A new take on measuring relative nutritional density: The feasibility of using a deep neural network to assess commercially-prepared pureed food concentrations

摘要

Dysphagia affects 590 million people worldwide and increases risk formalnutrition. Pureed food may reduce choking, however preparation differencesimpact nutrient density making quality assurance necessary. This paper is thefirst study to investigate the feasibility of computational pureed foodnutritional density analysis using an imaging system. Motivated by atheoretical optical dilution model, a novel deep neural network (DNN) wasevaluated using 390 samples from thirteen types of commercially prepared pureesat five dilutions. The DNN predicted relative concentration of the puree sample(20%, 40%, 60%, 80%, 100% initial concentration). Data were captured usingsame-side reflectance of multispectral imaging data at different polarizationsat three exposures. Experimental results yielded an average top-1 predictionaccuracy of 92.2+/-0.41% with sensitivity and specificity of 83.0+/-15.0% and95.0+/-4.8%, respectively. This DNN imaging system for nutrient densityanalysis of pureed food shows promise as a novel tool for nutrient qualityassurance.