Chen-Hua Chiang

Chen-Hua Chiang

Auto-generated profile, based on publicly available information

Chen-Hua Chiang

From the Department of Radiology (C.-H.C., S.B., A.H., A.T., J.C., S.Y.H.), Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
11 June 2021 in PLOS ONE
Verified
Journal Article
Open access
PDF file
CCChen-Hua Chiang
Chen-Hua Chiang
CWChi-Lun Weng
Chi-Lun Weng
HCHung-Wen Chiu
Hung-Wen Chiu
Modern radiologic images comply with DICOM (digital imaging and communications in medicine) standard, which, upon conversion to other image format, would lose its image detail and information such as patient demographics or type of image modality that DICOM format carries. As there is a growing interest in using large amount of image data for research purpose and acquisition of large amount of medical image is now a standard practice in the clinical setting, efficient handling and storage of large amount of image data is important in both the clinical and research setting. In this study, four classes of images were created, namely, CT (computed tomography) of abdomen, CT of brain, MRI (magnetic resonance imaging) of brain and MRI of spine. After converting these images into JPEG (Joint Photographic Experts Group) format, our proposed CNN architecture could automatically classify these 4 groups of medical images by both their image modality and anatomic location. We achieved excellent overall classification accuracy in both validation and test sets (> 99.5%), specificity and F1 score (> 99%) in each category of this dataset which contained both diseased and normal images. Our study has shown that using CNN for medical image classification is a promising methodology and could work on non-DICOM images, which could potentially save image processing time and storage space.

22 July 2025 in Journal of NeuroInterventional Surgery
Verified
Journal Article
Closed access
PDF file
CCChen-Hua Chiang
Chen-Hua Chiang
CLChia-Chuan Liu
Chia-Chuan Liu
CWChi-Lun Weng
Chi-Lun Weng
LCLung Chan
Lung Chan
LTLillian Tsao
Lillian Tsao
DCDavid Yen-Ting Chen
David Yen-Ting Chen
Background Current automatic software uses a fixed apparent diffusion coefficient (ADC) threshold (≤620×10⁻⁶ mm²/s) to quantify stroke volume before endovascular thrombectomy (EVT). Clinically, infarcts are visually detected by diffusion-weighted imaging (DWI) hyperintensity coupled with ADC hypointensity, which encompasses a broad range of ADC values, often exceeding the threshold. We investigated the clinical significance of large discrepancies between DWI hyperintense volumes and ADC ≤620×10⁻⁶ mm²/s volumes (DWI-ADC mismatch) on pre-EVT MRI. Methods This retrospective single-center study included consecutive patients with first-ever acute ischemic stroke who underwent EVT for symptomatic anterior circulation large vessel occlusion within 24 hours (January 2018–January 2020) and had high-quality pre- and post-EVT MRI. DWI-ADC mismatch was defined as a ratio of DWI hyperintense volume to ADC ≤620×10⁻⁶ mm²/s volume of ≥2. Demographics, clinical characteristics, procedural details, and outcomes were compared between mismatch and no-mismatch groups. Results Of 73 patients included, 20 (27.4%) had DWI-ADC mismatch. The mismatch group showed slower infarct growth (3.8 vs 7.5 mL/hour, P=0.04), significantly more parent artery stenosis (65% vs 20.8%, P<0.001), and a greater need for angioplasty/stenting (45% vs 17%, P=0.013). Imaging analysis showed a higher DWI reversal percentage (37.7% vs 21.2%, P=0.02) and a trend towards a lower ADC lowering percentage (25.3% vs 32.2%, P=0.05). Other characteristics were comparable. Conclusions Pre-EVT DWI-ADC mismatch identifies patients with slower infarct progression, distinct procedural requirements, and a higher likelihood of tissue reversal. Recognizing this pattern may improve EVT planning and optimize outcomes.

2 April 2026 in American Journal of Neuroradiology
Verified
Journal Article
Closed access
AHAzadeh Hajati
Azadeh Hajati
CCChen-Hua Chiang
Chen-Hua Chiang
SBSittaya Buathong
Sittaya Buathong
ATAzadeh Tabari
Azadeh Tabari
SYSeonghwan Yee
Seonghwan Yee
SHSusie Y. Huang
Susie Y. Huang

3 June 2026 in American Journal of Neuroradiology
Verified
Journal Article
Closed access
CCChen-Hua Chiang
Chen-Hua Chiang
SBSittaya Buathong
Sittaya Buathong
AHAzadeh Hajati
Azadeh Hajati
ATAzadeh Tabari
Azadeh Tabari
WLWei-Ching Lo
Wei-Ching Lo
SHSusie Y. Huang
Susie Y. Huang
Contrast-enhanced volume interpolated breath-hold examination (VIBE) is commonly used for evaluating internal auditory canal (IAC) pathology. Deep learning (DL)-reconstruction has been shown to reduce scan time while maintaining diagnostic quality. Our purpose was to compare the image quality of a DL-reconstructed post-contrast-enhanced VIBE sequence with standard 3D VIBE in a clinical setting. This study included 74 patients undergoing contrast-enhanced MRI for the assessment of IAC pathology. The participants were scanned with a standard 3D VIBE sequence and DL-reconstructed VIBE (DL-VIBE) sequences. Two neuroradiologists performed a head-to-head evaluation of both sequences in terms of image-quality metrics, identification of key anatomic structures, and pathologic enhancement when present. A third neuroradiologist adjudicated the discrepant results. Superiority testing was performed to compare DL-VIBE with standard VIBE in image quality metrics, and a noninferiority test with a 15% margin was performed to test the visibility of key anatomic structures and pathologic enhancing lesions. Interrater agreement and quantitative assessment of an image-wise SNR, contrast-noise ratio, IAC-specific SNR, and tumor volumetric measurements as applicable were performed. DL-VIBE enabled approximately 25% scan time reduction (from 3 minutes 2 seconds on the standard sequence to 2 minutes 17 seconds on the DL sequence). DL-VIBE was superior to standard VIBE for all image-quality metrics (P < .01). DL-VIBE was noninferior to standard VIBE for structural visibility of key structures (P < .001) and was noninferior for pooled pathologic enhancement (P = .049). Interrater agreements were moderate to almost perfect across all categories (κ = 0.515-1). DL-VIBE provided significantly higher mean image-wise SNR (71.8 [SD, 23.2] versus 39.4 [SD, 10.7], respectively; P < .001), while providing comparable lesion conspicuity and volumetric assessment. DL-VIBE is a strong alternative to standard 3D VIBE for clinical IAC assessment, providing superior image quality and comparable visibility of key anatomic structures and enhancing pathology while offering higher image SNR with 25% reduced scan time.

8 October 2025 in American Journal of Neuroradiology
Verified
Journal Article
Closed access
PDF file
SFShohei Fujita
Shohei Fujita
DPDaniel Polak
Daniel Polak
MNMarcel Dominik Nickel
Marcel Dominik Nickel
DSDaniel Nicolas Splitthoff
Daniel Nicolas Splitthoff
YHYantu Huang
Yantu Huang
SHSusie Y. Huang
Susie Y. Huang
BACKGROUND AND PURPOSE: Motion artifacts remain a key limitation in brain MRI, particularly during 3D acquisitions in cognitively impaired patients. Most deep learning (DL) reconstruction techniques improve signal-to-noise ratio but lack explicit mechanisms to correct for motion. This study aims to validate a DL reconstruction method that integrates retrospective motion correction into the reconstruction pipeline for 3D T1-weighted brain MRI. MATERIALS AND METHODS: This prospective, intra-individual comparison study included a controlled-motion cohort of healthy volunteers and a clinical cohort of patients undergoing evaluation for memory loss. Each cohort was scanned at distinct imaging sites between October 2022 and August 2023 in staggered periods. All participants underwent 4-fold under-sampled 3D magnetization-prepared rapid gradient-echo imaging with integrated Scout Accelerated Motion Estimation and Reduction (SAMER) acquisition. Image volumes were reconstructed using standard-of-care methods and the proposed DL approach. Quantitative morphometric accuracy was assessed by comparing brain segmentation results of instructed-motion scans to motion-free reference scans in the healthy volunteers. Image quality was rated by two board-certified neuroradiologists using a five-point Likert scale. Statistical analysis included Wilcoxon tests and intraclass correlation coefficients. RESULTS: A total of 41 participants (15 women [37%]; mean age, 58 years) and 154 image volumes were evaluated. The DL-based method with integrated motion correction significantly reduced segmentation error under moderate and severe motion (12.4% to 3.5% and 44.2% to 12.5%, respectively; P < .001). Visual ratings showed improved scores across all criteria compared with standard reconstructions (overall image quality, 4.26 ± 0.72 vs. 3.59 ± 0.82; P < .001). In 47% of cases, motion artifact severity was improved following DL-based processing. Inter-reader agreement ranged from moderate to substantial. CONCLUSIONS: Motion-informed DL reconstruction improved both morphometric accuracy and perceived image quality in 3D T1-weighted brain MRI. This technique may enhance diagnostic utility and reduce scan failure rates in motion-prone patients with cognitive impairment. ABBREVIATIONS: AD = Alzheimer’s disease; DL = deep learning; ICC = intra-class correlation coefficient; SAMER = scout accelerated motion estimation and reduction

14 March 2025 in medRxiv
Verified
Preprint
Open access
PDF file
CCChen-Hua Chiang
Chen-Hua Chiang
CLChia-Chuan Liu
Chia-Chuan Liu
CWChi-Lun Weng
Chi-Lun Weng
LCLung Chan
Lung Chan
LTLillian Tsao
Lillian Tsao
DCDavid Yen-Ting Chen
David Yen-Ting Chen
Background: Current automatic software uses a fixed apparent diffusion coefficient (ADC) threshold (≤620×10⁻⁶ mm²/s) to quantify stroke volume for endovascular thrombectomy (EVT) decision-making. In contrast, clinical evaluation relies on visual identification of DWI hyperintensity with corresponding ADC hypointensity—which may exhibit varied ADC values in hyperacute stroke. This study investigates the clinical and imaging significance of the significant discrepancy between DWI hyperintense volume and ADC ≤ 620 × 10⁻⁶ mm²/s volume on pre-EVT magnetic resonance imaging (MRI), defined as DWI-ADC mismatch. Method: This retrospective, single-center study analyzed consecutive patients diagnosed with acute ischemic stroke (AIS) between January 2018 and January 2020. Inclusion criteria consisted of patients with symptomatic anterior circulation large vessel occlusion (LVO) within 24 hours and treated by EVT, with high-quality pre- and post-EVT MRI scans. Infarct segmentation was performed by neuroradiologists. DWI-ADC mismatch was defined as a ratio of segmented DWI volume to ADC volume (≤620×10⁻⁶ mm²/s) ≥2. Data on demographics, baseline and follow-up clinical and imaging characteristics, procedural details, and outcomes were collected and analyzed. Result: 73 patients were included in the study, with 20 patients (27.4%) demonstrating a DWI-ADC mismatch. The mismatch group exhibited a significantly slower infarct growth rate (3.8 vs. 7.5 ml/h, P = 0.04), higher prevalence of parent artery stenosis (65% vs. 20.8%, P < 0.001) and a greater need for angioplasty and/or stenting (50% vs. 17%, P < 0.001). Imaging analysis showed a higher percentage of DWI reversal (37.7% vs. 21.2%, P = 0.02) and a trend toward lower ADC lowering percentage (25.3% vs. 32.2%, P = 0.05) in the mismatch group. Demographics, baseline and follow-up lesion volumes, and functional outcomes were comparable between groups. Conclusion: DWI-ADC mismatch identifies a distinct subgroup of stroke patients with slower infarct progression and unique procedural needs. Incorporating mismatch evaluation into pre-EVT imaging may refine patient selection and optimize treatment strategies.