1. Information about the Paper

Citation

Dong, Yujie, Adam Hoover, Jenna Scisco, and Eric Muth. “A new method for measuring meal intake in humans via automated wrist motion tracking.” Applied psychophysiology and biofeedback 37, no. 3 (2012): 205-215.

Abstract

Measuring the energy intake (kcal) of a person in day-to-day life is difficult. The best laboratory tool achieves 95 % accuracy on average, while tools used in daily living typically achieve 60–80 % accuracy. This paper describes a new method for measuring intake via automated tracking of wrist motion. Our method uses a watch-like device with a micro-electro-mechanical gyroscope to detect and record when an individual has taken a bite of food. Two tests of the accuracy of our device in counting bites found that our method has 94 % sensitivity in a controlled meal setting and 86 % sensitivity in an uncontrolled meal setting, with one false positive per every 5 bites in both settings. Preliminary data from daily living indicates that bites measured by the device are positively related to caloric intake illustrating the potential of the device to monitor energy intake. Future research should seek to further explore the relationship between bites taken and kilocalories consumed to validate the device as an automated measure of energy intake.

paper-review-4

2. My Review of the Paper

Summary:

The motivation behind this paper is to help overcome overweight and obesity by suppressing calory intake. Previous works are mostly manual ways. This paper proposed a better semi automatic way to calculate calory intake by using wrist rotation as a method to detect bite count. The main idea is that when eating, we rotate our wrist to face our mouth. Thus provides a pattern to be detected as bite count.

The goals of this paper are:

  1. To check whether the method is applicable to significant number of subjects
  2. To check if the method can be applied to variety of foods
  3. To check if there is any correlation between kilocalories and bite counts

This paper used three step of experiments:

  1. To measure ideal accuracy of the method in ideal situation, which is in a laboratory and all subjects were eating the same food. The accuracy was 94%.
  2. The subject can bring their own food. The experiment is still done in the laboratory. The accuracy was dropped to 86%.
  3. The experiment was done outside of the laboratory.

The result of correlation was moderate linear R=0.6 which most likely was caused by the different calory density of each food consumed by the subject.

Comments

Pro:

  1. This paper offers new method to automatically calculate meal intake by using wrist rotation to detect bite count.
  2. Offers a better way where previous works were more costly, required manual effort or involved inconvenience in practice.
  3. It has been cited by seventeen (17) following works, three (3) of them are done by the same group of this paper authors. It means that this paper has quite significant foundation to inspire following works. And the authors seem to continuously making it better. Three references of previous works are also by the same group of authors. 

Cons:

  1. Some limitations are already stated in the paper:
  2. Having the subject press a button to start eating detection
  3. Variation of calory density between different foods making it difficult to correctly compute how many calories intake per bite count
  4. Potensial effect of false positives due to several factors such as
  5. Using the other hand (which is not wearing the sensor) to eat
  6. Limitation to be addressed in future works are already stated in the paper:
  7. develop new algorithm to distinguish eating activity from other activities to enable detecting bite count without the need of pressing a button
  8. Offers a method to provide real-time feedback to stop eating after some bite count threshold has been achieved
  9. Actual calories count still need further laboratory calculation
  10. Need bigger battery and storage to allow storing raw sensor data to be calculated by a machine learning algorithm 

Suggestions:

  1. I suggest to count bite rate by dividing number of bites per time (per minute). Because the goal is to minimize calories intake. In addition, some research in medical field found that those who eat faster will consume more calories. However, this is only valid for men, not for women. Therefore, we need to also include gender/sex variable.
  2. My references:
  3. Eating slowly led to decreases in energy intake within meals in healthy women. http://www.ncbi.nlm.nih.gov/pubmed/18589027
  4. Slower eating rate reduces the food intake of men, but not women: implications for behavioral weight control. http://www.ncbi.nlm.nih.gov/pubmed/17517367