Novel AI-Driven Infant Meningitis Screening from High Resolution Ultrasound Imaging

Abstract

AbstractBackgroundInfant meningitis can be a life-threatening disease and requires prompt and accurate diagnosis to prevent severe outcomes or death. Gold-standard diagnosis requires lumbar punctures (LP), to obtain and analyze cerebrospinal fluid (CSF). Despite being standard practice, LPs are invasive, pose risks for the patient and often yield negative results, either because of the contamination with red blood cells derived from the puncture itself, or due to the disease’s relatively low incidence due to the protocolized requirement to do LPs to discard a life-threatening infection in spite its relatively low incidence. Furthermore, in low-income settings, where the incidence is the highest, LPs and CSF exams are rarely feasible, and suspected meningitis cases are generally treated empirically. There’s a growing need for non-invasive, accurate diagnostic methods.MethodologyWe developed a three-stage deep learning framework using Neosonics®ultrasound technology for 30 infants with suspected meningitis and a permeable fontanelle, from three Spanish University Hospitals (2021-2023). In Stage 1, 2194 images were processed for quality control using avessel/non-vesselmodel, with a focus on vessel identification and manual removal of images exhibiting artifacts such as poor coupling and clutter. This refinement process led to a focused cohort comprising 16 patients—6 cases (336 images) and 10 controls (445 images), yielding 781 images for the second stage. The second stage involved the use of a deep learning model to classify images based on WBC count threshold (set at 30 cells/mm3) intocontrolormeningitiscategories. The third stage integrated eXplainable Artificial Intelligence (XAI) methods, such as GradCAM visualizations, alongside image statistical analysis, to provide transparency and interpretability of the model’s decision-making process in our AI-driven screening tool.ResultsOur approach achieved 96% accuracy in quality control, 93% precision and 92% accuracy in image-level meningitis detection, and 94% overall patient-level accuracy. It identified 6 meningitis cases and 10 controls with 100% sensitivity and 90% specificity, demonstrating only a single misclassification. The use of GradCAM-based explainable AI (XAI) significantly enhanced diagnostic interpretability, and to further refine our insights, we incorporated a statistics-based XAI approach. By analyzing image metrics like entropy and standard deviation, we identified texture variations in the images, attributable to the presence of cells, which improved the interpretability of our diagnostic tool.ConclusionThis study supports the efficacy of a multistage deep learning model for the non-invasive screening of infant meningitis and its potential to guide indications of LPs. It also highlights the transformative potential of AI in medical diagnostic screening for neonatal healthcare and paves the way for future research and innovations.