Hyperspectral signature-band extraction and learning: an example of sugar content prediction of Syzygium samarangense.

This study proposes a method to extract the signature bands from the deep learning models of multispectral data converted from the hyperspectral data. The signature bands with two deep-learning models were further used to predict the sugar content of the Syzygium samarangense. Firstly, the hyperspectral data with the bandwidths lower than 2.5 nm were converted to the spectral data with multiple bandwidths higher than 2.5 nm to simulate the multispectral data. The convolution neural network (CNN) and the feedforward neural network (FNN) used these spectral data to predict the sugar content of the Syzygium samarangense and obtained the lowest mean absolute error (MAE) of 0.400° Brix and 0.408° Brix, respectively. Secondly, the absolute mean of the integrated gradient method was used to extract multiple signature bands from the CNN and FNN models for sugariness prediction. A total of thirty sets of six signature bands were selected from the CNN and FNN models, which were trained by using the spectral data with five bandwidths in the visible (VIS), visible to near-infrared (VISNIR), and visible to short-waved infrared (VISWIR) wavelengths ranging from 400 to 700 nm, 400 to 1000 nm, and 400 to 1700 nm. Lastly, these signature-band data were used to train the CNN and FNN models for sugar content prediction. The FNN model using VISWIR signature bands with a bandwidth of ± 12.5 nm had a minimum MAE of 0.390°Brix compared to the others. The CNN model using VISWIR signature bands with a bandwidth of ± 10 nm had the lowest MAE of 0.549° Brix compared to the other CNN models. The MAEs of the models with only six spectral bands were even better than those with tens or hundreds of spectral bands. These results reveal that six signature bands have the potential to be used in a small and compact multispectral device to predict the sugar content of the Syzygium samarangense.

Sugariness prediction of Syzygium samarangense using convolutional learning of hyperspectral images.

Sugariness is one of the most important indicators to measure the quality of Syzygium samarangense, which is also known as the wax apple. In general, farmers used to measure sugariness by testing the extracted juice of the wax apple products. Such a destructive way to measure sugariness is not only labor-consuming but also wasting products. Therefore, non-destructive and quick techniques for measuring sugariness would be significant for wax apple supply chains. Traditionally, the non-destructive method to predict the sugariness or the other indicators of the fruits was based on the reflectance spectra or Hyperspectral Images (HSIs) using linear regression such as Multi-Linear Regression (MLR), Principal Component Regression (PCR), and Partial Least Square Regression (PLSR), etc. However, these regression methods are usually too simple to precisely estimate the complicated mapping between the reflectance spectra or HSIs and the sugariness. This study presents the deep learning methods for sugariness prediction using the reflectance spectra or HSIs from the bottom of the wax apple. A non-destructive imaging system fabricated with two spectrum sensors and light sources is implemented to acquire the visible and infrared lights with a range of wavelengths. In particular, a specialized Convolutional Neural Network (CNN) with hyperspectral imaging is proposed by investigating the effect of different wavelength bands for sugariness prediction. Rather than extracting spatial features, the proposed CNN model was designed to extract spectral features of HSIs. In the experiments, the ground-truth value of sugariness is obtained from a commercial refractometer. The experimental results show that using the whole band range between 400 and 1700 nm achieves the best performance in terms of °Brix error. CNN models attain the °Brix error of ± 0.552, smaller than ± 0.597 using Feedforward Neural Network (FNN). Significantly, the CNN's test results show that the minor error in the interval 0 to 10°Brix and 10 to 11°Brix are ± 0.551 and ± 0.408, these results indicate that the model would have the capability to predict if sugariness is below 10°Brix or not, which would be similar to the human tongue. These results are much better than ± 1.441 and ± 1.379 by using PCR and PLSR, respectively. Moreover, this study provides the test error in each °Brix interval within one Brix, and the results show that the test error is varied considerably within different °Brix intervals, especially on PCR and PLSR. On the other hand, FNN and CNN obtain robust results in terms of test error.

Autofluorescence Detection Method for Dental Plaque Bacteria Detection and Classification: Example of Porphyromonas gingivalis, Aggregatibacter actinomycetemcomitans, and Streptococcus mutans.

The use of fluorescence spectroscopy for plaque detection is a fast and effective way to monitor oral health. At present, there is no uniform specification for the design of the excitation light source of related products for generating fluorescence. To carry out experiments on dental plaque, the fluorescence spectra of three different bacterial species (Porphyromonas gingivalis, Aggregatibacter actinomycetemcomitans, and Streptococcus mutans) were measured by hyperspectral imaging microscopy (HIM). Three critical issues were found in the experiments. One issue was the unwanted spectrum generated from a mercury line source; two four-order low-pass filters were evaluated for eliminating the unwanted spectrum and meet the experimental requirements. The second issue was the red fluorescence generated from the microscope slide made of borosilicate glass; this could affect the observation of the red fluorescence from the bacteria; quartz microscope slides were found to reduce the fluorescence intensity by about 2 dB compared with the borosilicate slide. The third issue of photobleaching in the fluorescence of the Porphyromonas gingivalis was studied. This study proposes a method of classifying three bacteria based on the spectral intensity ratios (510/635 and 500/635 nm) under the 405 nm excitation light was proposed in this study. The sensitivity and specificity of the classification were approximately 99% and 99%, respectively.

Using System Identification to Construct an Inherent Model of Pupillary Light Reflex to Explore Diabetic Neuropathy.

This study proposed a pupillary light reflex (PLR) inherent model based on the system identification method to demonstrate the dynamic physiological mechanism of the PLR, in which pupillary constriction and dilation are controlled by the sympathetic and parasympathetic nervous system. This model was constructed and verified by comparing the simulated and predicted PLR response with that of healthy participants. The least root-mean-square error (RMSE) of simulated PLR response was less than 0.7% when stimulus duration was under 3 ms. The RMSE of predicted PLR response increased by approximately 6.76%/s from the stimulus duration of 1 ms to 3 s, when the model directly used the parameters extracted from the PLR at the stimulus duration of 10 ms. When model parameters were derived from the regression by the measured PLR response, the RMSE kept under 8.5%. The model was applied to explore the PLR abnormalities of the people with Diabetic Mellitus (DM) by extracting the model parameters from 42 people with DM and comparing these parameters with those of 42 healthy participants. The parameter in the first-order term of the elastic force of the participants with DM was significantly lower than that of the healthy participants (p < 0.05). The sympathetic force and sympathetic action delay of the participants with DM were significantly larger (p < 0.05) and longer (p < 0.0001) than that of the healthy ones, respectively. The reason might be that the sympathetic nervous system, which controls the dilator muscle, degenerated in diabetic patients.

Band-Selection of a Portal LED-Induced Autofluorescence Multispectral Imager to Improve Oral Cancer Detection.

This aim of this study was to find effective spectral bands for the early detection of oral cancer. The spectral images in different bands were acquired using a self-made portable light-emitting diode (LED)-induced autofluorescence multispectral imager equipped with 365 and 405 nm excitation LEDs, emission filters with center wavelengths of 470, 505, 525, 532, 550, 595, 632, 635, and 695 nm, and a color image sensor. The spectral images of 218 healthy points in 62 healthy participants and 218 tumor points in 62 patients were collected in the ex vivo trials at China Medical University Hospital. These ex vivo trials were similar to in vivo because the spectral images of anatomical specimens were immediately acquired after the on-site tumor resection. The spectral images associated with red, blue, and green filters correlated with and without nine emission filters were quantized by four computing method, including summated intensity, the highest number of the intensity level, entropy, and fractional dimension. The combination of four computing methods, two excitation light sources with two intensities, and 30 spectral bands in three experiments formed 264 classifiers. The quantized data in each classifier was divided into two groups: one was the training group optimizing the threshold of the quantized data, and the other was validating group tested under this optimized threshold. The sensitivity, specificity, and accuracy of each classifier were derived from these tests. To identify the influential spectral bands based on the area under the region and the testing results, a single-layer network learning process was used. This was compared to conventional rules-based approaches to show its superior and faster performance. Consequently, four emission filters with the center wavelengths of 470, 505, 532, and 550 nm were selected by an AI-based method and verified using a rule-based approach. The sensitivities of six classifiers using these emission filters were more significant than 90%. The average sensitivity of these was about 96.15%, the average specificity was approximately 69.55%, and the average accuracy was about 82.85%.

Adjustable panoramic inspection system for submillimeter fasteners.

Fasteners are critical and indispensable locking components in mechanical assembly. Submillimeter fasteners are massively and widely used in electronic devices. This study proposed an adjustable panoramic inspection system for M2 to M0.8 submillimeter fasteners. The system mainly consists of a panoramic imaging module, a back-light module, and an image grabbing and computing module. The panoramic imaging module would form four equal optical path lengths to keep the same imaging amplification between the different directions of the field of view. The back-light module was designed to provide uniform illumination and enhance the contrast of the pitch edge between the fasteners and the background. The image grabbing and computing module with a high-speed camera was designed to be adjustable for different sizes of submillimeter fasteners. The realized system can offer the function of four images in one shot to make a panoramic scene, independent illumination for recognizing, inspect screws from M0.8 to M2.0 screws, and short time consumption of image processing, such as 3.284 ms for M0.8 screws and 2.384 ms for M2.0 screws, to achieve examination of 6000 pieces in 1 min.

Portable LED-induced autofluorescence spectroscopy for oral cancer diagnosis.

Oral cancer is a serious and growing problem in many developing and developed countries. To improve the cancer screening procedure, we developed a portable light-emitting-diode (LED)-induced autofluorescence (LIAF) imager that contains two wavelength LED excitation light sources and multiple filters to capture ex vivo oral tissue autofluorescence images. Compared with conventional means of oral cancer diagnosis, the LIAF imager is a handier, faster, and more highly reliable solution. The compact design with a tiny probe allows clinicians to easily observe autofluorescence images of hidden areas located in concave deep oral cavities. The ex vivo trials conducted in Taiwan present the design and prototype of the portable LIAF imager used for analyzing 31 patients with 221 measurement points. Using the normalized factor of normal tissues under the excitation source with 365 nm of the central wavelength and without the bandpass filter, the results revealed that the sensitivity was larger than 84%, the specificity was not smaller than over 76%, the accuracy was about 80%, and the area under curve of the receiver operating characteristic (ROC) was achieved at about 87%, respectively. The fact shows the LIAF spectroscopy has the possibilities of ex vivo diagnosis and noninvasive examinations for oral cancer.