A novel brain-inspired neural cognitive approach to SARS thermal image analysis

Thermal imaging is used extensively in the detection of infrared spectrum. This principle has found great and effective use in the screening of potential SARS patients. This paper investigates the application of several novel brain-inspired softcomputing techniques in the study of the correlation of superficial thermal images against the true internal body temperature. Given some backgrounds that the existing infrared systems used at various boarder checkpoints have high false-negative rate, the novel fuzzy neural networks (FNNs) employed in the back-end of the system have a role as a thermal analysis tool with high degree of accuracy. To achieve the automation and improve the accuracy in the feature extraction process of the infrared images, some forms of image processing technique based on the novel FNNs are also proposed in this paper. Extensive experimentations are undertaken to examine such intelligent medical decision support tool. Benchmarking was carried out on the novel FNN architectures which include pseudo outer-product fuzzy neural network (POPFNN), evolving fuzzy neural network (EFuNN), fuzzy adaptive learning control network (Falcon), generic self-organizing fuzzy neural network (GenSoFNN), and fuzzy cerebellar model articulation controller (FCMAC), and their results are promising. The performance of the GenSoFNN network is the most appealing among others. In addition, the experiments are conducted on real-life data taken from the Emergency Department (A&E), Tan Tock Seng Hospital (the designated SARS center in Singapore) to confirm the validity of such design in the real time.