KCKevin T. Chen

Kevin T. Chen

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Kevin T. Chen has a total of 113 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    A PET MOVIE for interpretable synthesis of late-frame [11C]-PiB PET images from early-frame counterparts

    by Bo-Wei Tsai, Yu-Nong Lin, Hsin-Ta Lin, Yi-Shih Li, Guan-Lin Huang, Chia-Ju Liu, Chi-Lun Ko, Ruoh-Fang Yen, Hsin-Hsi Tsai, Kevin T. Chen

  • Open Access

    Domain-aligned self-supervision: CT-specialized DINOv3 enhances PET/CT tumor segmentation

    by Yu-Nong Scarlett Lin, Shansong Wang, Yuheng Li, Mojtaba Safari, Qiang Li, Zach Eidex, Yizhou Wu, Ryan Sanford, Kevin T Chen, Xiaofeng Yang

    Objective. This study investigates how domain-aligned self-supervised pretraining influences downstream PET/CT tumor segmentation, with a focus on disentangling the roles of domain proximity and pretraining dataset scale. Approach. We compared nnU-Net segmentation models initialized from random weights or from DINOv3 encoders pretrained on natural images, x-ray mammography, CT, or PET, using datasets of different sizes. To disentangle the effects of pretraining domain and dataset size, CT-DINOv3 encoders were pretrained using different numbers of 2D CT slices, including subsets matched in size to the pretraining datasets from other domains. Models were evaluated on two PET/CT benchmarks, autoPET (whole-body, FDG/prostate-specific membrane antigen (PSMA)) and HECKTOR (head-and-neck primary tumors and metastatic lymph nodes), using five-fold cross-validation with frozen encoders. Main results. CT-pretrained models achieved the highest Dice scores for lymph nodes, FDG lesions, and PSMA lesions. For primary tumors, PET pretraining produced the highest observed Dice score, followed by CT pretraining, suggesting the potential benefit of in-domain representations when PET data are sufficient. Notably, PET pretraining exceeded size-matched CT pretraining for PSMA but remained below models pretrained on larger CT datasets for both PSMA and FDG targets. In contrast, natural-image pretraining did not outperform from-scratch training despite its much larger pretraining scale, indicating that scale alone is insufficient when domain proximity is low. Across comparisons, strong transfer reflected the combination of domain proximity and sufficient pretraining scale, especially for CT-based models with broad data availability. Significance. These results demonstrate that both domain proximity and dataset scale jointly determine the effectiveness of self-supervised pretraining. While higher domain alignment improves performance at comparable scale, sufficient data is required to realize this benefit. CT-based pretraining provides a favorable balance between domain relevance and data availability, leading to robust performance across tasks. This study provides practical guidance for selecting pretraining strategies in PET/CT segmentation.

  • Open Access

    Analysis of Upper Airway Morphology Using Four‐Dimensional Dynamic MRI With Active Deep Learning‐Based Automatic Segmentation

    by Cheng‐Yang Yu, Meng‐Chen Chung, Yunn‐Jy Chen, Han‐Wei Wang, Jonathan X. Zhou, Shih‐Lung Chen, Kevin T. Chen, Tiffany Ting‐Fang Shih

    Background: Upper‐airway morphology changes during breathing can be captured with cine 4D MRI. Active‐learning nnU‐Net reduces manual labeling while maintaining accuracy. Purpose: For automatic upper airway segmentation on free‐breathing cine 4D MRI using active learning and quantifying dynamic changes under two mouth positions. Study Type: Prospective cross‐sectional study. Population: Eighty‐four OSA (obstructive sleep apnea)‐free adults (28 M/56F; 18–80 years; 33 with sleep‐related breathing symptoms). Segmentation performance was evaluated on an internal test set ( n = 18). Fieldstrength/Sequence: 3T, free‐breathing time‐resolved imaging with interleaved stochastic trajectories (TWIST) sequence under closed‐ and open‐mouth positions. Assessment: Manual annotations by a technologist (radiologist‐verified) served as reference standard and training labels for an active‐learning nnU‐Net (68 training; four fixed validation). Total airway length, cross‐sectional area (CSA), and total airway volume were computed at each anatomical level and compared across mouth positions, sex, and sleep‐related symptom status, and independent predictors were identified. Statistical Tests: Paired/unpaired t or Mann–Whitney U test (two‐sided p = 0.05). Predictor selection by 10‐fold LASSO; effects estimated via ordinary least squares with cluster‐robust standard errors. Results: Segmentation achieved a dice 0.959 ± 0.019 (test set). Open‐mouth breathing significantly lengthened the total airway (7.92 ± 1.07 vs. 7.41 ± 0.93 cm) and reduced retropalatal CSA (1.51 ± 0.68 vs. 1.80 ± 0.69 cm 2 ). Coefficients of variation (CVs) for CSA and volume were significantly higher with 20‐s open‐mouth breathing. Males ( n = 28) exhibited significantly larger airway volumes than females (closed 27.94 ± 4.87 vs. 19.82 ± 3.26 cm 3 ; open 30.26 ± 5.94 vs. 20.94 ± 3.85 cm 3 ). Symptomatic individuals ( n = 33) had significantly longer airways (closed 7.96 ± 0.96 vs. 7.04 ± 0.70 cm; open 8.54 ± 1.01 vs. 7.52 ± 0.91 cm), narrower open‐mouth retropalatal CSA (1.24 ± 0.51 vs. 1.68 ± 0.72 cm 2 ), and greater retropalatal CSA dynamic variability. Multivariable regression confirmed mouth position, symptoms, and sex as independent predictors. Data Conclusion: Four‐dimensional cine MRI with active‐learning nnU‐Net can automatically quantify dynamic upper airway morphology, demonstrating systematic differences and dynamic variability. Evidence Level: 2. Technical Efficacy: Stage 2.


  • Open Access

    Paired PET‐MRI Deep Learning Model for Translating [ 11 C]PiB to [ 18 F]Florbetaben Amyloid Images

    by Cheng‐Han Tsai, Shao‐Yi Huang, Yu‐Nong Lin, Han‐Wei Wang, Ing‐Tsung Hsiao, Kevin T. Chen

    Background: Amyloid PET imaging has been extensively employed in the noninvasive assessment of amyloid‐beta accumulation in Alzheimer's disease. Various amyloid radiotracers are commonly used in clinical settings; however, the limited interchangeability among these radiotracers hinders the feasibility of long‐term clinical trials and multicenter comparisons. The Centiloid method was proposed for standardization, though providing a single score per image; voxel‐wise translation remains a formidable task. Purpose: This paper proposes a U‐Net model based on a deformable convolution network (DCNv3‐based U‐Net) for []‐Pittsburgh compound B‐to‐[]‐florbetaben image translation to augment existing datasets for large‐scale model training and provide image information when inconsistencies between visual assessments and the Centiloid scale occur. Methods: The DCNv3‐based U‐Net combined the benefits of deformable convolution that captures long‐range dependencies with efficient computation and the encoder–decoder architecture with skip connections for local–global feature learning and image synthesis. Results: The prediction images presented increased homogeneity to other previous models, closely resembling the texture of []‐florbetaben. Conclusions: The DCNv3‐based U‐Net demonstrated high performance in metrics measurement and statistical analyses for the PET image‐to‐image translation task. This work also justified the importance of MR images in providing structural information.

  • Open Access

    MRI-Styled PET: A Dual Modality Fusion Approach to PET Partial Volume Correction

    by Yu-Nong Lin, Shao-Yi Huang, Cheng-Han Tsai, Han-Wei Wang, Meng-Chen Chung, Enhao Gong, Ing-Tsung Hsiao, Kevin T. Chen

    Positron emission tomography (PET) with 18F-fludeoxyglucose (FDG) can visualize the spatial pattern of neurodegeneration-related glucose hypometabolism.We proposed the “MRI-styled PET”, leveraging anatomical information from T1-weighted magnetic resonance imaging to enhance the structural details and quantitative accuracy of FDG-PET, which is degraded by partial volume effects (PVE). The proposed framework comprised a baseline encoder-decoder image fusion model and several task-specific modules; notably, the alternative anatomical input significantly contributes to correcting the under/overestimation of gray/white matter while the adaptive multi-scale structural similarity loss utilized learnable ratios across various receptive fields to modulate attention to tissue contrast. Compared to traditional anatomy-guided postreconstruction PVE correction method (PVC-PET), MRI-styled PET demonstrated significantly higher structural similarity and peak signal-to-noise ratio than the baseline image fusion model (Baseline), showcasing the effectiveness of the proposed taskspecific modules. In several Alzheimers Disease-related brain regions, MRI-styled PET exhibited consistent increases in corrective effects regardless of disease stage, compared to Baseline and PVC-PET. In conclusion, this study represented an initial exploration of a deep-learning approach for correcting PVE in PET without prior knowledge regarding the correction method or the underlying radiotracer uptake and without assumptions about the system point-spread function. Our implementation is available at https://github.com/NTUMMIO/MRI-styled-PET


  • Open Access

    Early-Frame [18F]Florbetaben PET/MRI for Cerebral Blood Flow Quantification in Patients with Cognitive Impairment: Comparison to an [15O]Water Gold Standard

    by Ates Fettahoglu, Moss Zhao, Mehdi Khalighi, Hillary Vossler, Maria Jovin, Guido Davidzon, Michael Zeineh, Fernando Boada, Elizabeth Mormino, Victor W. Henderson, et al.

    Cerebral blood flow (CBF) may be estimated from early-frame PET imaging of lipophilic tracers, such as amyloid agents, enabling measurement of this important biomarker in participants with dementia and memory decline. Although previous methods could map relative CBF, quantitative measurement in absolute units (mL/100 g/min) remained challenging and has not been evaluated against the gold standard method of [15O]water PET. The purpose of this study was to develop and validate a minimally invasive quantitative CBF imaging method combining early [18F]florbetaben (eFBB) with phase-contrast MRI using simultaneous PET/MRI. Methods: Twenty participants (11 men and 9 women; 8 cognitively normal, 9 with mild cognitive impairment, and 3 with dementia; 10 β-amyloid negative and 10 β-amyloid positive; 69 ± 9 y old) underwent [15O]water PET, phase-contract MRI, and eFBB imaging in a single session on a 3-T PET/MRI scanner. Quantitative CBF images were created from the first 2 min of brain activity after [18F]florbetaben injection combined with phase-contrast MRI measurement of total brain blood flow. These maps were compared with [15O]water CBF using concordance correlation (CC) and Bland–Altman statistics for gray matter, white matter, and individual regions derived from the automated anatomic labeling (AAL) atlas. Results: The 2 methods showed similar results in gray matter ([15O]water, 55.2 ± 14.7 mL/100 g/min; eFBB, 55.9 ± 14.2 mL/100 g/min; difference, 0.7 ± 2.4 mL/100 g/min; P = 0.2) and white matter ([15O]water, 21.4 ± 5.6 mL/100 g/min; eFBB, 21.2 ± 5.3 mL/100 g/min; difference, −0.2 ± 1.0 mL/100 g/min; P = 0.4). The intrasubject CC for AAL-derived regions was high (0.91 ± 0.04). Intersubject CC in different AAL-derived regions was similarly high, ranging from 0.86 for midfrontal regions to 0.98 for temporal regions. There were no significant differences in performance between the methods in the amyloid-positive and amyloid-negative groups as well as participants with different cognitive statuses. Conclusion: We conclude that eFBB PET/MRI can provide robust CBF measurements, highlighting the capability of simultaneous PET/MRI to provide measurements of both CBF and amyloid burden in a single imaging session in participants with memory disorders. Visual Abstract


  • Generative Adversarial Network–Enhanced Ultra-Low-Dose [18F]-PI-2620 τ PET/MRI in Aging and Neurodegenerative Populations

    by K.T. Chen, R. Tesfay, M.E.I. Koran, J. Ouyang, S. Shams, C.B. Young, G. Davidzon, T. Liang, M. Khalighi, E. Mormino, et al.

    BACKGROUND AND PURPOSE: With the utility of hybrid τ PET/MR imaging in the screening, diagnosis, and follow-up of individuals with neurodegenerative diseases, we investigated whether deep learning techniques can be used in enhancing ultra-low-dose [18F]-PI-2620 τ PET/MR images to produce diagnostic-quality images. MATERIALS AND METHODS: Forty-four healthy aging participants and patients with neurodegenerative diseases were recruited for this study, and [18F]-PI-2620 τ PET/MR data were simultaneously acquired. A generative adversarial network was trained to enhance ultra-low-dose τ images, which were reconstructed from a random sampling of 1/20 (approximately 5% of original count level) of the original full-dose data. MR images were also used as additional input channels. Region-based analyses as well as a reader study were conducted to assess the image quality of the enhanced images compared with their full-dose counterparts. RESULTS: The enhanced ultra-low-dose τ images showed apparent noise reduction compared with the ultra-low-dose images. The regional standard uptake value ratios showed that while, in general, there is an underestimation for both image types, especially in regions with higher uptake, when focusing on the healthy-but-amyloid-positive population (with relatively lower τ uptake), this bias was reduced in the enhanced ultra-low-dose images. The radiotracer uptake patterns in the enhanced images were read accurately compared with their full-dose counterparts. CONCLUSIONS: The clinical readings of deep learning–enhanced ultra-low-dose τ PET images were consistent with those performed with full-dose imaging, suggesting the possibility of reducing the dose and enabling more frequent examinations for dementia monitoring.


  • CNS Machine Learning

    by Elizabeth Tong, Endre Grøvik, Kyrre Eeg Emblem, Kevin Chen, Audrey Fan, Yannan Yu, Guangming Zhu, Moss Zhao, Sanaz Niri, Greg Zaharchuk


  • Assessment of motion and model bias on the detection of dopamine response to behavioral challenge

    by Michael A Levine, Joseph B Mandeville, Finnegan Calabro, David Izquierdo-Garcia, Daniel B Chonde, Kevin T Chen, Inki Hong, Julie C Price, Beatriz Luna, Ciprian Catana

    Compartmental modeling analysis of 11C-raclopride (RAC) PET data can be used to measure the dopaminergic response to intra-scan behavioral tasks. Bias in estimates of binding potential (BPND) and its dynamic changes (ΔBPND) can arise both when head motion is present and when the compartmental model used for parameter estimation deviates from the underlying biology. The purpose of this study was to characterize the effects of motion and model bias within the context of a behavioral task challenge, examining the impacts of different mitigation strategies. Seventy healthy adults were administered bolus plus constant infusion RAC during a simultaneous PET/magnetic resonance (MR) scan with a reward task experiment. BPND and ΔBPND were estimated using an extension of the Multilinear Reference Tissue Model (E-MRTM2) and a new method (DE-MRTM2) was proposed to selectively discount the contribution of the initial uptake period. Motion was effectively corrected with a standard frame-based approach, which performed equivalently to a more complex reconstruction-based approach. DE-MRTM2 produced estimates of ΔBPND in putamen and nucleus accumbens that were significantly different from those estimated from E-MRTM2, while also decoupling ΔBPND values from first-pass k2′ estimation and removing skew in the spatial bias distribution of parametric ΔBPND estimates within the striatum.