[ad_1]
IBM released research this week investigating how artificial intelligence and machine learning could be used to improve maternal health in developing countries and predict the onset and progression of type 1 diabetes. In a study funded by the Bill and Melinda Foundation Gates, the IBM researchers built models to analyze demographic data sets from African countries, finding “data-supported” links between the number of years between pregnancies and the size of a woman’s social network with birth outcomes. In a separate work, another IBM team analyzed data across three decades and four countries to try to anticipate the onset of type 1 diabetes 3 to 12 months before it is normally diagnosed and then predict its progression. They state that one of the models accurately predicted progression 84% of the time.
Improving neonatal outcome
Despite a global decline in infant mortality rates, many countries are not on track to achieve the proposed targets to end preventable deaths among newborns and children under 5 years of age. Not surprisingly, progress toward these goals remains uneven, reflected in disparities in access to health services and unequal allocation of resources.
Towards possible solutions, the IBM researchers attempted to identify characteristics associated with neonatal mortality “as captured in nationally representative cross-sectional data.” They analyzed corpora from two recent demographic and health surveys (from 2014 and 2018) taken in 10 different sub-Saharan countries, constructing for each survey a model to classify (1) the mothers who reported a birth in the 5 years prior to the survey, ( 2) those who reported losing one or more children younger than 28 days, and (3) those who did not report losing a child. The researchers then inspected each model by visualizing the characteristics in the data that informed the model’s conclusions, as well as how changes in the values of the characteristics might have impacted neonatal mortality.
The researchers concluded that in most countries (for example, Nigeria, Senegal, Tanzania, Zambia, South Africa, Kenya, Ghana, Ethiopia, the Democratic Republic of the Congo, and Burkina Faso), neonatal deaths account for most of the loss. of children under 5 years of age and that neonatal death rates have historically remained high despite a decline in deaths of children under 5 years of age. They found that the number of births in the last 5 years was positively correlated with neonatal mortality, while household size was negatively correlated with neonatal mortality. In addition, they claimed to have established that mothers living in smaller households have a higher risk of neonatal mortality compared to mothers living in larger households, with factors such as the age and sex of the head of household that appear to influence the association between household size and neonatal mortality.
Study co-authors point to limitations of their work, such as the fact that surveys, which are self-reported, can miss key information such as access to health care and health care seeking behaviors. They also admit that the models could be identifying and exploiting undesirable patterns to make their predictions. Still, they claim to have made an important contribution to the research community by showing that joint machine learning can potentially derive information on neonatal outcomes from health surveys alone.
“Our work demonstrates the practical application of machine learning to generate knowledge by inspecting black box models and the applicability of the use of machine learning techniques to generate novel knowledge and alternative hypotheses about the phenomena captured in health data at the population level. ”The researchers wrote. in an article describing their efforts. “The positive correlation between the reported number of births and neonatal mortality reflected in our results confirms the previously known observation about birth spacing as a key determinant of neonatal mortality.”
Prediction of type 1 diabetes
Another IBM team sought to investigate the extent to which AI could be helpful in diagnosing and treating type 1 diabetes, which affects about 1 in 100 adults during their lifetime. Based on research showing that clinical type 1 diabetes is generally preceded by a condition called islet autoimmunity, in which the body constantly produces antibodies called islet autoantibodies, the team developed an algorithm that groups patients and determine the number of groups and their profiles to discover commonalities between different geographic groups.
The algorithm considered profiles based on the types of autoantibodies, the age at which autoantibodies developed, and imbalances in autoantibody positivity. After pooling the autoantibody-positive subjects, the researchers applied the model to data from 1,507 patients in studies conducted in the US, Sweden, and Finland. The precision of group transfer was reported to be high, with a mean of 84% mentioned above, suggesting that the AAb profile can be used to predict the progression of type 1 diabetes regardless of the population.
In a related study, this same team of researchers created a type 1 diabetes ontology that captures the patterns of certain biomarkers and uses them in conjunction with a model to discern characteristics. The co-authors state that when applied to the same data sets as the clustering algorithm, the ontology improves prediction performance up to 12 months in advance, allowing them to predict which patients might develop type 1 diabetes one year before they occur. is routinely detected.
It is important to note, of course, that imbalances in the data sets could have skewed the predictions. A team of scientists from the United Kingdom found that almost all eye disease datasets come from patients in North America, Europe and China, which means that eye disease diagnostic algorithms are less likely to work well for racial groups. from underrepresented countries. In another study, Stanford University researchers stated that most of the US data for studies involving medical uses of AI comes from California, New York and Massachusetts.
Co-authors of an audit last month recommend that professionals apply “rigorous” fairness analysis prior to implementation as a solution to bias. We hope that IBM researchers, should they choose to eventually implement their models, will heed their advice.
