Sodium Intake Estimation in Hospital Patients Using AI-Based Imaging: Prospective Pilot Study.

BACKGROUND: Measurement of sodium intake in hospitalized patients is critical for their care. In this study, artificial intelligence (AI)-based imaging was performed to determine sodium intake in these patients. OBJECTIVE: The applicability of a diet management system was evaluated using AI-based imaging to assess the sodium content of diets prescribed for hospitalized patients. METHODS: Based on the information on the already investigated nutrients and quantity of food, consumed sodium was analyzed through photographs obtained before and after a meal. We used a hybrid model that first leveraged the capabilities of the You Only Look Once, version 4 (YOLOv4) architecture for the detection of food and dish areas in images. Following this initial detection, 2 distinct approaches were adopted for further classification: a custom ResNet-101 model and a hyperspectral imaging-based technique. These methodologies focused on accurate classification and estimation of the food quantity and sodium amount, respectively. The 24-hour urine sodium (UNa) value was measured as a reference for evaluating the sodium intake. RESULTS: Results were analyzed using complete data from 25 participants out of the total 54 enrolled individuals. The median sodium intake calculated by the AI algorithm (AI-Na) was determined to be 2022.7 mg per day/person (adjusted by administered fluids). A significant correlation was observed between AI-Na and 24-hour UNa, while there was a notable disparity between them. A regression analysis, considering patient characteristics (eg, gender, age, renal function, the use of diuretics, and administered fluids) yielded a formula accounting for the interaction between AI-Na and 24-hour UNa. Consequently, it was concluded that AI-Na holds clinical significance in estimating salt intake for hospitalized patients using images without the need for 24-hour UNa measurements. The degree of correlation between AI-Na and 24-hour UNa was found to vary depending on the use of diuretics. CONCLUSIONS: This study highlights the potential of AI-based imaging for determining sodium intake in hospitalized patients.

Endoscopic Image Classification Based on Explainable Deep Learning.

Deep learning has achieved remarkably positive results and impacts on medical diagnostics in recent years. Due to its use in several proposals, deep learning has reached sufficient accuracy to implement; however, the algorithms are black boxes that are hard to understand, and model decisions are often made without reason or explanation. To reduce this gap, explainable artificial intelligence (XAI) offers a huge opportunity to receive informed decision support from deep learning models and opens the black box of the method. We conducted an explainable deep learning method based on ResNet152 combined with Grad-CAM for endoscopy image classification. We used an open-source KVASIR dataset that consisted of a total of 8000 wireless capsule images. The heat map of the classification results and an efficient augmentation method achieved a high positive result with 98.28% training and 93.46% validation accuracy in terms of medical image classification.