Grand Challenges: Computing on a Shoestring

Speaker: Cathy Holloway, Licia Capra, Danny Alexander, Delmiro Fernandez-Reyes
UCL Contact: Sarah Turnbull (Please note this is an internal UCL-CS event).
Date/Time: 06 Oct 17, 12:00 - 14:00
Venue: Torrington (1-19) 115 Galton LT

Abstract

The fourth of the Eight Challenges will be “Computing on a Shoestring” on the 6th October at 12:00 in 115 Galton LT in 1-19 Torrington Place. Food and drink to be provided. The talk will be split into 4 from the following speakers and topics, if you have not already done so, please either accept the invite on your calendar or email me to confirm you plan to come (for catering purposes):

1. Cathy Holloway: Street Rehab

Disability is acknowledged as a cause and consequence of poverty; and assistive technologies (ATs) essential to reducing global inequalities. However, most people who need them don’t have access to ATs and when they do receive them the environment is unsuitable for their use, and they have received little training. We present results from a GCRF pilot project in Delhi to use IoT technology to map wheelchair accessible routes. We compare this to work done in London and look ahead to how we will combine rehabilitation metrics with this data to provide training for people in cities such as Delhi.

2. Licia Capra : Poverty on the Cheap

Teaser: The number 1 Sustainable Development Goal put forward in 2016 by the United Nations is to end poverty. To be able to do so, one first needs to identify where policies and interventions are needed. At the moment, areas of deprivation are mostly identified by means of household surveys, whose cost is so high they take pace very infrequently (e.g., every 10 years) and covering only a sparse sample of the population. In this talk, we cover recent work done that uses alternative data sources (e.g., telecommunication data, satellite imagery, map data), to extract features and build models that accurately estimate economic deprivation, at fine spatio-temporal granularity and (almost) no cost.

3. Danny Alexander: Image Quality Transfer

One-off super-powered medical imaging devices can produce images of unprecedented resolution and information content. However, such devices have little impact on clinical practice where much more modest devices are in day to day use. Image quality transfer aims to learn mappings from low to high quality images. It offers a new imaging paradigm exploiting knowledge from small amounts of very high quality data to inform and enhance large amounts of readily available data. The idea has added potential in LMIC healthcare, which typically relies on decades-old imaging technology, by enhancement via learning mappings to data acquirable from state-of-the-art scanners available in HIC hospitals.

4. Delmiro Fernandez-Reyes : Fast Accurate and Scalable Diagnosis: Malaria Global Health Challenge

Half of the World population is at risk of contracting malaria. The global incidence of clinical malaria is estimated at about 300 million cases that lead around one million deaths each year. Up to eighty-five percent of the cases occur in sub-Saharan Africa with about 90% mortality in the under five years–of-age group due to severe malaria. Control of malaria remains a major public health issue in sub-Saharan Africa developing countries. Large all-year-round lethal malaria morbidity and mortality burden has hindered wellbeing and economic development within the West Sub-Saharan Africa. Key strategies for malaria control has been to reduce mortality by rapid treatment with antimalarial drugs, but this has been stalled by the lack of scalable accurate diagnosis methods.

Human-microscopic examination of blood smears remains the “gold standard” for malaria diagnosis but it has major severe drawbacks (1.very slow ie. taking up to two hours; 2.non-scalable in large urban settings of sub-Saharan Africa; 3.very susceptible to human operator fatigue; 4.requires skilled and experienced personnel and 5.not widely available in rural and remote areas). Consequently, anti-malarial therapy is often based on a presumptive diagnosis (without microscopic confirmation) which is wasteful of drugs and ineffective if the diagnosis was wrong, a drain on often precious health resources, fuels antimalarial resistance and have made control and elimination interventions unachievable.

We aim to achieve fast, accurate and scalable malaria diagnosis by replacing human-expert optical-microscopy with a robotic automated computer-expert system that assesses similar digital-optical-microscopy representations of the problem. Our research challenges focus on the integration of robotics, computer vision and machine learning to design, implement, deploy and test a fully automated system capable to tackle the major drawbacks of human-operated light-microscopy in the diagnosis of malaria. This in turn will underpin the design and development of future portable accurate and cost-effective malaria-detection devices.