pursuing.

Paper PDF

The use of affect-sensitive interfaces carries the promise of enhancing the human-computer interaction by delivering a system capable of identifying a user's emotions and adapt its content accordingly. Today's technology shows great potential to support children with autism, for example by using computer systems to improve their social skills. Generally, however, this technology does not encompass the potential of affect-sensitive interfaces. This is mainly due to Emotion Detection (ED) models built for the general population usually do not perform well when applied to children with autism, who express emotions differently. The aim of this project is therefore to build a person-independent Multimodal Emotion Detection system tailored for children with high-functioning autism for the ultimate goal of applying it to design affect-sensitive interfaces dedicated to children with autism. This is a work in progress and the project expects to build upon the current body of knowledge on methods to apply ED systems to this specific subset of the general population. We expect to apply the overall theoretical and practical design perspectives that arise from this research investigation (e.g. analysis of modalities and features extraction, behavioural cues based features, fusion layers and classifier techniques) to propose a guiding framework for future studies.

Paper PDF

One of the challenges faced by blind people is the difficulty in identifying objects with concise in-formation. They could only rely on the senses of hearing, smell, taste or touch to engage and get some perspectives of objects. Hence, this paper presents a mobile application called Iris to aid blind people in “visualising” their surroundings with descriptive objects. Iris combines the multiple object detection and optical character recognition capabilities of Microsoft Computer Vision API to turns smartphones into assistive devices for the blind to use in their daily activities.

Paper PDF

Action monitoring system used in households offers vital information for health monitoring particularly with aging residents. While visual inputs such as information provided by cameras can recognize the actions and position of a subject with high accuracy, they are not widely accepted due to privacy concerns. This paper proposes a DCNN for posture classification with the use of a low-resolution thermal sensor. The sensor aims to protect the subject’s privacy by capturing visual input in the infrared spectrum as well as having a low spatial resolution of 8x8 pixels. We consider simulation which recreates the experimental environment and produce artificial data for this postural-behavioral problem. The validity of this method is checked by considering 3 postures; standing, sitting, and laying down and cross-examining with field data. Additionally, we explore optimal position and angle of the sensor as well as the effects of color depth on accuracy. In our results we achieved over 93% classification accuracy by color conversion of the infrared array sensor image and successfully decreased loss due to displacement by DCNN. We discovered higher accuracies are achieved when the sensor is located 50cm below a subject’s height with a tilt angle of ±2°.

Paper PDF