Wenyan Jia1, Yuecheng Li1, Ruowei Qu1, Thomas Baranowski2, Lora E Burke3, Hong Zhang4, Yicheng Bai1, Juliet M Mancino3, Guizhi Xu5, Zhi-Hong Mao6, Mingui Sun1. 1. 1Department of Neurosurgery,University of Pittsburgh,3520 Forbes Avenue,Suite 202,Pittsburgh,PA 15213,USA. 2. 3Department of Pediatrics,Baylor College of Medicine,Houston,TX,USA. 3. 4School of Nursing,University of Pittsburgh,Pittsburgh,PA,USA. 4. 5Image Processing Center,Beihang University,Beijing,People's Republic of China. 5. 2Department of Biomedical Engineering,Hebei University of Technology,Tianjin,People's Republic of China. 6. 6Department of Electrical and Computer Engineering,University of Pittsburgh,Pittsburgh,PA,USA.
Abstract
OBJECTIVE: To develop an artificial intelligence (AI)-based algorithm which can automatically detect food items from images acquired by an egocentric wearable camera for dietary assessment. DESIGN: To study human diet and lifestyle, large sets of egocentric images were acquired using a wearable device, called eButton, from free-living individuals. Three thousand nine hundred images containing real-world activities, which formed eButton data set 1, were manually selected from thirty subjects. eButton data set 2 contained 29 515 images acquired from a research participant in a week-long unrestricted recording. They included both food- and non-food-related real-life activities, such as dining at both home and restaurants, cooking, shopping, gardening, housekeeping chores, taking classes, gym exercise, etc. All images in these data sets were classified as food/non-food images based on their tags generated by a convolutional neural network. RESULTS: A cross data-set test was conducted on eButton data set 1. The overall accuracy of food detection was 91·5 and 86·4 %, respectively, when one-half of data set 1 was used for training and the other half for testing. For eButton data set 2, 74·0 % sensitivity and 87·0 % specificity were obtained if both 'food' and 'drink' were considered as food images. Alternatively, if only 'food' items were considered, the sensitivity and specificity reached 85·0 and 85·8 %, respectively. CONCLUSIONS: The AI technology can automatically detect foods from low-quality, wearable camera-acquired real-world egocentric images with reasonable accuracy, reducing both the burden of data processing and privacy concerns.
OBJECTIVE: To develop an artificial intelligence (AI)-based algorithm which can automatically detect food items from images acquired by an egocentric wearable camera for dietary assessment. DESIGN: To study human diet and lifestyle, large sets of egocentric images were acquired using a wearable device, called eButton, from free-living individuals. Three thousand nine hundred images containing real-world activities, which formed eButton data set 1, were manually selected from thirty subjects. eButton data set 2 contained 29 515 images acquired from a research participant in a week-long unrestricted recording. They included both food- and non-food-related real-life activities, such as dining at both home and restaurants, cooking, shopping, gardening, housekeeping chores, taking classes, gym exercise, etc. All images in these data sets were classified as food/non-food images based on their tags generated by a convolutional neural network. RESULTS: A cross data-set test was conducted on eButton data set 1. The overall accuracy of food detection was 91·5 and 86·4 %, respectively, when one-half of data set 1 was used for training and the other half for testing. For eButton data set 2, 74·0 % sensitivity and 87·0 % specificity were obtained if both 'food' and 'drink' were considered as food images. Alternatively, if only 'food' items were considered, the sensitivity and specificity reached 85·0 and 85·8 %, respectively. CONCLUSIONS: The AI technology can automatically detect foods from low-quality, wearable camera-acquired real-world egocentric images with reasonable accuracy, reducing both the burden of data processing and privacy concerns.
Authors: Aiden R Doherty; Steve E Hodges; Abby C King; Alan F Smeaton; Emma Berry; Chris J A Moulin; Siân Lindley; Paul Kelly; Charlie Foster Journal: Am J Prev Med Date: 2013-03 Impact factor: 5.043
Authors: C Fukuike; N Kodama; Y Manda; Y Hashimoto; K Sugimoto; A Hirata; Q Pan; N Maeda; S Minagi Journal: J Oral Rehabil Date: 2014-12-26 Impact factor: 3.837
Authors: Brooke M Bell; Ridwan Alam; Nabil Alshurafa; Edison Thomaz; Abu S Mondol; Kayla de la Haye; John A Stankovic; John Lach; Donna Spruijt-Metz Journal: NPJ Digit Med Date: 2020-03-13
Authors: Donna Spruijt-Metz; Cheng K Fred Wen; Brooke M Bell; Stephen Intille; Jeannie S Huang; Tom Baranowski Journal: Am J Prev Med Date: 2018-08-19 Impact factor: 5.043
Authors: Masudul H Imtiaz; Delwar Hossain; Volkan Y Senyurek; Prajakta Belsare; Stephen Tiffany; Edward Sazonov Journal: Nicotine Tob Res Date: 2020-10-08 Impact factor: 4.244
Authors: Sai Krupa Das; Akari J Miki; Caroline M Blanchard; Edward Sazonov; Cheryl H Gilhooly; Sujit Dey; Colton B Wolk; Chor San H Khoo; James O Hill; Robin P Shook Journal: Adv Nutr Date: 2022-02-01 Impact factor: 11.567
Authors: Modou L Jobarteh; Megan A McCrory; Benny Lo; Mingui Sun; Edward Sazonov; Alex K Anderson; Wenyan Jia; Kathryn Maitland; Jianing Qiu; Matilda Steiner-Asiedu; Janine A Higgins; Tom Baranowski; Peter Olupot-Olupot; Gary Frost Journal: Curr Dev Nutr Date: 2020-02-07