Regresar

Artificial neural networks and digital image processing: An approach for indirect weight measurement

Abstract:

A person's weight is an important variable in determining the correct dosage of medication and anesthetics. However, weight is difficult to measure when people are in a medical emergency; thus, in such situations, weight is only subjectively estimated. We propose an image processing technique for the indirect measurement of the weight of a person in such scenarios. This algorithm computes the area that the subject fills in a normalized image; and a non-linear model relates this value to weight. The perspective of the camera and the noise in the image are also corrected. For the model, both Least Mean Square (LMS) and Artificial Neural Network (ANN) fitting techniques were explored. A mean relative error of 5.8% was obtained for the ANN method under k-Fold cross validation. Lack of CPU needs allows for the proposed algorithm to be implemented on smartphones, thereby making this technology accessible.