TransEBUS: The interpretation of endobronchial ultrasound image using hybrid transformer for differentiating malignant and benign mediastinal lesions.

The purpose of this study is to establish a deep learning automatic assistance diagnosis system for benign and malignant classification of mediastinal lesions in endobronchial ultrasound (EBUS) images. EBUS images are in the form of video and contain multiple imaging modes. Different imaging modes and different frames can reflect the different characteristics of lesions. Compared with previous studies, the proposed model can efficiently extract and integrate the spatiotemporal relationships between different modes and does not require manual selection of representative frames. In recent years, Vision Transformer has received much attention in the field of computer vision. Combined with convolutional neural networks, hybrid transformers can also perform well on small datasets. This study designed a novel deep learning architecture based on hybrid transformer called TransEBUS. By adding learnable parameters in the temporal dimension, TransEBUS was able to extract spatiotemporal features from insufficient data. In addition, we designed a two-stream module to integrate information from three different imaging modes of EBUS. Furthermore, we applied contrastive learning when training TransEBUS, enabling it to learn discriminative representation of benign and malignant mediastinal lesions. The results show that TransEBUS achieved a diagnostic accuracy of 82% and an area under the curve of 0.8812 in the test dataset, outperforming other methods. It also shows that several models can improve performance by incorporating two-stream module. Our proposed system has shown its potential to help physicians distinguishing benign and malignant mediastinal lesions, thereby ensuring the accuracy of EBUS examination.

Cold Atmospheric Plasma Jet Irradiation Decreases the Survival and the Expression of Oncogenic miRNAs of Oral Carcinoma Cells.

Despite recent advancements, therapies against advanced oral squamous cell carcinoma (OSCC) remain ineffective, resulting in unsatisfactory therapeutic outcomes. Cold atmospheric plasma (CAP) offers a promising approach in the treatment of malignant neoplasms. Although the effects of CAP in abrogating OSCC have been explored, the exact mechanisms driving CAP-induced cancer cell death and the changes in microRNA (miRNA) expression are not fully understood. We fabricated and calibrated an argon-CAP device to explore the effects of CAP irradiation on the growth and expression of oncogenic miRNAs in OSCC. The analysis revealed that, in OSCC cell lines following CAP irradiation, there was a significant reduction in viability; a downregulation of miR-21, miR-31, miR-134, miR-146a, and miR-211 expression; and an inactivation of the v-akt murine thymoma viral oncogene homolog (AKT) and extracellular signal-regulated kinase (ERK) signals. Pretreatment with blockers of apoptosis, autophagy, and ferroptosis synergistically reduced CAP-induced cell death, indicating a combined induction of variable death pathways via CAP. Combined treatments using death inhibitors and miRNA mimics, alongside the activation of AKT and ERK following the exogenous expression, counteracted the cell mortality associated with CAP. The CAP-induced downregulation of miR-21, miR-31, miR-187, and miR-211 expression was rescued through survival signaling. Additionally, CAP irradiation notably inhibited the growth of SAS OSCC cell xenografts on nude mice. The reduced expression of oncogenic miRNAs in vivo aligned with in vitro findings. In conclusion, our study provides new lines of evidence demonstrating that CAP irradiation diminishes OSCC cell viability by abrogating survival signals and oncogenic miRNA expression.

Passive Magnetic-Flux-Concentrator Based Electromagnetic Targeting System for Endobronchoscopy.

In this paper, we demonstrate an innovative electromagnetic targeting system utilizing a passive magnetic-flux-concentrator for tracking endobronchoscope used in the diagnosis process of lung cancer tumors/lesions. The system consists of a magnetic-flux emitting coil, a magnetic-flux receiving electromagnets-array, and high permeability silicon-steel sheets rolled as a collar (as the passive magnetic-flux-concentrator) fixed in a guide sheath of an endobronchoscope. The emitting coil is used to produce AC magnetic-flux, which is consequently received by the receiving electromagnets-array. Due to the electromagnetic-induction, a voltage is induced in the receiving electromagnets-array. When the endobronchoscope's guide sheath (with the silicon-steel collar) travels between the emitting coil and the receiving electromagnets-arrays, the magnetic flux is concentrated by the silicon-steel collar and thereby the induced voltage is changed. Through analyzing the voltage-pattern change, the location of the silicon-steel collar with the guide sheath is targeted. For testing, a bronchial-tree model for training medical doctors and operators is used to test our system. According to experimental results, the system is successfully verified to be able to target the endobronchoscope in the bronchial-tree model. The targeting errors on the x-, y- and z-axes are 9 mm, 10 mm, and 5 mm, respectively.

Using Micro-Molding and Stamping to Fabricate Conductive Polydimethylsiloxane-Based Flexible High-Sensitivity Strain Gauges.

In this study, polydimethylsiloxane (PDMS) and conductive carbon nanoparticles were combined to fabricate a conductive elastomer PDMS (CPDMS). A high sensitive and flexible CPDMS strain sensor is fabricated by using stamping-process based micro patterning. Compared with conventional sensors, flexible strain sensors are more suitable for medical applications but are usually fabricated by photolithography, which suffers from a large number of steps and difficult mass production. Hence, we fabricated flexible strain sensors using a stamping-process with fewer processes than photolithography. The piezoresistive coefficient and sensitivity of the flexible strain sensor were improved by sensor pattern design and thickness change. Micro-patterning is used to fabricate various CPDMS microstructure patterns. The effect of gauge pattern was evaluated with ANSYS simulations. The piezoresistance of the strain gauges was measured and the gauge factor determined. Experimental results show that the piezoresistive coefficient of CPDMS is approximately linear. Gauge factor measurement results show that the gauge factor of a 140.0 µm thick strain gauge with five grids is the highest.

Feature-enhanced adversarial semi-supervised semantic segmentation network for pulmonary embolism annotation.

This study established a feature-enhanced adversarial semi-supervised semantic segmentation model to automatically annotate pulmonary embolism (PE) lesion areas in computed tomography pulmonary angiogram (CTPA) images. In the current study, all of the PE CTPA image segmentation methods were trained by supervised learning. However, when CTPA images come from different hospitals, the supervised learning models need to be retrained and the images need to be relabeled. Therefore, this study proposed a semi-supervised learning method to make the model applicable to different datasets by the addition of a small number of unlabeled images. By training the model with both labeled and unlabeled images, the accuracy of unlabeled images was improved and the labeling cost was reduced. Our proposed semi-supervised segmentation model included a segmentation network and a discriminator network. We added feature information generated from the encoder of the segmentation network to the discriminator so that it could learn the similarities between the prediction label and ground truth label. The HRNet-based architecture was modified and used as the segmentation network. This HRNet-based architecture could maintain a higher resolution for convolutional operations to improve the prediction of small PE lesion areas. We used a labeled open-source dataset and an unlabeled National Cheng Kung University Hospital (NCKUH) (IRB number: B-ER-108-380) dataset to train the semi-supervised learning model, and the resulting mean intersection over union (mIOU), dice score, and sensitivity reached 0.3510, 0.4854, and 0.4253, respectively, on the NCKUH dataset. Then we fine-tuned and tested the model with a small number of unlabeled PE CTPA images in a dataset from China Medical University Hospital (CMUH) (IRB number: CMUH110-REC3-173). Comparing the results of our semi-supervised model with those of the supervised model, the mIOU, dice score, and sensitivity improved from 0.2344, 0.3325, and 0.3151 to 0.3721, 0.5113, and 0.4967, respectively. In conclusion, our semi-supervised model can improve the accuracy on other datasets and reduce the labor cost of labeling with the use of only a small number of unlabeled images for fine-tuning.

Using DeepLab v3 + -based semantic segmentation to evaluate platelet activation.

This research used DeepLab v3 + -based semantic segmentation to automatically evaluate the platelet activation process and count the number of platelets from scanning electron microscopy (SEM) images. Current activated platelet recognition and counting methods include (a) using optical microscopy or SEM images to identify and manually count platelets at different stages, or (b) using flow cytometry to automatically recognize and count platelets. However, the former is time- and labor-consuming, while the latter cannot be employed due to the complicated morphology of platelet transformation during activation. Additionally, because of how complicated the transformation of platelets is, current blood-cell image analysis methods, such as logistic regression or convolution neural networks, cannot precisely recognize transformed platelets. Therefore, this study used DeepLab v3 + , a powerful learning model for semantic segmentation of image analysis, to automatically recognize and count platelets at different activation stages from SEM images. Deformable convolution, a pretrained model, and deep supervision were added to obtain additional platelet transformation features and higher accuracy. The number of activated platelets was predicted by dividing the segmentation predicted platelet area by the average platelet area. The results showed that the model counted the activated platelets at different stages from the SEM images, achieving an error rate within 20%. The error rate was approximately 10% for stages 2 and 4. The proposed approach can thus save labor and time for evaluating platelet activation and facilitate related research.

Effectiveness of convolutional neural networks in the interpretation of pulmonary cytologic images in endobronchial ultrasound procedures.

BACKGROUND: Rapid on-site cytologic evaluation (ROSE) helps to improve the diagnostic accuracy in endobronchial ultrasound (EBUS) procedures. However, cytologists are seldom available to perform ROSE in many institutions. Recent studies have investigated the application of deep learning in cytologic image analysis. As such, the present study analyzed lung cytologic images obtained by EBUS procedures, and employed deep-learning methods to distinguish between benign and malignant cells and to semantically segment malignant cells. METHODS: Ninety-seven patients who underwent 104 EBUS procedures were enrolled. Four hundred and ninety-nine lung cytologic images obtained via ROSE, including 425 malignant and 74 benign, and most malignant were lung adenocarcinoma (64.3%). All the images were used to train a residual network model with 101 layers (ResNet101), with suitable hyperparameters selected to classify benign and malignant lung cytologic images. An HRNet model was also employed to mark the area of malignant cells. Automatic patch-cropping was adopted to facilitate dataset preparation. RESULTS: Malignant cells were successfully classified by ResNet101 with 98.8% classification accuracy, 98.8% sensitivity, and 98.8% specificity in patch-based classification; 95.5% classification accuracy in image-based classification; and 92.9% classification accuracy in patient-based classification. Malignant cell area was successfully marked by HRNet with a mean intersection over union of 89.2%. The automatic cropping method enabled the system to complete diagnosis within 1 s. CONCLUSIONS: This is the first study to combine lung cytologic image deep-learning classification with semantic segmentation. The model was optimized for high accuracy and the automatic cropping facilitates the clinical application of our model. The success in both lung cytologic images classification and semantic segmentation on our dataset shows a promising result for clinical application in the future.

Plasma-activated medium as adjuvant therapy for lung cancer with malignant pleural effusion.

This study compared effects of plasma-activated medium (PAM) with effects of conventional clinical thermal therapy on both lung cancer cells and benign cells for management of malignant pleural effusion (MPE). For MPE treatment, chemotherapy, photodynamic therapy, and thermal therapy are used but caused systemic side effects, patient photosensitivity, and edema, respectively. Recent studies show that plasma induces apoptosis in cancer cells with minor effects on normal cells and is cost-effective. However, the effects of plasma on MPE have not been investigated previously. This study applied a nonthermal atmospheric-pressure plasma jet to treat RPMI medium to produce PAM, carefully controlled the long-life reactive oxygen and nitrogen species concentration in PAM, and treated the cells. The influence of PAM treatment on the microenvironment of cells was also checked. The results indicated that PAM selectively inhibited CL1-5 and A549 cells, exerting minor effects on benign mesothelial and fibroblast cells. In contrast to selective lethal effects of PAM, thermal therapy inhibited both CL1-5 and benign mesothelial cells. This study also found that fibroblast growth factor 1 is not the factor explaining why PAM can selectively inhibit CL1-5 cells. These results indicate that PAM is potentially a less-harmful and cost-effective adjuvant therapy for MPE.

Enhancing the Microparticle Deposition Stability and Homogeneity on Planer for Synthesis of Self-Assembly Monolayer.

The deposition stability and homogeneity of microparticles improved with mask, lengthened nozzle and flow rate adjustment. The microparticles can be used to encapsulate monomers, before the monomers in the microparticles can be deposited onto a substrate for nanoscale self-assembly. For the uniformity of the synthesized nanofilm, the homogeneity of the deposited microparticles becomes an important issue. Based on the ANSYS simulation results, the effects of secondary flow were minimized with a lengthened nozzle. The ANSYS simulation was also used to investigate the ring-vortex generation and why the ring vortex can be eliminated by adding a mask with an aperture between the nozzle and deposition substrate. The experimental results also showed that particle deposition with a lengthened nozzle was more stable, while adding the mask stabilized deposition and diminished the ring-vortex contamination. The effects of flow rate and pressure were also investigated. Hence, the deposition stability and homogeneity of microparticles was improved.

Printing Peptide arrays with a complementary metal oxide semiconductor chip.

: In this chapter, we discuss the state-of-the-art peptide array technologies, comparing the spot technique, lithographical methods, and microelectronic chip-based approaches. Based on this analysis, we describe a novel peptide array synthesis method with a microelectronic chip printer. By means of a complementary metal oxide semiconductor chip, charged bioparticles can be patterned on its surface. The bioparticles serve as vehicles to transfer molecule monomers to specific synthesis spots. Our chip offers 16,384 pixel electrodes on its surface with a spot-to-spot pitch of 100 µm. By switching the voltage of each pixel between 0 and 100 V separately, it is possible to generate arbitrary particle patterns for combinatorial molecule synthesis. Afterwards, the patterned chip surface serves as a printing head to transfer the particle pattern from its surface to a synthesis substrate. We conducted a series of proof-of-principle experiments to synthesize high-density peptide arrays. Our solid phase synthesis approach is based on the 9-fluorenylmethoxycarbonyl protection group strategy. After melting the particles, embedded monomers diffuse to the surface and participate in the coupling reaction to the surface. The method demonstrated herein can be easily extended to the synthesis of more complicated artificial molecules by using bioparticles with artificial molecular building blocks. The possibility of synthesizing artificial peptides was also shown in an experiment in which we patterned biotin particles in a high-density array format. These results open the road to the development of peptide-based functional modules for diverse applications in biotechnology.

Peptide arrays with a chip.

Today, lithographic methods enable combinatorial synthesis of >50,000 oligonucleotides per cm(2), an advance that has revolutionized the whole field of genomics. A similar development is expected for the field of proteomics, provided that affordable, very high-density peptide arrays are available. However, peptide arrays lag behind oligonucleotide arrays. This is mainly due to the monomer-by-monomer repeated consecutive coupling of 20 different amino acids associated with lithography, which adds up to an excessive number of coupling cycles. A combinatorial synthesis based on electrically charged solid amino acid particles resolves this problem. A computer chip consecutively addresses the different charged particles to a solid support, where, when completed, the whole layer of solid amino acid particles is melted at once. This frees hitherto immobilized amino acids to couple all 20 different amino acids in one single coupling reaction to the support. The method should allow for the translation of entire genomes into a set of overlapping peptides to be used in proteome research.

Nonlinear pulse compression in pulse-inversion fundamental imaging.

Coded excitation can be applied in ultrasound contrast agent imaging to enhance the signal-to-noise ratio with minimal destruction of the microbubbles. Although the axial resolution is usually compromised by the requirement for a long coded transmit waveforms, this can be restored by using a compression filter to compress the received echo. However, nonlinear responses from microbubbles may cause difficulties in pulse compression and result in severe range side-lobe artifacts, particularly in pulse-inversion-based (PI) fundamental imaging. The efficacy of pulse compression in nonlinear contrast imaging was evaluated by investigating several factors relevant to PI fundamental generation using both in-vitro experiments and simulations. The results indicate that the acoustic pressure and the bubble size can alter the nonlinear characteristics of microbubbles and change the performance of the compression filter. When nonlinear responses from contrast agents are enhanced by using a higher acoustic pressure or when more microbubbles are near the resonance size of the transmit frequency, higher range side lobes are produced in both linear imaging and PI fundamental imaging. On the other hand, contrast detection in PI fundamental imaging significantly depends on the magnitude of the nonlinear responses of the bubbles and thus the resultant contrast-to-tissue ratio (CTR) still increases with acoustic pressure and the nonlinear resonance of microbubbles. It should be noted, however, that the CTR in PI fundamental imaging after compression is consistently lower than that before compression due to obvious side-lobe artifacts. Therefore, the use of coded excitation is not beneficial in PI fundamental contrast detection.