Browsing by Author "Fedewa, Michael"

Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
  • Thumbnail Image
    Item
    Critical Analysis of the Reliability and Prediction Error of a Novel Smartphone Application for Measuring Body Composition
    (University of Alabama Libraries, 2023) Metoyer, Casey; Esco, Michael; Fedewa, Michael
    A smartphone application was recently developed that measures body composition from a 2-dimensional image of a standing person (%FatIMAGE). Three experiments were performed to 1) determine the reliability of %FatIMAGE in two different body positions and days, 2) determine the reliability of %FatIMAGE across different cameras, room lighting levels, and colored backgrounds, and 3) assess the potential sources of error related to %Fat estimation with bioimpedance, skinfold, and IMAGE. In the first study, %FatIMAGE was measured in the anterior (ANTDay1) and posterior (POSTDay1) positions, then again within 48 hours of the first visit (POSTDay2). The ANTDay1 was significantly lower compared to the reference, whereas the POSTDay2 was not. In the second study, %FatIMAGE was estimated in different megapixel cameras (12MP, 8MP, 5MP, and 0.7MP), different room lighting conditions (i.e., low=LL, medium=ML, ambient=AL, and bright light=BL), and different colored backgrounds (White=WB, Black=BB, Green=GnB, Orange=OB, and Gray=GyB). Conditions were compared to a reference (8MP, AL, and WB). The 0.7MP was significantly higher than the reference, whereas the 12MP and 5MP were not. The LL was significantly higher than the reference, whereas the BL and ML were not. The BB, GnB, and GyB were significantly higher than the WB, whereas the OB was not. In a third study, %Fat was measured via bioimpedance, skinfold, %FatIMAGE, and a criterion 4-compartment model. The constant error (CE) was calculated (%Fat condition - criterion) for each method (SKFERROR, BIAERROR., and IMAGEERROR). The CE was correlated with markers of physical activity (PA) and muscular fitness (MF): the International physical activity questionnaire (IPAQT, IPAQV, IPAQM); Perceived functional ability questionnaire (1mileSUB and 3mileSUB); Physical Activity Rating questionnaire (PA-R): handgrip test (HG); and push-up test (PU). SKFERROR displayed a small correlation with PU, while IPAQV, 1mileSUB, 3mileSUB, PA-R, and PU were associated with IMAGEERROR. Regression analyses indicated that only PU contributed to the model for SKFERROR and IMAGEERROR. In conclusion, the IMAGE produced acceptable reliability across different camera, lighting, and background conditions but attention should be paid to clothing color in contrast with background color. In addition, further evaluation of MF measures is recommended, as higher PU were associated with greater IMAGEERROR.