Yi-Cheng Zhu

Mr. Yi-Cheng Zhu


Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

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

Representative Publications

View All Publications
22 September 2023 in Medical & Biological Engineering & Computing
Verified
Journal Article
Open access
PDF file
Alaa AlZoubiAlaa AlZoubi
Alaa AlZoubi
FLFeng Lu
Feng Lu
YZYi-Cheng Zhu
Yi-Cheng Zhu
TYTao Ying
Tao Ying
MAMohmmed Ahmed
Mohmmed Ahmed
Hongbo DuHongbo Du
Hongbo Du
Deep convolutional neural networks (DCNNs) have demonstrated promising performance in classifying breast lesions in 2D ultrasound (US) images. Exiting approaches typically use pre-trained models based on architectures designed for natural images with transfer learning. Fewer attempts have been made to design customized architectures specifically for this purpose. This paper presents a comprehensive evaluation on transfer learning based solutions and automatically designed networks, analyzing the accuracy and robustness of different recognition models in three folds. First, we develop six different DCNN models (BNet, GNet, SqNet, DsNet, RsNet, IncReNet) based on transfer learning. Second, we adapt the Bayesian optimization method to optimize a CNN network (BONet) for classifying breast lesions. A retrospective dataset of 3034 US images collected from various hospitals is then used for evaluation. Extensive tests show that the BONet outperforms other models, exhibiting higher accuracy (83.33%), lower generalization gap (1.85%), shorter training time (66 min), and less model complexity (approximately 0.5 million weight parameters). We also compare the diagnostic performance of all models against that by three experienced radiologists. Finally, we explore the use of saliency maps to explain the classification decisions made by different models. Our investigation shows that saliency maps can assist in comprehending the classification decisions. Graphical Abstract

18 July 2019 in Medical Science Monitor
Verified
Journal Article
Closed access
DZDao-Ming Zu
Dao-Ming Zu
LFLing-Ling Feng
Ling-Ling Feng
LZLei Zhang
Lei Zhang
SMSu-Li Ma
Su-Li Ma
YZYi-Cheng Zhu
Yi-Cheng Zhu
BACKGROUND This study aimed to evaluate superb microvascular imaging (SMI) as an adjunctive imaging method to evaluate mesenteric lymph nodes in children with mesenteric lymphadenitis compared with healthy children. MATERIAL AND METHODS A retrospective study compared children with mesenteric lymphadenitis (n=27) and healthy children (n=30). Lymph node size was determined using grayscale ultrasonography and parameters of lymph node vascularity were compared using color Doppler flow imaging (CDFI) and SMI. The diagnostic performance of ultrasound (US), US combined with SMI, and US combined with CDFI were compared. RESULTS Lymph nodes from children with mesenteric lymphadenitis (n=77) and normal lymph nodes (n=84) were evaluated by SMI, which showed that the least diameter of lymph nodes in cases of mesenteric lymphadenitis was 0.58±0.15 mm and of normal mesenteric lymph nodes was 0.47±0.08 mm (p<0.001). SMI identified 92.6% of abnormal mesenteric lymph nodes while CDFI detected 85.2%. US combined with SMI had the highest sensitivity (81.5%), and specificity (78.9%) compared with US alone (sensitivity, 63.0%; specificity, 64.9%), and compared with US combined with CDFI (sensitivity, 74.1%; specificity, 75.4%). US combined with SMI and US combined with CDFI achieved the same specificity (76.7%), which was higher than that of US alone (66.7%). CONCLUSIONS SMI was superior to color Doppler flow imaging in evaluating the microvasculature in lymphadenopathy in mesenteric lymphadenitis. SMI may be used as an adjunct to grayscale ultrasonography to assist in identifying mesenteric lymphadenopathy in pediatric patients.

23 April 2024 in Experimental and Therapeutic Medicine
Verified
Journal Article
Open access
PDF file
LZLi Zhou
Li Zhou
JSJun Shan
Jun Shan
DZDao-Ming Zu
Dao-Ming Zu
SDShu-Hao Deng
Shu-Hao Deng
YZYuan Zhang
Yuan Zhang
QJQuan Jiang
Quan Jiang
The present retrospective study was designed to explore the value of conventional ultrasound (US) and Virtual Touch Tissue Imaging and Quantification (VTIQ) in the assessment of mesenteric lymphadenitis (ML) in a paediatric population. A total of 103 patients with ML and 60 healthy paediatric patients were examined. VTIQ was performed to assess mesenteric lymph node (MLN) stiffness via shear‑wave velocity (SWV). Univariate and multivariate logistic regression analyses were conducted to reveal independent variables for the identification of ML. The diagnostic performance of US, and US combined with VTIQ, were compared. All the quantitative VTIQ parameters (including the SWVMean, SWVMax and SWVMin) were significantly greater for MLNs in the control group than for MLNs in the ML group (all P<0.001). The SWV values in the control group were nearly 2‑fold greater than that in the ML group. According to the multivariate logistic regression analysis, the longest diameter [odds ratio (OR)=6.042; P=0.046] was revealed to be the strongest independent predictor for ML, followed by the CRP level (OR=2.310; P<0.001) and the SWVMean (OR=0.106; P<0.001). According to the receiver operating characteristic analysis, the area under the curve (AUC) for US combined with VTIQ was 0.890 (95% CI: 0.831‑0.949) with a greater sensitivity of 91.26% and a greater specificity of 86.67% than that for US alone (AUC: 0.798; 95% CI: 0.724‑0.872; sensitivity: 79.61%; specificity: 80.00%). A significant negative correlation between increased VTIQ parameters and ML was observed. Utilizing VTIQ to assess MLN stiffness offers a non‑invasive, convenient, reliable and reproducible approach for identifying mesenteric lymphadenopathy.

1 November 2023 in Translational Cancer Research
Verified
Journal Article
Open access
PDF file
WZWenzhong Zhang
Wenzhong Zhang
YWYueming Wang
Yueming Wang
SDShuhao Deng
Shuhao Deng
YZYi-Cheng Zhu
Yi-Cheng Zhu
Background: As either oncogenes or tumor suppressor genes, long non-coding RNAs (lncRNAs) have a major role in both tumorigenesis and progression of human cancers, including breast cancer (BC). However, the statistical correlation between the lncRNA-lncRNA interaction and prognosis of BC remains unclear. Methods: We analyzed the fragments per kilobase per million (FPKM) lncRNA expression data in tumor tissue samples from 890 female patients with BC in The Cancer Genome Atlas (TCGA) between May 2021 and October 2022. The Cox proportional hazards model adjusted for age, race, clinical stage, neoadjuvant therapy, estrogen receptor (ER), and progesterone receptor (PR) was adopted to evaluate the lncRNA-lncRNA interaction regarding overall survival (OS) of BC. The multiple comparison was corrected by Bonferroni method. Results: RP11-10E18.7×RP11-481C4.2 was significantly associated with OS of BC patients [hazard ratio (HR)interaction =1.04, 95% confidence interval (CI): 1.03–1.06, P=3.35×10−9]. Then, gene-gene interaction analysis was performed for genes co-expressed with lncRNAs. FOXA1×U2SURP (HRinteraction =1.49, 95% CI: 1.28–1.73, P=2.16×10−7) was found to have a similar interactive pattern to RP11-10E18.7×RP11-481C4.2. after classifying the patients by intersection (3.47), we observed that the effect of FOXA1 opposite in patients with different U2SURP expression level (HRhighvs. low =0.58, 95% CI: 0.34–0.99, P=0.046 in low expression of U2SURP; HRhighvs. low =1.56, 95% CI: 1.18–2.87, P=0.029 in high expression of U2SURP). Conclusions: Our comprehensive study identified RP11-10E18.7×RP11-481C4.2 as a potential biomarker of BC prognosis. The results play an essential role in the impact of lncRNA-lncRNA interaction on BC survival. Our findings elucidated potential molecular mechanisms of BC progression under complex association patterns and provided potential dynamic and reversible therapeutic targets for BC patients.