Artificial intelligence to classify acquired intestinal injury in preterm neonates—a new perspective

Alain Cuna,Muralidhar H. Premkumar,Venkatesh Sampath
DOI: https://doi.org/10.1038/s41390-024-03148-w
IF: 3.953
2024-03-19
Pediatric Research
Abstract:The last two decades have seen striking advances in our understanding of necrotizing enterocolitis (NEC) including the role of innate immune signaling in NEC and distinguishing NEC from other intestinal injury phenotypes such as spontaneous intestinal perforation (SIP). 1,2,3 Nevertheless, it is recognized that NEC is a potpourri of several intestinal injury phenotypes that affect preterm and term infants varying with respect to pathophysiology, timing of onset and prognosis. 1,4,5 The resulting imprecision in diagnosis of intestinal injury phenotypes confounds targeted prevention and treatment strategies. This primarily arises from ambiguity and subjectivity in interpreting clinical and radiological signs that are observer-dependent. To address this limitation, newer definitions of NEC have been proposed that aim to distinguish intestinal injury phenotypes from potentially benign conditions such as feed intolerance. 1,6 The increasing awareness that computers (machines) can incorporate diverse data streams such as clinical, epidemiological, and radiological signs has resulted in an explosion of artificial intelligence-based approaches to neonatal diseases confounded by phenotypic heterogeneity and imprecision in diagnosis. Artificial intelligence refers to the capacity of a machine or computer to make interpretations and decisions using mathematical and logical processes akin to the human brain. Within artificial intelligence, machine learning is a subcategory where algorithms enable machines to learn and perform tasks using training data without explicit instructions. 7,8 Most current machine learning applications concerning NEC and SIP have predominantly utilized supervised learning. In supervised learning, algorithms are trained on labeled data, with NEC or SIP outcomes predetermined. 9,10,11,12,13 Conversely, only a few studies have employed either semi-supervised or unsupervised learning approaches, where the dataset may contain limited or no labeled data or predefined outcomes. 14,15,16
pediatrics
What problem does this paper attempt to address?