Thomas D. Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Thomas D. Wang has a total of 24 co-authors. Here are the top co-authors by number of shared publications:
- Kenn Richard Oldham
Department of Mechanical Engineering, University of Michigan, Ann Arbor, MI, USA
3 shared publications - Gaoming Li
Univ. of Michigan (United States)
2 shared publications - Mayur Birla
Department of Mechanical Engineering, University of Michigan, Ann Arbor, MI, USA
2 shared publications - Miki Lee
Department of Internal Medicine, University of Michigan, Ann Arbor, MI, USA
2 shared publications - Xiyu Duan
Biomedical Engineering, University of Michigan, Ann Arbor, MI, USA
2 shared publications - Chen Gong
Department of Mechanical Engineering, University of Washington, Seattle, WA 98195, USA
1 shared publication
Latest Publications
- Open Access
- 23 August 2022
- Published by MDPI AG in Diagnostics
- Vol. 12 (9) , 2031
Deep-Learning-Based Real-Time and Automatic Target-to-Background Ratio Calculation in Fluorescence Endoscopy for Cancer Detection and Localization
Esophageal adenocarcinoma (EAC) is a deadly cancer that is rising rapidly in incidence. The early detection of EAC with curative intervention greatly improves the prognoses of patients. A scanning fiber endoscope (SFE) using fluorescence-labeled peptides that bind rapidly to epidermal growth factor receptors showed a promising performance for early EAC detection. Target-to-background (T/B) ratios were calculated to quantify the fluorescence images for neoplasia lesion classification. This T/B calculation is generally based on lesion segmentation with the Chan–Vese algorithm, which may require hyperparameter adjustment when segmenting frames with different brightness and contrasts, which impedes automation to real-time video. Deep learning models are more robust to these changes, while accurate pixel-level segmentation ground truth is challenging to establish in the medical field. Since within our dataset the ground truth contained only a frame-level diagnosis, we proposed a computer-aided diagnosis (CAD) system to calculate the T/B ratio in real time. A two-step process using convolutional neural networks (CNNs) was developed to achieve automatic suspicious frame selection and lesion segmentation for T/B calculation. In the segmentation model training for Step 2, the lesion labels were generated with a manually tuned Chan–Vese algorithm using the labeled and predicted suspicious frames from Step 1. In Step 1, we designed and trained deep CNNs to select suspicious frames using a diverse and representative set of 3427 SFE images collected from 25 patient videos from two clinical trials. We tested the models on 1039 images from 10 different SFE patient videos and achieved a sensitivity of 96.4%, a specificity of 96.6%, a precision of 95.5%, and an area under the receiver operating characteristic curve of 0.989. In Step 2, 1006 frames containing suspicious lesions were used for training for fluorescence target segmentation. The segmentation models were tested on two clinical datasets with 100 SFE frames each and achieved mean intersection-over-union values of 0.89 and 0.88, respectively. The T/B ratio calculations based on our segmentation results were similar to the manually tuned Chan–Vese algorithm, which were 1.71 ± 0.22 and 1.72 ± 0.28, respectively, with a p-value of 0.872. With the graphic processing unit (GPU), the proposed two-step CAD system achieved 50 fps for frame selection and 15 fps for segmentation and T/B calculation, which showed that the frame rejection in Step 1 improved the diagnostic efficiency. This CAD system with T/B ratio as the real-time indicator is designed to guide biopsies and surgeries and to serve as a reliable second observer to localize and outline suspicious lesions highlighted by fluorescence probes topically applied in organs where cancer originates in the epithelia.
- 23 August 2022
- Published by MDPI AG in Diagnostics
- Vol. 12 (9) , 2031
Open AccessDeep-Learning-Based Real-Time and Automatic Target-to-Background Ratio Calculation in Fluorescence Endoscopy for Cancer Detection and Localization
Esophageal adenocarcinoma (EAC) is a deadly cancer that is rising rapidly in incidence. The early detection of EAC with curative intervention greatly improves the prognoses of patients. A scanning fiber endoscope (SFE) using fluorescence-labeled peptides that bind rapidly to epidermal growth factor receptors showed a promising performance for early EAC detection. Target-to-background (T/B) ratios were calculated to quantify the fluorescence images for neoplasia lesion classification. This T/B calculation is generally based on lesion segmentation with the Chan–Vese algorithm, which may require hyperparameter adjustment when segmenting frames with different brightness and contrasts, which impedes automation to real-time video. Deep learning models are more robust to these changes, while accurate pixel-level segmentation ground truth is challenging to establish in the medical field. Since within our dataset the ground truth contained only a frame-level diagnosis, we proposed a computer-aided diagnosis (CAD) system to calculate the T/B ratio in real time. A two-step process using convolutional neural networks (CNNs) was developed to achieve automatic suspicious frame selection and lesion segmentation for T/B calculation. In the segmentation model training for Step 2, the lesion labels were generated with a manually tuned Chan–Vese algorithm using the labeled and predicted suspicious frames from Step 1. In Step 1, we designed and trained deep CNNs to select suspicious frames using a diverse and representative set of 3427 SFE images collected from 25 patient videos from two clinical trials. We tested the models on 1039 images from 10 different SFE patient videos and achieved a sensitivity of 96.4%, a specificity of 96.6%, a precision of 95.5%, and an area under the receiver operating characteristic curve of 0.989. In Step 2, 1006 frames containing suspicious lesions were used for training for fluorescence target segmentation. The segmentation models were tested on two clinical datasets with 100 SFE frames each and achieved mean intersection-over-union values of 0.89 and 0.88, respectively. The T/B ratio calculations based on our segmentation results were similar to the manually tuned Chan–Vese algorithm, which were 1.71 ± 0.22 and 1.72 ± 0.28, respectively, with a p-value of 0.872. With the graphic processing unit (GPU), the proposed two-step CAD system achieved 50 fps for frame selection and 15 fps for segmentation and T/B calculation, which showed that the frame rejection in Step 1 improved the diagnostic efficiency. This CAD system with T/B ratio as the real-time indicator is designed to guide biopsies and surgeries and to serve as a reliable second observer to localize and outline suspicious lesions highlighted by fluorescence probes topically applied in organs where cancer originates in the epithelia.
- 1 September 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ASME Transactions on Mechatronics
- Vol. 26 (3) , 1445-1454
Image Processing Metrics for Phase Identification of a Multiaxis MEMS Scanner Used in Single Pixel Imaging
This article applies image processing metrics to tracking of perturbations in mechanical phase delay in a multiaxis microelectromechanical system scanner. The compact mirror is designed to scan a laser beam in a Lissajous pattern during the collection of endoscopic confocal fluorescence images, but environmental perturbations to the mirror dynamics can lead to image registration errors and blurry images. A binarized, threshold-based blur metric and variance-based sharpness metric are introduced for detecting scanner phase delay. Accuracy of local optima of the metric for identification of phase delay is examined, and relative advantages for processing accuracy and computational complexity are assessed. Image reconstruction is demonstrated using both generic images and sample tissue images, with significant improvement in image quality for tissue imaging. Implications of nonideal Lissajous scan effects on phase detection and image reconstruction are discussed.
- 1 September 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ASME Transactions on Mechatronics
- Vol. 26 (3) , 1445-1454
Image Processing Metrics for Phase Identification of a Multiaxis MEMS Scanner Used in Single Pixel Imaging
This article applies image processing metrics to tracking of perturbations in mechanical phase delay in a multiaxis microelectromechanical system scanner. The compact mirror is designed to scan a laser beam in a Lissajous pattern during the collection of endoscopic confocal fluorescence images, but environmental perturbations to the mirror dynamics can lead to image registration errors and blurry images. A binarized, threshold-based blur metric and variance-based sharpness metric are introduced for detecting scanner phase delay. Accuracy of local optima of the metric for identification of phase delay is examined, and relative advantages for processing accuracy and computational complexity are assessed. Image reconstruction is demonstrated using both generic images and sample tissue images, with significant improvement in image quality for tissue imaging. Implications of nonideal Lissajous scan effects on phase detection and image reconstruction are discussed.
- 12 March 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ASME Transactions on Mechatronics
- Vol. 25 (2) , 661-672
Motion Estimation for a Compact Electrostatic Microscanner via Shared Driving and Sensing Electrodes in Endomicroscopy
In this article, we present a method to estimate high-frequency rotary motion of a highly compact electrostatic microscanner using the same electrodes for both actuation and sensing. The accuracy of estimated rotary motion is critical for reducing blur and distortion in image reconstruction applications with the microscanner given its changing dynamics due to perturbations such as temperature. To overcome the limitation that no dedicated sensing electrodes are available in the proposed applications due to size constraints, the method adopts electromechanical amplitude modulation to separate motion signal from parasitic capacitance feedthrough, and a novel nonlinear measurement model is derived to characterize the relationship between large out-of-plane angular motion and circuit output. To estimate motion, an extended Kalman filter and an unscented Kalman filter are implemented, incorporating a process model based on the microscanner's parametric resonant dynamics and the measurement model. Experimental results show that compared to estimation without using the measurement model, our method can improve the rotary motion estimation accuracy of the microscanner significantly, with a reduction of root-mean-square error (RMSE) in phase shift of 86.1%, and a reduction of RMSE in angular position error of 78.5%.
- 12 March 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/ASME Transactions on Mechatronics
- Vol. 25 (2) , 661-672
Motion Estimation for a Compact Electrostatic Microscanner via Shared Driving and Sensing Electrodes in Endomicroscopy
In this article, we present a method to estimate high-frequency rotary motion of a highly compact electrostatic microscanner using the same electrodes for both actuation and sensing. The accuracy of estimated rotary motion is critical for reducing blur and distortion in image reconstruction applications with the microscanner given its changing dynamics due to perturbations such as temperature. To overcome the limitation that no dedicated sensing electrodes are available in the proposed applications due to size constraints, the method adopts electromechanical amplitude modulation to separate motion signal from parasitic capacitance feedthrough, and a novel nonlinear measurement model is derived to characterize the relationship between large out-of-plane angular motion and circuit output. To estimate motion, an extended Kalman filter and an unscented Kalman filter are implemented, incorporating a process model based on the microscanner's parametric resonant dynamics and the measurement model. Experimental results show that compared to estimation without using the measurement model, our method can improve the rotary motion estimation accuracy of the microscanner significantly, with a reduction of root-mean-square error (RMSE) in phase shift of 86.1%, and a reduction of RMSE in angular position error of 78.5%.
- 4 July 2015
- Published by Springer Nature in Biotechnology Letters
- Vol. 37 (11) , 2311-2320
A CD44 specific peptide developed by phage display for targeting gastric cancer
Objective To develop a peptide probe that could be used for gastric cancer detection via binding to CD44 protein with specificity and affinity. Results A 12-mer phage peptide library was screened against immobilized CD44 protein. Bound phage counts using ELISA were performed to identify phage clones carrying the most highly selective peptide, which termed RP-1. Immunofluorescence and flow cytometry analysis indicated that the consensus peptide RP-1 could bind to CD44-positive gastric cancer cells with mean fluorescence intensities significantly higher than that of CD44-negative cells. CD44 knockdown led to decreased binding activity of RP-1 to the same cell line. Tissue array technique was used to identify the relationship (r = 0.556) between peptide binding and CD44 detection on gastric cancer tissues. Further, the hyaluronan-binding domain of CD44 was docked with RP-1 using computer modeling/docking approaches, revealing a RP-1/CD44 interaction with geometrical and energy match (−8.6 kcal/mol). Conclusions The RP-1 peptide we screened exhibits affinity and specificity to CD44 on cells and has the potential to be used as a candidate probe for gastric cancer cell targeting.
- 4 July 2015
- Published by Springer Nature in Biotechnology Letters
- Vol. 37 (11) , 2311-2320
A CD44 specific peptide developed by phage display for targeting gastric cancer
Objective To develop a peptide probe that could be used for gastric cancer detection via binding to CD44 protein with specificity and affinity. Results A 12-mer phage peptide library was screened against immobilized CD44 protein. Bound phage counts using ELISA were performed to identify phage clones carrying the most highly selective peptide, which termed RP-1. Immunofluorescence and flow cytometry analysis indicated that the consensus peptide RP-1 could bind to CD44-positive gastric cancer cells with mean fluorescence intensities significantly higher than that of CD44-negative cells. CD44 knockdown led to decreased binding activity of RP-1 to the same cell line. Tissue array technique was used to identify the relationship (r = 0.556) between peptide binding and CD44 detection on gastric cancer tissues. Further, the hyaluronan-binding domain of CD44 was docked with RP-1 using computer modeling/docking approaches, revealing a RP-1/CD44 interaction with geometrical and energy match (−8.6 kcal/mol). Conclusions The RP-1 peptide we screened exhibits affinity and specificity to CD44 on cells and has the potential to be used as a candidate probe for gastric cancer cell targeting.
- 13 February 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Journal of Microelectromechanical Systems
- Vol. 23 (2) , 256-258
Large Stroke Vertical PZT Microactuator With High-Speed Rotational Scanning
A thin-film piezoelectric microactuator using a novel combination of active vertical translational scanning and passive resonant rotational scanning is presented. Thin-film lead-zirconate-titanate unimorph bending beams surrounding a central platform provide nearly 200-μm displacement at 18 V with bandwidth greater than 200 Hz. Inside the platform, a mirror mount, or mirror surface, supported by silicon dioxide spring beams can be excited to resonance by low-voltage; high-frequency excitation of the outer PZT beams. Over ±5.5° mechanical resonance is obtained at 3.8 kHz and ±2 V. The combination of large translational vertical displacements and high-speed rotational scanning is intended to support real-time cross-sectional imaging in a dual axes confocal endomicroscope.
- 13 February 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Journal of Microelectromechanical Systems
- Vol. 23 (2) , 256-258
Large Stroke Vertical PZT Microactuator With High-Speed Rotational Scanning
A thin-film piezoelectric microactuator using a novel combination of active vertical translational scanning and passive resonant rotational scanning is presented. Thin-film lead-zirconate-titanate unimorph bending beams surrounding a central platform provide nearly 200-μm displacement at 18 V with bandwidth greater than 200 Hz. Inside the platform, a mirror mount, or mirror surface, supported by silicon dioxide spring beams can be excited to resonance by low-voltage; high-frequency excitation of the outer PZT beams. Over ±5.5° mechanical resonance is obtained at 3.8 kHz and ±2 V. The combination of large translational vertical displacements and high-speed rotational scanning is intended to support real-time cross-sectional imaging in a dual axes confocal endomicroscope.
- 24 December 2008
- Published by SAGE Publications in Cancer Biomarkers
- Vol. 4 (6) , 341-350
In vivo cancer biomarkers of esophageal neoplasia
The emergence of in vivo cancer biomarkers is promising tool for early detection, risk stratification, and therapeutic intervention in the esophagus, where adenocarcinoma is increasing at a rate that is faster than any other in industrialized nations
- 24 December 2008
- Published by SAGE Publications in Cancer Biomarkers
- Vol. 4 (6) , 341-350
In vivo cancer biomarkers of esophageal neoplasia
The emergence of in vivo cancer biomarkers is promising tool for early detection, risk stratification, and therapeutic intervention in the esophagus, where adenocarcinoma is increasing at a rate that is faster than any other in industrialized nations