From Risk Scores to Agents: What "AI" Really Means in Nephrology
A clinician-facing lecture that moves from classical risk scores and machine learning toward deep learning, retrieval, and agentic systems in nephrology.
Every dot is one output at month resolution. Every bar is one continuous role, degree, course or service period. Hover or focus any mark to identify it; select it to open its portfolio record. On a narrow screen, scroll sideways to travel from 2022 to today.
Showing 108 of 108 records
A clinician-facing lecture that moves from classical risk scores and machine learning toward deep learning, retrieval, and agentic systems in nephrology.
Collaborative medical AI research and development across clinical language, specialist agents, imaging, forecasting and physiology, with projects at distinct stages of research and evaluation.
An open-source Windows learning application for nephrology, bringing study, case discussion, examination preparation and evidence-based reference material into one desktop workspace.
A population-level analysis of notifiable infectious disease trends in Poland from 2014 to 2022, covering the period shaped by COVID-19 interventions.
Interpretable machine-learning models are developed to forecast ICU acute kidney injury with actionable lead time, including multi-center validation to test generalizability.
Retrospective MIMIC-IV analysis compares multiple iron-status definitions for one-year mortality risk stratification in congestive heart failure.
A prototype cardiology assistant integrating longitudinal clinical context, imaging, retrieval and specialised model tools with clinician review.
Deep-learning research using chest-radiograph-predicted age as an imaging biomarker of mortality risk.
A compact multimodal model for radiology report classification and image-grounded report generation across languages.
Student research examining whether the gap between radiograph-predicted age and chronological age relates to mortality.
MIMIC-IV cohort analysis of inflammatory and nutritional indices for 90-day mortality prediction in heart failure.
Research combining chest radiographs with clinical variables available near imaging to distinguish viral from bacterial pneumonia.
Conference abstract reporting population-level changes in infectious-disease notifications during and after pandemic interventions.
The IMSC abstract of the Polish infectious-disease surveillance study, with the author order printed in that book.
Conference abstract describing transparent early warning for acute kidney injury with multicentre validation and a 24-hour prediction horizon.
Deyaa presented the chest X-ray age study in COPM’s online oral session. The certificate and programme document the student’s presentation.
Arman Sürmeli’s WIMC presentation of an evidence-grounded multimodal cardiology assistant; the related IMSC abstract names Hourani as coauthor.
A mentee-presented deep learning project validating chest X-ray-predicted age as an imaging biomarker for mortality.
A mentee-presented multimodal reasoning model for chest X-ray labelling and report generation.
Arman Sürmeli’s presentation of the NPI study is retained as related student research, with both supplied and official abstract versions available.
A multi-center validation presentation on interpretable machine learning for forecasting ICU acute kidney injury with clinically actionable lead time.
A generative multimodal AI presentation on synthesizing a full nine-view echocardiography study from ECG and clinical context with uncertainty awareness.
A mentee-presented multimodal model combining chest X-ray and early clinical data to distinguish viral from bacterial pneumonia.
A WCN poster listing and published abstract on transparent AKI early warning with a 24-hour actionable horizon.
Student-presented research on chest-radiograph-predicted age and mortality, with Hourani named as coauthor and supervisor.
A clinician-facing lecture that moves from classical risk scores and machine learning toward deep learning, retrieval, and agentic systems in nephrology.
A benchmark of language models assigning structured CheXpert-style labels to chest-radiograph reports, examining reliability in report-labelling workflows.
Research on age predicted from chest radiographs and its relationship with mortality, including supervised student abstracts, presentations and recognition.
A multimodal language-model project developed through additional training for chest X-ray understanding and radiology report interpretation.
Perception-oriented work connecting vision models, tool use, and system orchestration for embodied and agentic workflows.
An ongoing study of generating nine echocardiographic views from ECG and clinical context, with uncertainty represented in the generated outputs.
An agentic research project combining clinical information in several forms with external tools to investigate cardiology decision workflows.
Population-level time-series work examining notifiable infectious diseases in Poland and the timing of COVID-19 interventions, connected to the preprint, journal article and conference outputs.
Clinical registry curation and outcome modelling with liver-transplant cohorts, supporting reproducible translational analysis.
Post-training multimodal models with attention to grounded reasoning, evaluation quality, and deployment-minded reliability.
A tool-augmented language-model research project for specialist education, with answers grounded in medical evidence and external references.
Collaborative medical AI research and development across clinical language, specialist agents, imaging, forecasting and physiology, with projects at distinct stages of research and evaluation.
Doctoral work on prediction of renal outcomes with an emphasis on fairness, interpretability, and decisions that can be acted on in practice.
An open-source Windows learning application for nephrology, bringing study, case discussion, examination preparation and evidence-based reference material into one desktop workspace.
Clinical AI has spent a decade scoring labels. The next frontier is action-aware patient-state modeling: physiology, measurement, intervention, uncertainty, and consequence.
An in-depth survey of clinical note extraction methods--rule-based, BERT, and LLMs--with error patterns, taxonomy of failures, and implementation strategy.
Behavioral finance explains how cognitive shortcuts and emotional responses distort risk perception and lead even highly educated traders to gamble badly.
Inaugural committee board supporting responsible, evidence-based AI in nephrology.
Postgraduate medical training appointment.
Retrospective cohort study characterizes the presentation and outcomes of patients with brain abscess, highlighting factors linked to hospital course and prognosis.
A retrospective study of a blood-cell-derived inflammatory marker for mortality risk assessment in heart failure with preserved ejection fraction.
Retrospective MIMIC-IV analysis evaluates HALP and modified HALP scores for ninety-day mortality prediction in heart failure.
Preprint version of the infectious disease surveillance analysis, emphasizing how NPIs coincide with shifts in reported incidence across multiple conditions.
Preprint investigating ferritin and transferrin saturation thresholds against mortality outcomes in congestive heart failure.
A conference abstract on integrating signals, imaging, structured measurements and clinical notes for cardiac ICU outcome prediction.
Retrospective analysis examining whether iron-deficiency classifications improve risk stratification in congestive heart failure.
A multimodal heart-failure decision-support research proposal using ECG, chest radiographs and clinical data.
A retrospective cohort analysis of pre-ICU anticoagulation and outcomes in patients with ischaemic stroke.
Coauthored research presented by Arman Sürmeli on iron-deficiency phenotyping and risk prediction in congestive heart failure.
A time-aware prediction project focused on mortality risk in acute coronary syndrome patients.
A hardware-aware multimodal early-warning concept for myocardial infarction patients, built around real-time prediction of critical complications.
A multimodal prediction talk on estimating echocardiographic parameters from ECG and vital-sign signals.
Coauthored MIMIC-IV research presented by Arman Sürmeli on chronic anticoagulation and ICU outcomes in ischaemic stroke.
Awarded for presenting the AI-accelerated early-warning prediction study in the cardiology session.
Recognised for the same presented study in the congress plenary session.
Medical education at the Medical University of Warsaw, grounding subsequent work in clinical research and medical AI.
Doctoral research on clinically useful prediction of kidney outcomes, with attention to interpretability and actionable evaluation.
Study design and methodological planning across clinical specialties and departments.
Production GPU-backed machine-learning services, Kubernetes, Kubeflow and autoscaling infrastructure.
Acknowledgement for organising support for the Clinical AI Solutions Event 2025.
Scientific peer review.
Long-term mentoring of medical students.
Health outreach and first-aid education.
A machine-learning based bleeding-risk score for atrial fibrillation patients receiving oral anticoagulation.
Third place for the presented atrial fibrillation bleeding-risk study in the cardiology session.
Clinical microbiology teaching and student research coordination.
Curriculum and seminar coordination across biostatistics, time-series analysis and machine learning, as described in the archived Europass CV.
Student–department coordination, research-project management and statistical work, as described in the archived Europass CV.
Refresher-course completion certificate for human research training.
A clinical case report on a rare fungal infection, presented within the early WIMC conference work.
An early epidemiology presentation on how non-pharmaceutical interventions reshaped incidence patterns across notifiable infectious diseases in Poland.
Appreciation for participation in the WIMC ECG workshop.
Continuing medical education participation at ESC Congress 2023.
Active participation certificate from OpenCARDIO 2023.
Active participation certificate from the Symbioza conference.
A quantitative framework for assessing market, economic and tenant-related risks in commercial property portfolios.
Quantitative analysis of property-market trends using time-series, regression and machine-learning models to inform investment decisions.
Exploratory development of machine-learning risk scores using MIMIC-IV and eICU, investigating sepsis and cardiovascular complications.
Departmental confirmation and examination record showing the best pathomorphology result of the year.
Data collection, governance and time-series analytics supporting property portfolio decisions.
Recognition for coordination and contribution to AMSA MUW volunteering initiatives.
Recognition for organising and delivering cardiology seminars and workshops.
Recognition for peer-support mentoring and coordination of ultrasound, research and active oral session projects.
Organisational and promotional support as ambassador for the first edition of IMSC in Kraków.
Participation certificate for the 2022–2023 MACR-S programme.
Appreciation for outstanding leadership as MACR-S club president.
Confirmation of leadership of the MUW Students Research Program and research-methodology teaching support.
Teaching and research-methodology support in the biostatistics and telemedicine department, documented in the archived Clinical and Quantitative CV.
Certificate recognising leadership of the active oral presentation participant division; it does not identify a delivered talk.
Volunteer coordination and community service.
Completed Basic Life Support training at the University of Sharjah Clinical Training Center in August 2023, documented by an AHA provider eCard and the course attendance certificate.
Basic-course completion certificate for conflicts of interest training.
Participation certificate for the healthcare-associated infections course.
E-learning completion certificate for palliative medicine.
E-learning completion certificate for patient education.
Attendance certificate for the 17th Warsaw International Medical Congress.
Continuing medical education participation at ESC Congress 2022.
Workshop certificate for point-of-care ultrasound and RUSH protocol training.
Departmental confirmation of the highest dermatology grade for the 2021–2022 academic year.
Financial analysis, cross-department accounting support and real-estate tax reporting.
Participation certificate for the English Division Cardiology Club.
Membership recognition including seminars on shoulder fractures and observation of elective orthopedic surgeries.
Participation certificate for the Internal Medicine Club 7th Floor.
Basic and advanced life support and ultrasound tutoring, documented as a past teaching role in the archived Clinical and Quantitative CV.
Recognition for active membership in the English Division Cardiology Club.
E-learning completion certificate for work safety and hygiene during COVID-19.