Search

The current study aimed to investigate the relationship between body parameters and the current–time product (mAs) in chest digital radiography using a non-contact infrared thickness-measurement sensor. An anthropomorphic chest phantom was first used to understand variations in mAs over multiple positionings during chest radiography when using the automatic exposure control (AEC) technique. In a human study, 929 consecutive male subjects who underwent regular chest examinations were enrolled, and their height (H), weight (W), and body mass index (BMI) were recorded. In addition, their chest thickness (T) was measured at exhalation using a non-contact infrared sensor, and chest radiography was then performed using the AEC technique. Finally, the relationship between four body parameters (T, BMI, T*BMI, and W/H) and mAs was investigated by fitting the body parameters to mAs using three curve models. The phantom study showed that the maximum mAs was 1.76 times higher than the lowest mAs during multiple positionings in chest radiography. In the human study, all chest radiographs passed the routine quality control procedure and had an exposure index between 100 and 212. In curve fitting, the comparisons showed that W/H had a closer relationship with mAs than the other body parameters, while the first-order power model with W/H fitted to mAs performed the best and had an R-square of 0.9971. We concluded that the relationship between W/H and mAs in the first-order power model may be helpful in predicting the optimal mAs and reducing the radiation dose for chest radiography when using the AEC technique.

Consumers’ purchase behavior increasingly relies on online reviews. Accordingly, there are more and more deceptive reviews which are harmful to customers. Existing methods to detect spam reviews mainly take the problem as a general text classification task, but they ignore the important features of spam reviews. In this paper, we propose a novel model, which splits a review into three parts: first sentence, middle context, and last sentence, based on the discovery that the first and last sentence express stronger emotion than the middle context. Then, the model uses four independent bidirectional long-short term memory (LSTM) models to encode the beginning, middle, end of a review and the whole review into four document representations. After that, the four representations are integrated into one document representation by a self-attention mechanism layer and an attention mechanism layer. Based on three domain datasets, the results of in-domain and mix-domain experiments show that our proposed method performs better than the compared methods.