AI and machine-learning researcher completing a PhD at the University of Manchester, specialising in Alzheimer’s disease prognosis and early detection from plasma biomarkers combined with genetic and clinical data. First author of two journal articles, with four international conference presentations. Experienced in external cohort validation, leakage-safe nested cross-validation, ensemble learning, SHAP interpretation and translating technical findings for clinical and cross-functional scientific audiences.
Research & professional experience
Doctoral Researcher · Machine Learning for Alzheimer’s Disease Prognosis & Early Detection
University of Manchester · Supervisor: Dr Fumie Costen
- Develop machine-learning systems for amyloid pathology, MCI-to-Alzheimer’s conversion and three-class cognitive staging using plasma biomarkers, APOE genotype and clinical data.
- Designed and externally evaluated four algorithms for amyloid-β PET positivity; random forest reached AUC 0.95 internally on ADNI (n=340), while MLP reached AUC 0.90 on independent CNTN data (n=127) without retraining.
- Built voting and stacking ensembles for disease-progression prediction; a strict nested-CV pipeline achieved AUC 0.888 for three-year MCI-to-AD conversion on ADNI, with the Base + Plasma feature group performing comparably to Base + CSF.
- Quantified label circularity in ADNI cognitive-stage labels and reported realistic non-circular performance: AUC-OVR 0.7455 internally and 0.702 in external zero-shot evaluation.
- Engineered reproducible preprocessing for sparse, multi-site clinical data and used SHAP to communicate model drivers to clinical, biochemistry and neuroscience collaborators.
Sep 2023—present
Expected Q4 2026
Graduate Researcher · Wearable Digital Health
De Montfort University · Industry project with ReTiSense
- Built an end-to-end system spanning Stridalyzer PRISM pressure-sensing insoles, a cross-platform Android/web application and an ML pipeline for controlled motion classification.
- Benchmarked four architectures on 12,000 pressure-map images across six controlled motion classes using stratified five-fold cross-validation; collaborated directly with ReTiSense’s engineering team.
Jan—Jun 2023
Research Intern · Edge AI & Computer Vision
De Montfort University
- Designed a privacy-oriented fall-detection prototype running offline from a single thermal camera, combining non-linear Difference-of-Gaussians processing with a lightweight temporal-memory algorithm.
- Self-collected a 21-clip dataset and systematically tuned four hyperparameters across three scenarios; kept identifiable RGB imagery and cloud transfer out of the design.
Jun—Sep 2022
Selected publications
Xu, J., & Costen, F. (2026). Machine Learning-Based Multiclass Classification of Cognitive Stages Using Plasma Biomarkers, Clinical Assessments, and Genetic Features: A Repeated, Nested Cross-Validation Study in ADNI with External Evaluation in CNTN. Diagnostics, 16(12), 1755. doi:10.3390/diagnostics16121755
Xu, J., Doig, A. J., Michopoulou, S., Proitsi, P., & Costen, F. (2025). Accurate and robust prediction of Amyloid-β brain deposition from plasma biomarkers and clinical information using machine learning. Frontiers in Aging Neuroscience, 17, 1559459. doi:10.3389/fnagi.2025.1559459
International conference presentations
Oral presentation and panel discussion. Predicting MCI-to-AD Conversion Using Ensemble Learning and Multi-Modal Biomarkers. Alzheimer’s Disease International Conference, Lyon, 2026. Selected as the session’s sole PhD-stage speaker and served as the panel’s sole AI/ML speaker.
Poster. An Ensemble Machine Learning Model for the Prediction of the Conversion from Mild Cognitive Impairment to Alzheimer’s Disease Within 6 Years. AD/PD, Copenhagen, 2026.
Poster. Blood-Based Staging of Alzheimer’s Disease—and the Label Circularity That Inflates It. Alzheimer’s Association International Conference, London, 2026.
Poster. Estimation of Amyloid-β PET Status Using Plasma Biomarkers and Clinical Information. AD/PD, Vienna, 2025.
Education
PhD, Electrical & Electronic Engineering
University of Manchester, UK · Machine learning for Alzheimer’s disease
2023—expected Q4 2026
MSc, Intelligent Systems & Robotics
De Montfort University, UK · Distinction, 81/100
2022—2023
Exchange Student
Purdue University Northwest, USA · GPA 4.0/4.0
2020—2021
BEng, Automation
Nanjing University of Science and Technology Zijin College, China
2016—2020
Technical skills
- Programming
- Python, MATLAB, C/C++, Java
- ML / DL
- PyTorch, scikit-learn, random forest, XGBoost, LightGBM, SVM, MLP, CNN, stacking and voting ensembles
- Evaluation
- Nested and repeated cross-validation, hyperparameter optimisation, external validation, sensitivity analysis, SMOTE, Brier score, SHAP
- Data & tooling
- pandas, NumPy, SciPy, Matplotlib, seaborn, Git, Linux, Docker, LaTeX
- Biomedical domain
- Plasma biomarkers (p-Tau217, p-Tau181, Aβ42/40, NfL, GFAP), APOE genotype, ADNI and CNTN cohorts, Alzheimer’s progression, wearable and thermal sensing
Teaching, service, awards & languages
Teaching: Teaching Assistant, Digital Medical Image Processing, University of Manchester (Jan 2024–Jan 2026); mentored more than ten MSc research projects.
Ad hoc peer review: IET Image Processing; IEEE Transactions on Radiation and Plasma Medical Sciences.
Awards: Best Postgraduate Student Prize, De Montfort University (2023); Dean’s List and Semester Honors, Purdue University Northwest (2021); Jiangsu Provincial Second Prize, Blue Bridge Cup C/C++ Programming (2019).
Languages: Mandarin Chinese (native); English (full professional working proficiency).