使用形态学分析对严重营养不良进行诊断
对于营养不良的人群而言，尤其发达国家的肥胖人群，传统的营养不良筛查及诊断标准敏感性及特异度较低。通过电脑图像计算方法可以精确量化机体组织或器官的脂肪及瘦组织，这种方法被称为“形态学分析”。
美国密歇根大学为了验证此法对于严重营养不良的诊断价值，回顾性分析了2014～2016年415例急诊患者。所有人均通过CT进行形态学分析，内容包含身高、内脏脂肪面积、背部集群面积和皮下脂肪厚度，并有注册营养师在CT检查48小时之内评估患者的营养状况。
接受者操作特征（ROC）曲线下面积（AUC）提示“形态学分析”对于严重营养不良具有较高的诊断价值（AUC=0.78），优于临床诊断（0.74）及实验室检查（0.62）。
因此，不同程度的营养不良患者均出现皮下脂肪及内脏脂肪的丢失，然而只有严重营养不良的患者才会出现肌肉减少症。
JPEN J Parenter Enteral Nutr. 2017;41(2):276-277.
Diagnosis of severe malnutrition using analytic morphomics.
Joshua M. Glazer; Erica Raymond; Christopher Lee; Stewart C. Wang; Jill R. Cherry-Bukowiec.
University of Michigan Health System, Ann Arbor, Michigan, USA.
PURPOSE: A reliable, objective, and practical way to diagnose the presence and degree of malnutrition has the potential to significantly change individualized patient care plans and monitor effectiveness of our interventions. Traditional criteria used for screening and diagnosis of malnutrition are both insensitive and nonspecific and are further obfuscated by the prevalence of obesity in developed nations. Analytic morphomics uses computational image-processing algorithms to provide precise and detailed measurements of organs and body tissues and thus allows precise quantification of both lean soft tissue and fat. We sought to determine whether morphomic parameters can be used to identify patients with severe malnutrition.
METHODS: In a retrospective study, we performed analytic morphomics on data collected from 415 patients admitted from the emergency department at the University of Michigan from 2014-2016 who had a registered dietitian (RD) evaluation within 48 hours of a computed tomography (CT) scan for any reason. The general methodology of analytic morphomics has been previously described by our group. Expected age- and sex-related body composition changes were controlled for by normalizing to a publicly available reference analytic morphomics population (RAMP). The gold standard for assigning presence and degree of malnutrition was the diagnosis documented by the RD at the time of evaluation. We used logistic regression with elastic net regularization and cross-validation to develop predictive models for severe malnutrition on a 60% randomly selected training set. The remaining 40% of patients were prospectively held as an internal validation set on which to determine statistical performance measures of our predictive models.
RESULTS: During the study period, 21.7% of patients admitted from the emergency department were evaluated by an RD during their hospital encounter; of these, 25.2% (2878/11,407) had a qualifying CT scan obtained within 48 hours of this nutrition status assessment. A random sample of this population (n = 425) was used to developed 4 distinct models using the above-detailed regression. "Clinical" includes age, sex, race, and body mass index (BMI). "Lab" relies on serum albumin as a continuous variable. "Morphomics" consists of body depth, visceral fat area, dorsal muscle group area, and subcutaneous fat depth. A composite "Full" model integrates all clinical, lab, and morphomic data. In the validation set, patients with severe malnutrition were identified with area under the receiver operating characteristic curve (AUROC) values of 0.74, 0.62, 0.78, and 0.83, respectively. Loss of subcutaneous and visceral fat is seen across all severities of malnutrition while loss of lean muscle mass (sarcopenia) appears to be the defining characteristic of severe malnutrition.
CONCLUSIONS: We used analytic morphomics to demonstrate that severe malnutrition can be objectively quantified with medical imaging. By using preexisting and commonly obtained CT scans, no additional risk or cost is incurred to the patient or medical system. Our results suggest a role for analytic morphomics in automating expert consultation and expediting nutrition support in patients presenting with radiologic evidence of severe malnutrition. Furthermore, analytic morphomics may improve diagnostic accuracy, severity characterization, interrater reliability, and perhaps even inform future consensus definitions of malnutrition.
DOI: 10.1177/0148607116686023