The prevalence of malnutrition is estimated at 30-50% of hospitalized patients in China. Disease-related malnutrition increases the risk of infection, mortality, length of hospitalization as well as the economic burden. National Nutrition Plan proposed to reduce malnutrition, but a clear, effective roadmap and protocol has not existed yet. Several factors impede to resolve the above challenges. They include :1) the low efficiency of current malnutrition diagnosis methods; 2) the lack of dynamic, standard method that can evaluate nutritional status in quantitative way. To this end, the investigators aim to establish an artificial-intelligence malnutrition diagnosis system to improve the application of malnutrition Clinical Pathway. Firstly, the investigators will establish a multidimensional malnutrition large data set, based on our previously built national hospital nutrition screening data set.

It will contain deep 3D facial images, semi-structured and structured electronic medical record. Then, the investigators will use ensemble learning algorithm to establish a fully automatic, artificial-intelligence malnutrition diagnosis model that includes both etiological and phenotypic diagnosis.

Inclusion Criteria:

• Adults (≥18 years old);
• Within 48 hours of admission;
• Inpatients at high risk of malnutrition, such as malignant tumors, chronic obstructive pulmonary disease, etc;
• Han nationality;
• Able to given informed consent.

Exclusion Criteria:

• Patients with artificial facial changes (such as plastic surgery , head and neck radiotherapy , head and neck trauma);
• Diseases with special facial changes (such as acromegaly);
• High dose glucocorticoid users;
• Patients with facial edema;
• Emergency admission with an expected length of stay of less than 3 days;
• Other conditions researchers thought could not be included

Hospitalized patients in Peking Union Medical College