Kelvin Szolnoky

Kelvin Szolnoky

Auto-generated profile, based on publicly available information

Kelvin Szolnoky

Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

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

Representative Publications

View All Publications
7 July 2025 in BMJ Open
Verified
Journal Article
Open access
PDF file
NMNita Mulliqi
Nita Mulliqi
ABAnders Blilie
Anders Blilie
XJXiaoyi Ji
Xiaoyi Ji
KSKelvin Szolnoky
Kelvin Szolnoky
HOHenrik Olsson
Henrik Olsson
MEMartin Eklund
Martin Eklund
Introduction Histopathological evaluation of prostate biopsies using the Gleason scoring system is critical for prostate cancer diagnosis and treatment selection. However, grading variability among pathologists can lead to inconsistent assessments, risking inappropriate treatment. Similar challenges complicate the assessment of other prognostic features like cribriform cancer morphology and perineural invasion. Many pathology departments are also facing an increasingly unsustainable workload due to rising prostate cancer incidence and a decreasing pathologist workforce coinciding with increasing requirements for more complex assessments and reporting. Digital pathology and artificial intelligence (AI) algorithms for analysing whole slide images show promise in improving the accuracy and efficiency of histopathological assessments. Studies have demonstrated AI’s capability to diagnose and grade prostate cancer comparably to expert pathologists. However, external validations on diverse data sets have been limited and often show reduced performance. Historically, there have been no well-established guidelines for AI study designs and validation methods. Diagnostic assessments of AI systems often lack preregistered protocols and rigorous external cohort sampling, essential for reliable evidence of their safety and accuracy. Methods and analysis This study protocol covers the retrospective validation of an AI system for prostate biopsy assessment. The primary objective of the study is to develop a high-performing and robust AI model for diagnosis and Gleason scoring of prostate cancer in core needle biopsies, and at scale evaluate whether it can generalise to fully external data from independent patients, pathology laboratories and digitalisation platforms. The secondary objectives cover AI performance in estimating cancer extent and detecting cribriform prostate cancer and perineural invasion. This protocol outlines the steps for data collection, predefined partitioning of data cohorts for AI model training and validation, model development and predetermined statistical analyses, ensuring systematic development and comprehensive validation of the system. The protocol adheres to Transparent Reporting of a multivariable prediction model of Individual Prognosis Or Diagnosis+AI (TRIPOD+AI), Protocol Items for External Cohort Evaluation of a Deep Learning System in Cancer Diagnostics (PIECES), Checklist for AI in Medical Imaging (CLAIM) and other relevant best practices. Ethics and dissemination Data collection and usage were approved by the respective ethical review boards of each participating clinical laboratory, and centralised anonymised data handling was approved by the Swedish Ethical Review Authority. The study will be conducted in agreement with the Helsinki Declaration. The findings will be disseminated in peer-reviewed publications (open access).

7 April 2026 in European Urology Open Science
Verified
Journal Article
Open access
KSKelvin Szolnoky
Kelvin Szolnoky
ABAnders Blilie
Anders Blilie
NMNita Mulliqi
Nita Mulliqi
TTToyonori Tsuzuki
Toyonori Tsuzuki
HSHemamali Samaratunga
Hemamali Samaratunga
KKKimmo Kartasalo
Kimmo Kartasalo
Cribriform morphology in prostate cancer is a histological feature that indicates poor prognosis and contraindicates active surveillance. However, it remains underreported and subject to significant interobserver variability among pathologists. We aimed to develop and validate an artificial intelligence (AI)-based system to improve cribriform pattern detection. We created a deep learning model using an EfficientNetV2-S encoder with multiple instance learning for end-to-end whole-slide classification. The model was trained on 640 digitised prostate core needle biopsies from 430 patients, collected across three cohorts. It was validated internally (261 slides from 171 patients) and externally (266 slides, 104 patients from three independent cohorts). Internal validation cohorts included laboratories or scanners from the development set, while external cohorts used completely independent instruments and laboratories. Annotations were provided by three expert uropathologists with known high concordance. We assessed model performance using the area under the receiver operating characteristic curve (AUC) and Cohen’s κ with 95% confidence intervals calculated through bootstrapping. Additionally, we conducted an inter-rater analysis and compared the model’s performance against nine expert uropathologists on 88 slides from the internal validation cohort. The model showed strong internal validation performance (AUC: 0.97, 95% CI: 0.95, 0.99; Cohen’s κ: 0.81, 95% CI: 0.72, 0.89) and robust external validation (AUC: 0.90, 95% CI: 0.86, 0.93; Cohen’s κ: 0.55, 95% CI: 0.45, 0.64). In our inter-rater analysis, the model achieved the highest average agreement (Cohen’s κ: 0.66, 95% CI: 0.57, 0.74), outperforming all nine pathologists whose Cohen’s κ ranged from 0.35 to 0.62. Limitations include the retrospective design and that the cross-scanner reproducibility and inter-rater analyses were conducted exclusively on internal validation data, potentially overestimating performance in these analyses. Our AI model demonstrates pathologist-level performance for cribriform morphology detection in prostate cancer. This approach could enhance diagnostic reliability, standardise reporting, and improve treatment decisions for patients with prostate cancer.

6 June 2025 in BMJ Open
Verified
Journal Article
Open access
PDF file
JAJonatan Attergrim
Jonatan Attergrim
KSKelvin Szolnoky
Kelvin Szolnoky
LSLovisa Strömmer
Lovisa Strömmer
OBOlof Brattström
Olof Brattström
GWGunilla Wihlke
Gunilla Wihlke
MWMartin Gerdin Wärnberg
Martin Gerdin Wärnberg
Objective To develop models to predict opportunities for improvement in trauma care and compare the performance of these models to the currently used audit filters. Design Retrospective registry-based study. Setting Single-centre, Scandinavian level one equivalent trauma centre. Participants 8220 adult trauma patients screened for opportunities for improvement between 2013 and 2022. Primary and secondary outcome measures Two machine learning models (logistic regression and XGBoost) and the currently used audit filters were compared. Internal validation by an expanding window approach with annual updates was used for model evaluation. Performance measured by discrimination, calibration, sensitivity and false positive rate of opportunities for improvement prediction. Results A total of 8220 patients, with a mean age of 45 years, were analysed; 69% were men with a mean injury severity score of 12. Opportunities for improvement were identified in 496 (6%) patients. Both the logistic regression and XGBoost models were well-calibrated, with intercalibration indices of 0.02 and 0.02, respectively. The models demonstrated higher areas under the receiver operating characteristic curve (AUCs) (logistic regression: 0.71; XGBoost: 0.74). The XGBoost model had a lower false positive rate at a similar sensitivity (false positive rate: 0.63). The audit filters had an AUC of 0.62 and a false positive rate of 0.67. Conclusions The logistic regression and XGBoost models outperformed audit filters in predicting opportunities for improvement among adult trauma patients and can potentially be used to improve systems for selecting patients for trauma peer review.

22 May 2025 in Trauma Surgery & Acute Care Open
Verified
Journal Article
Open access
PDF file
KSKelvin Szolnoky
Kelvin Szolnoky
EJElias Joneborg
Elias Joneborg
JAJonatan Attergrim
Jonatan Attergrim
HAHussein Albaaj
Hussein Albaaj
LSLovisa Strömmer
Lovisa Strömmer
MWMartin Gerdin Wärnberg
Martin Gerdin Wärnberg
Introduction: Trauma is a leading cause of death in individuals aged 45 and younger, contributing significantly to the global disease burden. Local trauma quality improvement programs have been implemented to improve clinical practice and patient outcomes. Multidisciplinary peer reviews, included in quality improvement programs, aim to identify opportunities for improvement in trauma patient care and implement corrective measures. This study assesses the incidence and trends of these opportunities across clinically important trauma cohorts. Methods: We conducted a retrospective cohort study using data from the trauma registry at Karolinska University Hospital in Solna, Sweden, between 2017 and 2022. Patients screened for opportunities for improvement were categorized into common trauma cohorts. Logistic regression was used to analyze trends in the occurrence of opportunities for improvement over the years in each cohort. The relationship between opportunities for improvement and trauma cohorts was also assessed. Results: Out of 7192 patients included, 404 (6%) had at least one opportunity for improvement. A statistically significant decrease in opportunities for improvement per year was observed overall (OR 0.90; 95% CI 0.84 to 0.95). Significant decreases were identified in patients with blunt multisystem trauma without traumatic brain injury (TBI) (OR 0.82; 95% CI 0.72 to 0.93), isolated severe TBI (OR 0.61; 95% CI 0.41 to 0.91), and severe penetrating injuries (OR 0.68; 95% CI 0.50 to 0.92). The blunt multisystem with TBI cohort showed a non-significant increase. After adjusting for Injury Severity Score, only the blunt multisystem without TBI cohort remained significantly associated with opportunities for improvement (OR 1.69; 95% CI 1.24 to 2.31). Conclusion: The incidence of opportunities for improvement in trauma care showed a significant decrease, indicating that the current trauma quality improvement program at Karolinska University Hospital may be effective in reducing opportunities for improvement. Patients with blunt multisystem trauma without TBI were at higher risk for opportunities for improvement compared with other trauma cohorts. Level of evidence: Level IV: retrospective study with up to three negative criteria.

1 October 2025 in Trauma Surgery & Acute Care Open
Verified
Journal Article
Open access
JAJosef Al-Khalili
Josef Al-Khalili
JAJonatan Attergrim
Jonatan Attergrim
JBJohanna Berg
Johanna Berg
KSKelvin Szolnoky
Kelvin Szolnoky
MWMartin Gerdin Wärnberg
Martin Gerdin Wärnberg
Background: Trauma audit filters, based on standards of care, serve as quality indicators and screening tools for patient cases to be assessed at interdisciplinary morbidity and mortality conferences. These conferences aim to identify potential opportunities for improvement in the care of trauma patients. This study aimed to determine the performance of individual audit filters in predicting opportunities for improvement and their relevance as screening tools for morbidity and mortality conferences. Methods: We conducted a single-center registry-based retrospective cohort study based on registered data collected between 2013 and 2022 from the Karolinska University Hospital in Sweden. Performance was measured using sensitivity, specificity, positive predictive value, negative predictive value, and Matthews correlation coefficient for 10 audit filters used at the Karolinska University Hospital. Results: Our analysis included 8309 patients. Audit filters demonstrated inconsistent agreement with opportunities for improvement with a Matthews correlation coefficient ranging from −0.088 (95% CI −0.129 to −0.044) for “Injury Severity Score >15 and no team activation” to 0.092 (95% CI 0.058 to 0.126) for “> 60 min until first major intervention.” The Matthews correlation coefficient for all audit filters combined was 0.118 (95% CI 0.104 to 0.131). Conclusion: No individual audit filter or combination thereof performed well in predicting opportunities for improvement, which suggests that the current method of screening for morbidity and mortality conferences should be revisited. Level of evidence: Level III, prognostic/epidemiological

24 December 2025 in BMJ Open
Verified
Journal Article
Open access
XJXiaoyi Ji
Xiaoyi Ji
RZRenata Zelic
Renata Zelic
OAOskar Aspegren
Oskar Aspegren
NMNita Mulliqi
Nita Mulliqi
MFMichelangelo Fiorentino
Michelangelo Fiorentino
KKKimmo Kartasalo
Kimmo Kartasalo
Introduction: Prostate cancer diagnosis and treatment planning depend on accurate histopathological assessment of needle biopsies, particularly through the Gleason scoring system. The inherently subjective nature of the grading creates variability between pathologists, potentially resulting in suboptimal patient management decisions. These reproducibility challenges extend beyond Gleason scoring to encompass other critical diagnostic and prognostic markers, including cancer volume quantification and detection of cribriform morphology patterns and perineural invasion. Artificial intelligence (AI) applications in digital pathology have emerged as promising solutions for enhancing diagnostic consistency and accuracy, with recent research demonstrating that automated systems can match expert-level performance in prostate biopsy evaluation. Nevertheless, comprehensive validation studies have revealed concerning limitations in model generalisability when deployed across different clinical environments and patient populations. Recent systematic reviews revealed widespread risk-of-bias limitations and insufficient external validation in AI diagnostic studies, highlighting critical needs for accumulated evidence supporting generalisability before clinical implementation. Rigorous external validation with preregistered protocols using independent datasets from diverse clinical settings remains essential to establish the reliability and safety of AI-assisted prostate pathology systems. Methods and analysis: This study protocol establishes a framework for the retrospective external validation of an AI system developed for prostate biopsy assessment, to be conducted on the case-control samples of the National Prostate Cancer Register of Sweden, ProMort study (1998-2015). The primary aim is to evaluate the AI model’s diagnostic accuracy and Gleason grading performance using completely independent datasets separate from any model development or previously used validation cohorts. The diversity of the validation samples, spanning multiple geographic regions, temporal collection periods and reference standards, allows evaluation of model robustness across varied clinical contexts. Secondary aims encompass evaluating AI performance in cancer length estimation and detection of cribriform patterns and perineural invasion. This protocol delineates procedures for data collection, reference standard clarification and prespecified statistical analyses, ensuring comprehensive validation and reliable performance assessment. The study design conforms to established reporting guidelines Checklist for Artificial Intelligence in Medical Imaging (CLAIM) and Standards for Reporting Diagnostic Accuracy Studies using Artificial Intelligence (STARD-AI), and recognised best practices for AI validation in medical imaging. Ethics and dissemination: Data collection and usage were approved by the Swedish Regional Ethics Review Board and the Swedish Ethical Review Authority (permits 2012/1586-31/1, 2016/613-31/2, 2019-01395, 2019-05220). The study adheres to the Declaration of Helsinki principles, and findings will be made available in open access peer-reviewed publications.