RESUMEN
Supervised machine learning classification is the most common example of artificial intelligence (AI) in industry and in academic research. These technologies predict whether a series of measurements belong to one of multiple groups of examples on which the machine was previously trained. Prior to real-world deployment, all implementations need to be carefully evaluated with hold-out validation, where the algorithm is tested on different samples than it was provided for training, in order to ensure the generalizability and reliability of AI models. However, established methods for performing hold-out validation do not assess the consistency of the mistakes that the AI model makes during hold-out validation. Here, we show that in addition to standard methods, an enhanced technique for performing hold-out validation-that also assesses the consistency of the sample-wise mistakes made by the learning algorithm-can assist in the evaluation and design of reliable and predictable AI models. The technique can be applied to the validation of any supervised learning classification application, and we demonstrate the use of the technique on a variety of example biomedical diagnostic applications, which help illustrate the importance of producing reliable AI models. The validation software created is made publicly available, assisting anyone developing AI models for any supervised classification application in the creation of more reliable and predictable technologies.
RESUMEN
Autism is a group of complex neurodevelopmental disorders characterized by impaired social interaction and restricted/repetitive behavior. We performed a large-scale retrospective analysis of 1,996 clinical neurological structural magnetic resonance imaging (MRI) examinations of 781 autistic and 988 control subjects (aged 0-32 years), and extracted regionally distributed cortical thickness measurements, including average measurements as well as standard deviations which supports the assessment of intra-regional cortical thickness variability. The youngest autistic participants (<2.5 years) were diagnosed after imaging and were identified retrospectively. The largest effect sizes and the most common findings not previously published in the scientific literature involve abnormal intra-regional variability in cortical thickness affecting many (but not all) regions of the autistic brain, suggesting irregular gray matter development in autism that can be detected with MRI. Atypical developmental patterns have been detected as early as 0 years old in individuals who would later be diagnosed with autism.
RESUMEN
Autism is a group of complex neurodevelopmental disorders characterized by impaired social interaction, restricted and repetitive behavior. We performed a large-scale retrospective analysis of 1,996 structural magnetic resonance imaging (MRI) examinations of the brain from 1,769 autistic and neurologically typically developing patients (aged 0-32 years), and extracted regional volumetric measurements distributed across 463 brain regions of each patient. The youngest autistic patients (<2.5 years) were diagnosed after imaging and identified retrospectively. Our study demonstrates corpus callosum volumetric abnormalities among autistic patients that are associated with brain overgrowth in early childhood (0-5 years old), followed by a shift towards known decreased volumes in later ages. Results confirm known increases in ventricular volumes among autistic populations and extends those findings to increased volumes of the choroid plexus. Our study also demonstrates distributed volumetric abnormalities among autistic patients that affect a variety of key regional white and grey matter areas of the brain potentially associated with known symptoms of autism.
