Yujing Zou
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
Research Subjects
Yujing Zou has a total of 60 co-authors. Here are the top co-authors by number of shared publications:
- Confidence Raymond
Medical Artificial Intelligence Lab
2 shared publications - Dong Zhang
McGill University
2 shared publications Farhad Maleki
Department of Computer Science, University of Calgary, Alberta, Canada
2 shared publications- Harry Glickman
Medical Physics Unit, Department of Oncology, Faculty of Medicine, McGill University, Montreal, Quebec, Canada
2 shared publications - Jill Sunday
Technical University of Mombasa
2 shared publications - Kendi Muchungi
Brain Mind Institute, The Aga Khan University
2 shared publications
Latest Publications
- Open Access
- 6 June 2026
- Published by Springer Nature in npj Precision Oncology
- Vol. 10 (1)
Prognostic data extraction harnessing a privacy-preserving large language model: a clinician-AI collaborative retrospective evaluation in head and neck oncology
Privacy regulations and limited expert-validation constrain the deployment of large language models (LLMs) for electronic health record structuring. We evaluated locally deployed LLMs to extract 30 prognostic variables from 1,360 head and neck cancer reports (882 patients) using zero-shot prompting. A stratified 50-case subset was reviewed by three radiation oncologists (50 cases, 30 fields, 3 reviewers; 4,500 decisions) to form a majority-vote reference for Llama3.3-70B, which achieved 98.6% F1 with high clinician agreement and processed reports in 53 s/report. Among seven additional models (2.6B-70B) benchmarked against this reference, GPT-OSS-20.9B (F1 89.4%) and MedGemma-27B (F1 88.5%) performed best. Integrating LLM-extracted HPV status, smoking history, and Charlson Comorbidity Score into a multivariate Cox Proportional Hazards model (age, sex, T/N stage) improved disease-free survival (likelihood ratio test p = 0.014; ΔC-index + 0.071) and locoregional failure-free survival (p = 0.026; ΔC-index + 0.108) with 1,000-bootstrap internal validation. This clinician-AI collaborative evaluation shows that on-premises LLMs enable privacy-preserving and efficient tumour board support, longitudinal data curation, and outcome prediction.
- 6 June 2026
- Published by Springer Nature in npj Precision Oncology
- Vol. 10 (1)
Open AccessPrognostic data extraction harnessing a privacy-preserving large language model: a clinician-AI collaborative retrospective evaluation in head and neck oncology
Privacy regulations and limited expert-validation constrain the deployment of large language models (LLMs) for electronic health record structuring. We evaluated locally deployed LLMs to extract 30 prognostic variables from 1,360 head and neck cancer reports (882 patients) using zero-shot prompting. A stratified 50-case subset was reviewed by three radiation oncologists (50 cases, 30 fields, 3 reviewers; 4,500 decisions) to form a majority-vote reference for Llama3.3-70B, which achieved 98.6% F1 with high clinician agreement and processed reports in 53 s/report. Among seven additional models (2.6B-70B) benchmarked against this reference, GPT-OSS-20.9B (F1 89.4%) and MedGemma-27B (F1 88.5%) performed best. Integrating LLM-extracted HPV status, smoking history, and Charlson Comorbidity Score into a multivariate Cox Proportional Hazards model (age, sex, T/N stage) improved disease-free survival (likelihood ratio test p = 0.014; ΔC-index + 0.071) and locoregional failure-free survival (p = 0.026; ΔC-index + 0.108) with 1,000-bootstrap internal validation. This clinician-AI collaborative evaluation shows that on-premises LLMs enable privacy-preserving and efficient tumour board support, longitudinal data curation, and outcome prediction.
- Open Access
- 1 May 2026
- Published by Elsevier in Physics and Imaging in Radiation Oncology
- Vol. 39
Trends in brachytherapy utilization in Canada from 2011 to 2020
- 1 May 2026
- Published by Elsevier in Physics and Imaging in Radiation Oncology
- Vol. 39
Open AccessTrends in brachytherapy utilization in Canada from 2011 to 2020
- Open Access
- 19 June 2025
- Published by Elsevier in Physics and Imaging in Radiation Oncology
- Vol. 35, 100793
Tumour nuclear size heterogeneity as a biomarker for post-radiotherapy outcomes in gynecological malignancies
- 19 June 2025
- Published by Elsevier in Physics and Imaging in Radiation Oncology
- Vol. 35, 100793
Open AccessTumour nuclear size heterogeneity as a biomarker for post-radiotherapy outcomes in gynecological malignancies
- 9 February 2025
- Published by Springer Nature
- p. 110-123
Generative Style Transfer for MR Image Segmentation: A Case of Glioma Segmentation in Sub-Saharan Africa
- 9 February 2025
- Published by Springer Nature
- p. 110-123
Generative Style Transfer for MR Image Segmentation: A Case of Glioma Segmentation in Sub-Saharan Africa
- Open Access
Generative Style Transfer for MRI Image Segmentation: A Case of Glioma Segmentation in Sub-Saharan Africa
In Sub-Saharan Africa (SSA), the utilization of lower-quality Magnetic Resonance Imaging (MRI) technology raises questions about the applicability of machine learning methods for clinical tasks. This study aims to provide a robust deep learning-based brain tumor segmentation (BraTS) method tailored for the SSA population using a threefold approach. Firstly, the impact of domain shift from the SSA training data on model efficacy was examined, revealing no significant effect. Secondly, a comparative analysis of 3D and 2D full-resolution models using the nnU-Net framework indicates similar performance of both the models trained for 300 epochs achieving a five-fold cross-validation score of 0.93. Lastly, addressing the performance gap observed in SSA validation as opposed to the relatively larger BraTS glioma (GLI) validation set, two strategies are proposed: fine-tuning SSA cases using the GLI+SSA best-pretrained 2D fullres model at 300 epochs, and introducing a novel neural style transfer-based data augmentation technique for the SSA cases. This investigation underscores the potential of enhancing brain tumor prediction within SSA's unique healthcare landscape.
Open AccessGenerative Style Transfer for MRI Image Segmentation: A Case of Glioma Segmentation in Sub-Saharan Africa
In Sub-Saharan Africa (SSA), the utilization of lower-quality Magnetic Resonance Imaging (MRI) technology raises questions about the applicability of machine learning methods for clinical tasks. This study aims to provide a robust deep learning-based brain tumor segmentation (BraTS) method tailored for the SSA population using a threefold approach. Firstly, the impact of domain shift from the SSA training data on model efficacy was examined, revealing no significant effect. Secondly, a comparative analysis of 3D and 2D full-resolution models using the nnU-Net framework indicates similar performance of both the models trained for 300 epochs achieving a five-fold cross-validation score of 0.93. Lastly, addressing the performance gap observed in SSA validation as opposed to the relatively larger BraTS glioma (GLI) validation set, two strategies are proposed: fine-tuning SSA cases using the GLI+SSA best-pretrained 2D fullres model at 300 epochs, and introducing a novel neural style transfer-based data augmentation technique for the SSA cases. This investigation underscores the potential of enhancing brain tumor prediction within SSA's unique healthcare landscape.
- Open Access
- 9 December 2019
- Published by Springer Nature in BMC Immunology
- Vol. 20 (1) , 1-9
NK cell predicts the severity of acute graft-versus-host disease in patients after allogeneic stem cell transplantation using antithymocyte globulin (ATG) in pretreatment scheme
Background: Graft-versus-host disease (GVHD) is one of the most complex complications after allogeneic stem cell transplantation. Current standard of grading system is based on clinical symptoms in skin, liver and intestinal. However, it’s difficult to differ GVHD and its extent just by clinical manifestation. Here we retrospectively analyzed cell immune function in patients implemented allogeneic stem cell transplantation in Ningbo first Hospital from Jan 2013 to Jan 2018. Results: the data are collected from 51 patients (mean age was 42; 45.1% women). The average NK cell percentage was 39.31% in severe GVHD (Grade III-IV), was 16.98% in mild GVHD (GradeI-II), while was 21.15% in No GVHD group. The statistical analysis showed difference among each grade. Further analysis was performed in Antithymocyte globulin (ATG) treated group and control group. We showed NK Cell percentage was sharply different in ATG treated group: 47.34% in severe GVHD, 11.98% in mild GVHD group, while 18.3% in no GVHD group. However, in control group, the average percentage of NK cells was 23.27% in severe GVHD, was 23.22%in mild GVHD group, while was 21.13% in no GVHD group. Conclusion: The data supports that ATG can prevent GVHD by increasing NK cell percentage. The percentage of NK cell seemed to be a useful probe to evaluate the severity of GVHD in allogeneic stem cell transplantation patients using ATG in pretreatment.
- 9 December 2019
- Published by Springer Nature in BMC Immunology
- Vol. 20 (1) , 1-9
Open AccessNK cell predicts the severity of acute graft-versus-host disease in patients after allogeneic stem cell transplantation using antithymocyte globulin (ATG) in pretreatment scheme
Background: Graft-versus-host disease (GVHD) is one of the most complex complications after allogeneic stem cell transplantation. Current standard of grading system is based on clinical symptoms in skin, liver and intestinal. However, it’s difficult to differ GVHD and its extent just by clinical manifestation. Here we retrospectively analyzed cell immune function in patients implemented allogeneic stem cell transplantation in Ningbo first Hospital from Jan 2013 to Jan 2018. Results: the data are collected from 51 patients (mean age was 42; 45.1% women). The average NK cell percentage was 39.31% in severe GVHD (Grade III-IV), was 16.98% in mild GVHD (GradeI-II), while was 21.15% in No GVHD group. The statistical analysis showed difference among each grade. Further analysis was performed in Antithymocyte globulin (ATG) treated group and control group. We showed NK Cell percentage was sharply different in ATG treated group: 47.34% in severe GVHD, 11.98% in mild GVHD group, while 18.3% in no GVHD group. However, in control group, the average percentage of NK cells was 23.27% in severe GVHD, was 23.22%in mild GVHD group, while was 21.13% in no GVHD group. Conclusion: The data supports that ATG can prevent GVHD by increasing NK cell percentage. The percentage of NK cell seemed to be a useful probe to evaluate the severity of GVHD in allogeneic stem cell transplantation patients using ATG in pretreatment.
- Open Access
- 13 November 2019
- Published by American Society of Hematology in Blood
- Vol. 134 (Supplement) , 4433
Microrna-191 Regulates T-Cell Clonal Expansion during Graft-Versus-Host Disease
Allogeneic hematopoietic cell transplantation is a potentially curative treatment choice for a wide variety of hematological malignancies. However, graft-versus-host disease (GVHD), which is mediated by donor alloreactive T cells, limits the success of this procedure. Previous studies have demonstrated that several microRNAs (miRs) modulate graft-versus-host disease. miR-191 was previously reported to be able to support T cell survival after TCR stimulation. We hypothesize that miR191 regulates T cell response during GVHD. To test this hypothesis, we first studied miR-191 expression in alloreactive T cells. The result demonstrated that miR-191 was up-regulated in donor T cells isolated from murine GVHD recipients, suggesting that miR-191 may play a role in GVHD induction. We further studied the role of miR-191in GVHD using miR-191 deficient T cells (KO). Lethally irradiated (8.5 Gy) BALB/c mice were injected intravenously with 1×107 T cell-depleted bone marrow (TCDBM) cells along with 1×106 purified T cells from wild-type (WT) or KO mice, which are in C57BL/6 background. Interestingly, all recipients in the WT group died within 35 days after transplantation, while only one out of ten animals died in the KO group during an observation period of 56 days. Body weights and clinical scores were also improved in KO T cell recipients when compared with the WT controls. Similar results were also observed in a second GVHD model (C57BL/6→C3H/HeJ). To understand the mechanism by which miR-191 KO T cells have decreased ability to mediate GVHD, we first measured the ability of KO T cells to respond to alloantigens in vitro in a mixed lymphocytes reaction assay. Dramatically decreased alloresponse was observed with KO T cells as compared with WT T cells. Similarly, decreased clonal expansion was observed in KO T cells in vivo upon challenge with alloantigens as measured by bioluminescent imaging (Figure 1A). These results were further supported by data from a co-transfer experiment, in which equal numbers of WT and KO T cells were transplanted into the same GVHD recipient. At day7 after transplantation, KO T cells showed significantly reduced expansion in the spleen and liver compared with WT T cells. Reduced alloresponses mediated by KO T cells may not due to decreased proliferative capability directly as an in vivo carboxyfluorescein succinimidyl ester (CFSE) assay showed a comparable cell division between WT and KO T cells upon challenge with alloantigens. Rather, increased cell death is responsible for decreased alloresponse observed in KO T cells because dramatically increased number of dead cells was observed in KO group compared with WT group upon response to alloantigens in vitro and vivo. To determine the genes that are regulated by miR-191, we did a screening based on the prediction. Humans and mice share more than 100 predicted targets for miR-191. We chose top 20 of these targets for RT-qPCR screening. The result demonstrated that Taf5 was a target gene of miR-191. Expression of TAF5 protein was down-regulated in activated KO T cells when compared with the WT T cells. Finally, we investigated whether miR-191 KO T cells preserve graft-versus-leukemia effects. 1×106 T cells from WT or KO mice were transplanted into lethally irradiated BALB/c mice along with 1×107 TCDBM cells and 1×105 host-type BCL-1 cells. While all recipients that received only TCDBM and tumor cells developed lethal leukemia/lymphoma, none of WT and KO T cells recipients developed tumor. In conclusion, our findings reveal a critical role of miR-191 during GVHD process and demonstrate that miR-191 is a novel therapeutic target for GVHD.
- 13 November 2019
- Published by American Society of Hematology in Blood
- Vol. 134 (Supplement) , 4433
Open AccessMicrorna-191 Regulates T-Cell Clonal Expansion during Graft-Versus-Host Disease
Allogeneic hematopoietic cell transplantation is a potentially curative treatment choice for a wide variety of hematological malignancies. However, graft-versus-host disease (GVHD), which is mediated by donor alloreactive T cells, limits the success of this procedure. Previous studies have demonstrated that several microRNAs (miRs) modulate graft-versus-host disease. miR-191 was previously reported to be able to support T cell survival after TCR stimulation. We hypothesize that miR191 regulates T cell response during GVHD. To test this hypothesis, we first studied miR-191 expression in alloreactive T cells. The result demonstrated that miR-191 was up-regulated in donor T cells isolated from murine GVHD recipients, suggesting that miR-191 may play a role in GVHD induction. We further studied the role of miR-191in GVHD using miR-191 deficient T cells (KO). Lethally irradiated (8.5 Gy) BALB/c mice were injected intravenously with 1×107 T cell-depleted bone marrow (TCDBM) cells along with 1×106 purified T cells from wild-type (WT) or KO mice, which are in C57BL/6 background. Interestingly, all recipients in the WT group died within 35 days after transplantation, while only one out of ten animals died in the KO group during an observation period of 56 days. Body weights and clinical scores were also improved in KO T cell recipients when compared with the WT controls. Similar results were also observed in a second GVHD model (C57BL/6→C3H/HeJ). To understand the mechanism by which miR-191 KO T cells have decreased ability to mediate GVHD, we first measured the ability of KO T cells to respond to alloantigens in vitro in a mixed lymphocytes reaction assay. Dramatically decreased alloresponse was observed with KO T cells as compared with WT T cells. Similarly, decreased clonal expansion was observed in KO T cells in vivo upon challenge with alloantigens as measured by bioluminescent imaging (Figure 1A). These results were further supported by data from a co-transfer experiment, in which equal numbers of WT and KO T cells were transplanted into the same GVHD recipient. At day7 after transplantation, KO T cells showed significantly reduced expansion in the spleen and liver compared with WT T cells. Reduced alloresponses mediated by KO T cells may not due to decreased proliferative capability directly as an in vivo carboxyfluorescein succinimidyl ester (CFSE) assay showed a comparable cell division between WT and KO T cells upon challenge with alloantigens. Rather, increased cell death is responsible for decreased alloresponse observed in KO T cells because dramatically increased number of dead cells was observed in KO group compared with WT group upon response to alloantigens in vitro and vivo. To determine the genes that are regulated by miR-191, we did a screening based on the prediction. Humans and mice share more than 100 predicted targets for miR-191. We chose top 20 of these targets for RT-qPCR screening. The result demonstrated that Taf5 was a target gene of miR-191. Expression of TAF5 protein was down-regulated in activated KO T cells when compared with the WT T cells. Finally, we investigated whether miR-191 KO T cells preserve graft-versus-leukemia effects. 1×106 T cells from WT or KO mice were transplanted into lethally irradiated BALB/c mice along with 1×107 TCDBM cells and 1×105 host-type BCL-1 cells. While all recipients that received only TCDBM and tumor cells developed lethal leukemia/lymphoma, none of WT and KO T cells recipients developed tumor. In conclusion, our findings reveal a critical role of miR-191 during GVHD process and demonstrate that miR-191 is a novel therapeutic target for GVHD.