SOFTWARE OPTIMIZATION & VALIDATION STUDIES
ASC Oral Presentation
Total Abnormal Cells in Bladder Cancer Screening: The Role of Artificial Intelligence-Based Logistic Models in Urine Cytology
Resources and News
Educational resources and news from AIxMed to keep you informed
and up-to-date on our AI empowered digital cytology solutions.
and up-to-date on our AI empowered digital cytology solutions.
.png)
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
AIxURO URINE CYTOLOGY CLINICAL STUDIES
May 2025
ICC 2025 Oral Presentation
AIxURO: Enhancing Diagnostic Accuracy in Bladder Caner Detection with AI-Assisted Urine Cytology
ICC 2025 Oral Presentation
AIxURO: Enhancing Diagnostic Accuracy in Bladder Caner Detection with AI-Assisted Urine Cytology
STUDY DESIGN:
RESULTS:
CONCLUSIONS:
AI SOFTWARE DEVELOPMENT
May 2025
ICC 2025 - Oral Presentation
Does AI performance in Thyroid Cancer Diagnosis Differ Between Single-Layer and Multi-Z-Stacked Imaging?
ICC 2025 - Oral Presentation
Does AI performance in Thyroid Cancer Diagnosis Differ Between Single-Layer and Multi-Z-Stacked Imaging?
STUDY DESIGN:
RESULTS:
CONCLUSIONS:
THE PARIS SYSTEM FOR URINE CYTOLOGY NUCLEAR TO CYTOPLASMIC RATIO
May 2025
Academic Article: Cancer Cytopathology
Does Artificial Intelligence Redefine Nuclear-to-Cytoplasmic Ratio Threshold for Diagnosing High-Grade Urothelial Carcinoma?
Academic Article: Cancer Cytopathology
Does Artificial Intelligence Redefine Nuclear-to-Cytoplasmic Ratio Threshold for Diagnosing High-Grade Urothelial Carcinoma?
STUDY DESIGN:
- 106 urine cytology slides (from 46 lower and 60 upper urinary tract HGUC/CIS cases) digitized and analyzed using AIxURO, an AI-based WSI tool.
- AI quantified atypical and suspicious urothelial cells, their nuclear-to-cytoplasmic(N/C) ratios, and nuclear areas.
- Morphologic data were correlated with biopsy-confirmed HGUC/CIS diagnoses. Statistical comparisons were conducted using Kruskal–Wallis tests.
RESULTS:
- Suspicious vs Atypical Cells (Total):
- Fewer suspicious cells detected (median 20.5 vs 242.0, p<.001)
- Higher median N/C ratio: 0.66 vs 0.58 (p<.001)
- Larger nuclear area: 102.3 µm² vs 85.7 µm² (p<.001)
- By Cytology Category (AUC, SHGUC, HGUC):
- Suspicious cells consistently had higher N/C ratios and nuclear areas than atypical cells.
- Median N/C ratio of suspicious cells: 0.66~0.67 and atypical cells: 0.58~0.59 across groups
- Median nuclear areas of suspicious cells: AUC 108.9 µm², SHGUC 99.2 µm², HGUC 101.6 µm² and atypical cells: AUC 88.5 µm², SHGUC 86.8 µm², HGUC 83.5 µm²
- Upper vs Lower Tract:
- No significant difference in N/C ratios
- Nuclear areas were smaller in UUT vs LUT cases (e.g., suspicious: 98.0 µm² vs 108.0 µm², p<.001)
- No significant difference in N/C ratios
- Biopsy Correlation:
- CIS had the largest nuclear areas among suspicious (116.3 µm²) and atypical (101.5 µm²) cells
- Cell number and N/C ratio did not differ significantly across CIS, CIS-HGUC, and HGUC biopsy categories
CONCLUSIONS:
AIxURO provides objective measurement of N/C ratio and nuclear area in urine cytology, challenging the current TPS threshold (>0.7) for diagnosing HGUC. The study suggests that a revised N/C ratio cutoff of 0.66 may be more appropriate for SHGUC/HGUC categorization using AI. Findings also support the use of consistent thresholds across upper and lower urinary tract cases. Nuclear area offers additional discriminative value, particularly for differentiating CIS from HGUC.
November 2024
Total Abnormal Cells in Bladder Cancer Screening: The Role of Artificial Intelligence-Based Logistic Models in Urine Cytology
ASC 2024 Oral Presentation / Hwang YT et al. / National Taipei University
STUDY DESIGN:
An AI-based logistic regression model was developed to determine the optimal number of abnormal urothelial cells (HGUC, SGHUC, or AUC) required to accurately predict the presence of bladder cancer. The study goal was to evaluate the accuracy and effectiveness of the logistic models in predicting bladder cancer from clinical urine cytology test datasets.
- 2025 cytology slide images (471 positives, 1334 negatives) were analyzed by the AI algorithm
- Abnormal cells were categorized as suspicious (SHGUC+) or atypical (AUC)
- Preparation types: Cytospin, ThinPrep, BD CytoRich, and TP-Urocyte
- Cell numbers were standardized by preparation sample area (mm2)
- Calculations:
- Suspicious cell numbers
- Log of suspicious cell numbers
- Atypical cell numbers
- Log of atypical cell numbers
- Performance Metrics: Accuracy, sensitivity, specificity, ROC curve
RESULTS:
- Logistic model- suspicious cells: 91.69 Accuracy, 64.29 Sensitivity, 95.44 Specificity
- Log Suspicious cells: 65.33 Accuracy, 95.24 Sensitivity, 61.24 Specificity
- Logistic model-atypical cells: 94.27 Accuracy, 69.05 Sensitivity, 97.72 Specificity
- Log Atypical cells: 76.22 Accuracy, 88.1 Sensitivity, 74.59 Specificity
- Cell cut-off value per preparation type:some text
- Suspicious cutoff value: 0.189 (Cytospin 5 cells, TP 59, CytoRich 25, UroCyte 10
- Atypical cutoff value: 1.823 (Cytospin 52 cells, TP 573, 242 CytoRich, 49 UroCyte
CONCLUSIONS:
Suspicious cell cut-off values support TPS criteria, with more than 5 suspicious cells categorized as SHGUC or higher indicating high likelihood of HGUC
Atypical cell cut-off values can aid in identifying potential false-negative cases when suspicious cell numbers are below cut-off values but atypical numbers are higher.
November 2024
THE PARIS SYSTEM FOR URINE CYTOLOGY NUCLEAR TO CYTOPLASMIC RATIO
ASC Poster Presentation
How does artificial intelligence redefine N:C ratio thresholds for diagnosing high-grade urothelial carcinoma?
How does artificial intelligence redefine N:C ratio thresholds for diagnosing high-grade urothelial carcinoma?
ASC 2024 Poster Presentation Hang JF, Yang WL, Yeh CH et al. TVGHTE, Natl Tung University, Show Chwan Mem Hospital, Tungs MetroHarbor Hospital N:C ratio pilot study_2024
STUDY DESIGN:
- 109 urine cytology cytospin cases with surgically-confirmed HGUC, HGUC-CIS, “positive” or CIS within a 6 month pre-or post-urine collection window
- Digitally analyzed by AIxURO AI-deep learning software for N/C ratio and nuclear area for the cytologic specimen diagnoses of AUC, SHGUC, and HGUC
- Comparison of the cytologic and surgical diagnoses by mean cell numbers per slide, mean N/C ratio, and mean nuclear area (µm2)
RESULTS:
- Cell Quantification: AIxURO categorized fewer suspicious cells (mean 160.0) compared to atypical cells (mean 929.6)
- N/C Ratio and Nuclear Area: Suspicious cells had higher mean N/C ratio (0.66 vs 0.58) and higher mean nuclear area (107.2 vs 66,9 µm2) than atypical cells
- Mean N/C for suspicious cells across all biopsies = 0.65 to 0.68
- Larger nuclear areas were also noted in SHGUC (13.9% difference) and HGUC (24.4% difference)
- Carcinoma in situ (CIS) had the largest mean nuclear area for both suspicious (117.3 µm2) and atypical (101.1 µm2) cells
CONCLUSIONS:
A lower mean N/C ratio for suspicious cells as an indicator of biopsy-proven HGUC using TPS should be considered, especially for AI-assisted urine cytology software programs. AIxURO categorized suspicious cells have a higher nuclear area than categorized atypical cells. Very large nuclear areas may be an indicator of CIS.
November 2024
WHOLE SLIDE IMAGE SCANNER STUDIES
Pathology Visions Poster Presentation
Do different digital imagers used with the AIxURO system impact performance in urine cytology?
Do different digital imagers used with the AIxURO system impact performance in urine cytology?
DPA 2024 Poster Presentation / Abstract Yang WL, Yang WC, Huang SM, et al. Show Chwan Hospital and AIxMED (pilot studies 2 & 3)
STUDY DESIGN:
- 183 urine cytology slides (140 CytoRich, 43 SurePath), with expert panel consensus diagnoses as NHGUC (83), AUC (45), SHGUC (27), and HGUC (28)
- Slides scanned by Leica AT2 and Hamamatsu S360 digital scanners with images inferred by the AI algorithm (AIxURO), using The Paris System (TPS) 2.0 criteria
- AIxURO separated abnormal cells into “suspicious” and “atypical” categories for investigator review
- 3-Arm Study:
- 6 reviewers (1 cytopathologist and 5 cytologists) interpreted the slides using conventional microscopy only (Arm 1), Leica whole slide image (WSI) with AIxURO AI-assistance (Arm 2), and Hamamatsu WSI with AIxURO AI-assistance (Arm 3) with a 2-week wash-out period between reviews
RESULTS:
3294 total reviews (microscopy + WSI + AIxURO); binary distribution as positive (AUC, SHGUC, or HGUC) or negative (NHGUC)
- Sensitivity and Specificity of AIxURO with Leica: 85.0% and 90.0%
- Sensitivity and Specificity of AIxURO with Hamamatsu: 82.9% and 89.6%
- Sensitivity and Specificity of microscopy only: 83.7% and 89.4%
- Total Time for review of AIxURO with Leica: 37.4 seconds
- Total Time for Review of AIxURO with Hamamatsu: 53.1 seconds
- Total Time for Review of microscopy only: 82.0 seconds
- Reviewers (3) experienced with the AIxURO system had higher sensitivity (91.3% vs 78.7%) but lower specificity (87.7% vs 92.3%) and spent less time (30.6 sec vs 44.4 sec) than the 3 without experience
CONCLUSIONS:
Use of AI-assisted software (AIxURO) to classify abnormal urothelial cells into atypical or suspicious categories improved the overall sensitivity and specificity for a binary interpretation over conventional microscopy. Additionally, AI-assistance significantly reduced reporting time by 52.4% (AIxURO + Leica AT2) and 35.2% (AIxURO + Hamamatsu) compared to microscopy. Experienced users of AIxURO showed better overall performance, suggesting that training and experience with AI-assistance will improve performance.
November 2024
AIxURO URINE CYTOLOGY CLINICAL STUDIES
Pathology Visions Poster Presentation
Transforming bladder cancer diagnosis via AI-aided digital cytology integration in clinical workflow
Transforming bladder cancer diagnosis via AI-aided digital cytology integration in clinical workflow
DPA 2024 Poster Presentation / Abstract Sigfred Lajara, MD; Samer Khader, MD; Rajiv Dhir, MD; Daniel Geisler, MD; Jacqueline Cuda, SCT; Caroline Horne, SCT UPMC pilot study
STUDY DESIGN:
- 200 ThinPrep urine cytology slides (100 NHGUC, 35 AUC, 32 SHGUC, 33 HGUC) with diagnostic confirmation by 3 expert pathologists were digitized to WSIs using Mikroscan SLxCyto and analyzed by an AI-assisted program, AIxURO
- 1 cytopathologist (CP) and 2 cytologists (CT) reviewed urine slides or AI-assisted images with a 2-week washout period between reviews
- Arm 1: Microscopic review
- Arm 2: AI-assisted software (AIxURO) review
- Performance Metrics: Comparison of 2 diagnostic positive thresholds: AUC+ (AUC, SHGUC, and HGUC) and SHGUC+ (SHGUC and HGUC) and total time for interpretation
RESULTS:
- AI-assisted AIxURO showed higher sensitivity (85% vs 79.3%) but lower specificity (92% vs 98%) than microscopy overall, using AUC+ as the positive threshold
- AIxURO showed lower sensitivity (74.9% vs 76.9%) and specificity (96% vs 97.5%) than microscopy using SHGUC+ as the positive threshold.
- Review Time: AIxURO markedly reduced review time (37.4 s) compared to microscopy (102.6 s) overall. CTs spent almost double the time for microscopy (121.6 s) compared to the CP (64.6 s).
CONCLUSIONS:
AIxURO showed a 5.7% increase in sensitivity and 8.6% decrease in specificity compared to microscopy, suggesting that while AIxURO helps to ID more positive cases, it results in higher false positives. At SHGUC+ threshold, AIxURO showed a 2% decrease in sensitivity and 1.5% decrease in specificity, suggesting that reviewers diagnose more cases as AUC than with microscopy. AIxURO saves reviewers 63.5% of evaluation time compared to microscopy, at more consistent evaluation times. CTs spend substantially less evaluation time than CP using AIxURO.
November 2024
AIxURO URINE CYTOLOGY CLINICAL STUDIES
Pathology Visions Poster Presentation
Revolutionizing bladder cancer diagnosis: Insights on urine cytology from a digital pathology lab
Revolutionizing bladder cancer diagnosis: Insights on urine cytology from a digital pathology lab
DPA 2024 Poster Presentation / Abstract CorePlus pilot study
STUDY DESIGN:
- 100 CytoRich urine cytology slides with 3 expert consensus interpretations as ground truth (68 NHGUC, 11 AUC, 7 SHGUC, 14 HGUC) scanned with Mikroscan SLxCyto to form WSI, which was in turn analyzed with deep learning AI model (AIxURO) trained on The Paris System criteria for urine cytology reporting
- 1 cytopathologist and 2 cytologists microscopically reviewed the glass slide and digitally reviewed the WSI and AIxURO images, with a 2-week washout between each, resulting in 300 diagnostic pairs per observer
- Diagnostic pairs were compared with the ground truth; total time to report recorded; and sensitivity, specificity recorded per observer
RESULTS:
- Sensitivity: For a binary diagnosis (negative = NHGUC; positive = AUC, SHGUC, HGUC), AIxURO showed higher sensitivity overall (88.5% vs 86.5%) than microscopy for all observers
- CT1: 84.4% vs. 87.5%
- CT2: 90.6% vs. 90.6%
- CP: 90.6% vs. 81.3%
- Specificity was lower overall for AIxURO (93.6% vs. 97.6%) compared to microscopy
- CT1: 91.2% vs. 94.1%
- CT2: 98.5% vs. 98.5%
- CP: 91.2% vs. 100.0%
- Reporting time: AIxURO substantially reduced mean reporting time for all observers compared to microscopy (13.6 seconds vs 83.3 seconds); this was especially pronounced for negative cases (7.7 seconds vs 72.3 seconds
CONCLUSIONS:
AI-assisted software (AIxURO) substantially reduces overall interpretation and reporting time by 83.3% compared to conventional microscopy. AIxURO also improves sensitivity for a binary (positive vs negative) diagnosis but is less specific than microscopic interpretation.
November 2024
AI SOFTWARE DEVELOPMENT
Academic Article: In Vivo
An artificial intelligence-assisted diagnostic system improves upper urine tract cytology diagnosis
An artificial intelligence-assisted diagnostic system improves upper urine tract cytology diagnosis
In Vivo 2024; 38(x): published ahead of print; doi:10.21873/invivo.11xxx Chang KY, Yang CS, Lai HY et al. VGHTC; Nat’l Chung Hsing University; Chung Shan Medical University
STUDY DESIGN:
- Retrospective cohort study
- 185 upper tract urine cytology slides (168 NHGUC, 14 AUC, 2 SHGUC, 1 HGUC) with one expert cytopathologist (CP) and one experienced cytologist (CT) confirmation of interpretation; discrepancies in diagnosis were resolved by multiheaded microscopy review by expert panel
- Digitized using Aperio AT2 scanner (Leica Biosystems) at 40X and single Z-plane
- Deep-learning training
RESULTS:
- Cases ranked by AI-driven software into low risk (N/C 0.5 to 0.7) or high risk (N/C > 0.7)
- 37 discrepant results after AI analysis (AIxURO)
- Discrepancies (AIxURO vs conventional):
- Cytopathologist:
- Overcalled 1 NHGUC as SHGUC
- Undercalled 2 AUC as NHGUC
- Cytologist:
- Overcalled 3 NHGUC as AUC and 2 AUC as SHGUC
- Undercalled 9 AUC as NHGUC, and 1 SHGUC
- o NHGUC (20 of 168; 11.9% discrepancy rate)
- Cytopathologist:
CONCLUSIONS:
Diagnostic accuracy is achieved in at least 85.7% for atypical and suspicious cells in the AUC and above categories, with AUC showing the least concordance (21.4% accuracy). The use of AI-assistance markedly reduced the miscall rate for the CP but not the CT compared to reported misdiagnosis rates as high as 27.6% (57 million cases in China).
July 2024
AIxURO URINE CYTOLOGY CLINICAL STUDIES
Academic Article: Cancer Cytopathology
Evaluating artificial intelligence–enhanced digital urine cytology for bladder cancer diagnosis
Evaluating artificial intelligence–enhanced digital urine cytology for bladder cancer diagnosis
STUDY DESIGN:
- 116 urine (76 cytospin; 40 CytoRich) cytology slides with 3-armed microscopy, corresponding WSI and AI-digital (AIxURO) review by 1 experienced cytopathologist and 2 cytologists
- Performance metrics calculated for each arm included binary (negative vs positive) diagnosis, inter-and intra-observer agreement, and screening time
RESULTS:
- Atypical Urothelial Cells (AUC): AIxURO improved diagnostic sensitivity (from 25-30.6% to 63.9%), PPV (from 21.6-24.3% to 31.1%), and NPV (91.3-19.6%) to 95.3%)
- Suspicious for High-Grade Urothelial Carcinoma (SHGUC): AIxURO improved sensitivity (from 15.2-27.3% to 33.3%), PPV (from 31.3-47.4% to 61.1%), and NPV (from 91.6%-92.7% to 93.3%).
- Binary Diagnosis (Negative vs Positive [AUC, SHGUC, or HGUC]): AIxURO improved sensitivity (from 77.8-82.2% to 90.0%) and NPV (from 91.7-93.4% to 95.8%)
- Interobserver agreement: Moderate concurrence across all methods of evaluation (ĸ = 0.57-0.61); cytopathologist showed the highest intraobserver agreement (ĸ = 0.75-0.88)
- Screening time: AIxURO significantly reduced screening time compared to conventional microscopy for all observers (by 52.3% to 83.2%); AUC case
CONCLUSIONS:
The most significant finding is the marked reduction in screening time for AI-enhancement (AIxURO) compared with conventional microscopy (up to 83% less time required). Implementation of AI enhancement (AIxURO) for urine cytology interpretation improves diagnostic sensitivity, PPV and NPV for AUC and SHGUC, but not HGUC.
AIxURO improves the sensitivity and NPV of a binary interpretation of negative or positive. The interobserver agreement across all methods of review (microscopy, WSI and AIxURO) is moderate (ĸ = 0.57-0.61) with the cytopathologist showing the highest intra-observer agreement.
June 2024
AIxURO URINE CYTOLOGY CLINICAL STUDIES
ECDP Poster Presentation
A comparative study of artificial intelligence integration in bladder cancer screening: Enhancing accuracy and efficiency
A comparative study of artificial intelligence integration in bladder cancer screening: Enhancing accuracy and efficiency
ECDP 2024 Poster Presentation / Abstract Show Chwan Hospital, clinical pilot study 2
STUDY DESIGN:
- 200 urine cytology slides (100 positive, 100 negative) were scanned to create whole slide images (WSI) that were analyzed by an artificial intelligence (AI)-assisted software program (AIxURO) to detect and quantify characteristics of abnormal urothelial cells
- Three study arms, each performed by 3 reviewers (1 cytopathologist, 2 cytologists) rendering the Paris System (TPS) 2.0 interpretation (2-week washout period between each arm):
- ARM 1- Glass slide microscopic interpretation
- ARM 2- Whole slide image interpretation
- ARM 3- WSI with AI-assisted interpretation (AIxURO)
- Performance Metrics: Total screening/reporting time, sensitivity and specificity compared to the ground truth diagnosis
RESULTS:
- Average screening and reporting time was significantly reduced by 25.8%-58.7% (p < 0.05)
- Microscopy only and AI-assisted (AIxURO) outperformed WSI-only review in both sensitivity and specificity
- AIxURO was slightly less sensitive than microscopy (66.0 - 87.0% vs. 86.0 -89.0%) but more specific (89.0 – 95.0% vs. 81.0 – 88.0%).
- Use of AIxURO reclassified some ground-truth diagnoses from HGUC or SHGUC to AUC or NHGUC.
CONCLUSIONS:
The use of an AI-assisted software platform (AIxURO) for detection of bladder carcinoma improves overall specificity in comparison with microscopic glass slide or whole slide imaging review, while significantly reducing screening and reporting time.
June 2024
SOFTWARE OPTIMIZATION & VALIDATION STUDIES
ECDP Poster Presentation
Developing an optimal artificial intelligence-based logistic model for automated bladder cancer screening in digital urine cytology
Developing an optimal artificial intelligence-based logistic model for automated bladder cancer screening in digital urine cytology
ECDP 2024 Poster Presentation Receiver-Operator Curve (ROC) study to assess AIxURO logistic model’s accuracy to predict bladder cancer
STUDY DESIGN:
- 1856 urine cytology cases (1466 negative and 390 positive)- AI training set
- 169 urine cytology cases (88 negative, 81 positive)- Validation set
- AIxURO classifies abnormal urothelial cells into 2 categories based on The Paris System 2.0: “Suspicious” (SHGUC or HGUC) and “Atypical “(AUC), with the final interpretation deferred to a pathologist
- Logistic regression performed to predict presence of cancer, including variables such as total # suspicious cells, total # atypical cells, and predictive accuracy using sensitivity and specificity
RESULTS:
- Optimal performance of the training set (based on the total number of atypical cells) was 10 cells (cytospin) and 49 cells (CytoRich)
- Training Set Sensitivity and Specificity: 75.9% and 73.0%
- Validation Set Sensitivity and Specificity: 75.3% and 87.5%
CONCLUSIONS:
The logistic model supports the optimal cut-off values of at least 10 cells (cytospin) and 49 cells (CytoRich) for the number of atypical cells required for a high concordance with bladder cancer as the final outcome.
Upcoming Events
Upcoming event
International Congress of Cytology
11
-
15
May 2025
Florence Italy
May
11
-
15
,
2025
Latest news
November 2024
American Society of Cytopathology (ASC)
Dr. Barbara Crothers is the recipient of the ASC’s highest award- 2024 Papanicolaou AwardApril 2024
