Collaborating Authors


Machine Learning for Febrile Infants – A New Paradigm?


Perhaps many people are like me in that hearing the word "machine learning" for the first time brings forth images of Skynet from The Terminator movies or Haley Joel Osment's character from the Steven Spielberg's film A.I. Artificial Intelligence. However, machine learning has now become a regular part of our vernacular when it comes to predictive modeling in many conditions. Ramgopal et al use machine learning methods to derive and validate a new prediction model for risk stratification of febrile infants 60 days of age. Using various machine learning approaches, the authors developed a prediction model with high sensitivity and specificity compared with recent prediction models for febrile infants. So, are machine learning models the new paradigm for risk stratification of febrile infants? The results are intriguing, particularly the high specificity of the model, but further work must be done, as explained nicely by Chamberlain et al in an accompanying commentary (10.1542/peds.2020-012203).

Rethinking Greulich and Pyle: A Deep Learning Approach to Pediatric Bone Age Assessment Using Pediatric Trauma Hand Radiographs


To develop a deep learning approach to bone age assessment based on a training set of developmentally normal pediatric hand radiographs and to compare this approach with automated and manual bone age assessment methods based on Greulich and Pyle (GP). In this retrospective study, a convolutional neural network (trauma hand radiograph–trained deep learning bone age assessment method [TDL-BAAM]) was trained on 15 129 frontal view pediatric trauma hand radiographs obtained between December 14, 2009, and May 31, 2017, from Children's Hospital of New York, to predict chronological age. A total of 214 trauma hand radiographs from Hasbro Children's Hospital were used as an independent test set. The test set was rated by the TDL-BAAM model as well as a GP-based deep learning model (GPDL-BAAM) and two pediatric radiologists (radiologists 1 and 2) using the GP method. All ratings were compared with chronological age using mean absolute error (MAE), and standard concordance analyses were performed.

AI tool detects child abuse images with 99% accuracy


CSAM Image Classifier: Machine learning classification model developed by Thorn and leveraged within Safer that returns a prediction for whether a …

AI tool detects child abuse images with 99% accuracy


A new AI-powered tool claims to detect child abuse images with around 99 percent accuracy. The tool, called Safer, is developed by non-profit Thorn to assist businesses which do not have in-house filtering systems to detect and remove such images. According to the Internet Watch Foundation in the UK, reports of child abuse images surged 50 percent during the COVID-19 lockdown. In the 11 weeks starting on 23rd March, its hotline logged 44,809 reports of images compared with 29,698 last year. Many of these images are from children who've spent more time online and been coerced into releasing images of themselves.

Kids Are Especially Tough to Interview About Abuse. Are Robots the Solution?

Mother Jones

Cindy Bethel was 6 when her babysitter's neighbor started molesting her. Worried what else would happen if she told her parents, she confided in her stuffed panda instead. Sometimes she acted out the abuse with Barbie and Ken dolls. A few years later, the same teen neighbor raped her on a woodpile outside his house. She didn't tell anyone about the assault until long after she moved away from her Ohio hometown.

Zoom calls are being interrupted to broadcast child abuse images, users say

The Independent - Tech

Zoom calls are being interrupted by attackers who broadcast child abuse imagery, users say. The reports are just the latest example of the phenomenon of "Zoombombing", where strangers break into calls and show often distressing images and videos. The National Crime Agency (NCA) has now confirmed that it is investigating the reports and following up with other instances of such attacks. The BBC reported that several users of video calling app Zoom had recently experienced incidents where their meeting had been interrupted by abuse footage. It said one of the meetings in question had been publicised on social media - something a number of online safety groups and Zoom itself urge users not to do.

Artificial Intelligence to Define Very Young Brains


"This is one of the first times that artificial intelligence has been used to better define the different parts of a newborn's brain on an MRI: namely the grey matter, white matter and cerebrospinal fluid," said Dr. Gregory A. Lodygensky, a neonatologist at CHU Sainte-Justine and professor at Universit-- de Montr--al. 'The new technique that uses artificial intelligence allows babies' brains to be examined quickly, accurately and reliably. Scientists see it as a major asset for supporting research that not only addresses brain development in neonatal care, but also the effectiveness of neuroprotective strategies.' "Until today, the tools available were complex, often intermingled and difficult to access," he added. In collaboration with Professor Jose Dolz, an expert in medical image analysis and machine learning at --TS, the researchers were able to adapt the tools to the specificities of the neonatal setting and then validate them. In evaluating a range of tools available in artificial intelligence, CHU Sainte-Justine researchers found that these tools had limitations, particularly with respect to pediatric research.

Causal datasheet: An approximate guide to practically assess Bayesian networks in the real world Artificial Intelligence

In solving real-world problems like changing healthcare-seeking behaviors, designing interventions to improve downstream outcomes requires an understanding of the causal links within the system. Causal Bayesian Networks (BN) have been proposed as one such powerful method. In real-world applications, however, confidence in the results of BNs are often moderate at best. This is due in part to the inability to validate against some ground truth, as the DAG is not available. This is especially problematic if the learned DAG conflicts with pre-existing domain doctrine. At the policy level, one must justify insights generated by such analysis, preferably accompanying them with uncertainty estimation. Here we propose a causal extension to the datasheet concept proposed by Gebru et al (2018) to include approximate BN performance expectations for any given dataset. To generate the results for a prototype Causal Datasheet, we constructed over 30,000 synthetic datasets with properties mirroring characteristics of real data. We then recorded the results given by state-of-the-art structure learning algorithms. These results were used to populate the Causal Datasheet, and recommendations were automatically generated dependent on expected performance. As a proof of concept, we used our Causal Datasheet Generation Tool (CDG-T) to assign expected performance expectations to a maternal health survey we conducted in Uttar Pradesh, India.

A survey of statistical learning techniques as applied to inexpensive pediatric Obstructive Sleep Apnea data Machine Learning

Obstructive sleep apnea (OSA), a form of sleep-disordered breathing characterized by recurrent episodes of partial or complete airway obstruction during sleep, is a serious health problem, affecting an estimated 1-5% of elementary school-aged children [9, 2]. Even mild forms of untreated pediatric OSA may cause high blood pressure, behavioral challenges, or impeded growth. Compared to adults, the symptoms of childhood-onset OSA are more varied and change continuously with development, making diagnosis a difficult challenge. The complexity of the data from surveys, biomedical measurements, 3D facial photos, and time-series data calls for state of the art techniques from mathematics and data science. Clinical data, including that considered in confirming or ruling out a diagnosis of pediatric OSA, consist of high-dimensional multi-mode data with mixtures of variables of disparate types (e.g., nominal and categorical data of different scales, interval data, time-to-event and longitudinal outcomes) also called mixed or noncommensurate data.

Neural Proximal Gradient Descent for Compressive Imaging

Neural Information Processing Systems

Recovering high-resolution images from limited sensory data typically leads to a serious ill-posed inverse problem, demanding inversion algorithms that effectively capture the prior information. Learning a good inverse mapping from training data faces severe challenges, including: (i) scarcity of training data; (ii) need for plausible reconstructions that are physically feasible; (iii) need for fast reconstruction, especially in real-time applications. We develop a successful system solving all these challenges, using as basic architecture the repetitive application of alternating proximal and data fidelity constraints. We learn a proximal map that works well with real images based on residual networks with recurrent blocks. Extensive experiments are carried out under different settings: (a) reconstructing abdominal MRI of pediatric patients from highly undersampled k-space data and (b) super-resolving natural face images.