Ali Mokhtari

Ali Mokhtari

Auto-generated profile, based on publicly available information

Ali Mokhtari

ARTORG Center for Biomedical Engineering Research, University of Bern, Bern, Switzerland

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

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

Representative Publications

View All Publications
10 June 2025 in Computers in Biology and Medicine
Verified
Journal Article
Open access
AMAli Mokhtari
Ali Mokhtari
Pascal CorsoPascal Corso
Pascal Corso
BJBernd Jung
Bernd Jung
LFLorenzo Ferrari
Lorenzo Ferrari
SZShaokai Zheng
Shaokai Zheng
Dominik ObristDominik Obrist
Dominik Obrist
Internal carotid artery (ICA) stenoses are a major source of stroke. In addition to traditional cardiovascular risk factors, hemodynamic parameters such as wall shear stress (WSS) are recognized as critical contributors to the progression and rupture of ICA plaques. This study evaluates the ability of 4D flow MRI to predict the velocity and WSS fields in three models of carotid bifurcations representing stenosis ratios of 0% (healthy), 40% (mildly stenosed), and 80% (severely stenosed) in the ICA. The stenosis ratio significantly impacts hemodynamics, increasing peak velocity, WSS, and flow disturbances. Compared to the healthy model, peak velocity rises by 38% in severe stenosis, generating a high-velocity jet and post-stenotic recirculation. Spatially averaged WSS increases by 21.5% in the mildly stenosed model and 99% in the severely stenosed model. Additionally, flow separation and vortex formation intensify as stenosis severity increases. A robust experimental setup with carotid bifurcation phantoms has been developed to eliminate the major sources of uncertainty in the boundary conditions. This enables a detailed quantitative comparison of 4D flow MRI measurements (obtained with a 7T scanner) against results from a CFD simulation. Moreover, the CFD results were validated against results obtained from Direct Numerical Simulations (DNS). Therefore, the CFD data could be used as a reliable high-fidelity reference for the flow within the carotid models. The comparison indicates that 4D flow MRI on a 7T scanner captures the major flow patterns with reasonable accuracy, though quantitative discrepancies increase with the stenosis ratio. Specifically, the peak velocity error rises from 6.1% in healthy arteries to 11.9% in severely stenosed arteries. Similarly, the error in spatially averaged WSS magnitude in the bifurcation region increases from 18% in healthy cases to 38% in severely stenosed cases. These findings underscore the need for validated approaches to accurately analyze complex flow patterns within carotid arteries, especially as even high-quality 4D flow MRI measurements using a 7T scanner with 0.5 mm resolution show diminishing accuracy in such challenging cases.

7 May 2026 in Journal of Open Source Software
Verified
Journal Article
Open access
PDF file
ASAnselm Walter Stark
Anselm Walter Stark
MIMarc Ilic
Marc Ilic
AMAli Mokhtari
Ali Mokhtari
PKPooya Mohammadi Kazaj
Pooya Mohammadi Kazaj
CGChristoph Gräni
Christoph Gräni
ISIsaac Shiri
Isaac Shiri

5 January 2026 in Diagnostics
Verified
Journal Article
Open access
PDF file
BSBasil Signer
Basil Signer
AMAli Mokhtari
Ali Mokhtari
SCSimone Cazzaniga
Simone Cazzaniga
FBFlurin Brand
Flurin Brand
GCGemma Caro
Gemma Caro
SJS. Morteza Seyed Jafari
S. Morteza Seyed Jafari
Background/Objectives: Trichoscopy is an important diagnostic tool for hair and scalp disorders, but it requires significant expertise. Publicly available large language models (LLMs) are becoming more popular among both physicians and patients, yet their usefulness in trichology is unknown. We aimed to evaluate the diagnostic accuracy of four publicly available LLMs when interpreting trichoscopic images, as well as to compare their performance with that of dermatology residents, board-certified dermatologists, and trichology experts. Method: In this prospective comparative study, a preprocessed set of trichoscopic images was assessed in an online image-based survey. To reduced recognition bias from public image repositories, all images were structurally transformed while preserving diagnostic features. Fifteen dermatologists (five residents, four board-certified dermatologists, six trichology experts) provided a suspected diagnosis (SD), and up to three the differential diagnoses (DD). Four LLMs (ChatGPT-4o, Claude Sonnet 4, Gemini 2.5 Flash, and Grok-3) evaluated the images under the same conditions. Results: The overall diagnostic accuracy among 15 dermatologists was 58.1% (95% CI, 53.0–63.0) for SD and 68.3% (95% CI, 63.4–72.8) for SD + DD. Experts significantly outperformed residents and board-certified dermatologists. AI models achieved an accuracy of 18.2% (95% CI, 11.8–26.9) for SD and 44.4% (95% CI, 35.0–54.3) for SD + DD. Gemini 2.5 Flash performed best, with an accuracy of 62.5% for SD + DD. Agreement among dermatologists increased with experience (AC1 up to 0.65 for experts), while agreement among AI models was moderate to good (AC1 up to 0.70). Agreement between AI models and dermatologists was only slight to fair (AC1 = 0.06 for SD and 0.21 for SD + DD). All human-AI differences were statistically significant (p < 0.001). Conclusions: In trichology, publicly available LLMs currently underperform compared to human experts, especially in providing a single correct diagnosis. These models require further development and specialized training before they can reliably assist with trichological diagnoses in routine care.

19 October 2026 in Annals of Biomedical Engineering
Verified
Journal Article
Open access
PDF file
AMAli Mokhtari
Ali Mokhtari
ABAriel Bergmann
Ariel Bergmann
CSChristoph Strecker
Christoph Strecker
JAJonathan Andrae
Jonathan Andrae
AHAndreas Harloff
Andreas Harloff
DODominik Obrist
Dominik Obrist
Purpose: Carotid geometry and local hemodynamics are closely associated with atherosclerosis, yet how anatomy and disturbed flow jointly mark disease presence and relate to pro-atherogenic flow in still-non-stenosed bifurcations remains unresolved. This study quantifies the cross-sectional association between three-dimensional carotid bifurcation geometry and pro-atherogenic flow and identifies anatomical biomarkers for non-invasive risk stratification. Methods: Geometric analyses included 346 carotid bifurcations, comprising 240 stenosed and 106 non-stenosed bifurcations. Patient-specific CFD was performed in 176 bifurcations: 101 stenosed and 75 non-stenosed. Eight geometric descriptors, including a new offset flow divider capturing bifurcation apex asymmetry, and four ICA bulb hemodynamic indices (TAWSS, OSI, RRT, BFI) were extracted; logistic and linear regression assessed the association of geometry with (i) stenosis presence and (ii) pro-atherogenic flow. Results: Bifurcation angle, ICA angle, tortuosity, and offset flow divider were significantly elevated in stenosed bifurcations (all $$p<0.001$$ p < 0.001 ); a four-feature geometric model showed moderate discrimination between stenosed and non-stenosed bifurcations (cross-validated $$\textrm{AUROC}=0.767$$ AUROC = 0.767 ). The post-stenotic expansion ratio was more strongly associated with local flow disturbance than the degree of luminal narrowing. In non-stenosed bifurcations, multivariable geometric models explained 49 to $$60\%$$ 60 % of the variance in the flow-normalized low-TAWSS, elevated-OSI, and elevated-RRT outcomes, with bulb area and area ratio providing the largest independent contributions, indicating that geometric variation is strongly associated with candidate pro-atherogenic flow patterns even in the absence of detectable luminal narrowing. Conclusions: Carotid geometry carries two complementary signals: markers of disease presence (bifurcation angle, ICA angle) and candidate anatomical substrates of pro-atherogenic flow in still-non-stenosed bifurcations (bulb area, area ratio). Together with the post-stenotic expansion ratio, these geometric features define a compact, image-derivable set of candidate biomarkers for future early carotid risk stratification.

26 December 2025 in Cosmetics
Verified
Journal Article
Open access
PDF file
AJAlexandra Junge
Alexandra Junge
AMAli Mokhtari
Ali Mokhtari
SCSimone Cazzaniga
Simone Cazzaniga
ABAshraf Mohamed Badawi
Ashraf Mohamed Badawi
FBFlurin Brand
Flurin Brand
SJS. Morteza Seyed Jafari
S. Morteza Seyed Jafari
Introduction: Artificial intelligence (AI) has developed into an increasingly important tool in dermatology. While new technologies integrated within laser devices are emerging, there is a lack of data on the applicability of publicly available AI models. Methods: The prospective study used an online questionnaire where participants evaluated diagnosis and treatment for 25 dermatological cases shown as pictures. The same questions were given to AI models: ChatGPT-4o, Claude Sonnet 4, Gemini 2.5 Flash, and Grok-3. Results: Dermatologists outperformed AI in diagnostic accuracy (suspected primary diagnosis-SD 75.6%) in pooled dermatologists vs. pooled AI (SD 57.0%), with laser specialists achieving the highest accuracy (SD 82.0%) and residents the lowest (SD 66.0%). There was a high heterogeneity across AI models. Gemini approached dermatologist performance (SD 72.0%), while Claude showed a low accuracy (SD 40.0%). While AI models reached near 100% accuracy in some classic/common diagnoses (e.g., acne, rosacea, spider angioma, infantile hemangioma), their accuracy dropped to near 0% on rare or context-dependent cases (e.g., blue rubber bleb nevus syndrome, angiosarcoma, hirsutism, cutaneous siderosis). Inter-rater agreement was high among laser experts in terms of diagnostic accuracy and treatment choice. Agreement between residents and AI models was highest for diagnostic accuracy and treatment choice, while it was lowest between experts and AI models. Conclusions: Before AI-based tools can be integrated into daily practice, particularly regarding diagnosis and appropriate laser treatment recommendations, specific supervised medical training of the AI model is necessary, as open-source platforms currently lack the ability to contextualize presented data.

7 July 2026 in Computers in Biology and Medicine
Verified
Journal Article
Open access
AMAli Mokhtari
Ali Mokhtari
ASAnselm W. Stark
Anselm W. Stark
DODominik Obrist
Dominik Obrist
MBMarius Reto Bigler
Marius Reto Bigler
SMStefano F. de Marchi
Stefano F. de Marchi
CGChristoph Gräni
Christoph Gräni
Traditional anatomical assessment of stenosis in right anomalous aortic origin of coronary arteries (R-AAOCA) may not capture the complex, eccentrically shaped lumen and its perimeter-dependent resistance, and may therefore underestimate flow limitation. R-AAOCA involves both fixed stenosis, similar to atherosclerotic plaques (assessable with invasive adenosine-derived fractional flow reserve, FFRAdenosine), and additional stress-induced dynamic stenosis; both components can be captured with dobutamine-derived FFR (FFRDobutamine). We hypothesized that fluid dynamics-informed geometric parameters provide a physiologically grounded description with enhanced discriminative value compared with conventional geometric parameters for predicting FFRAdenosine and FFRDobutamine. We performed a retrospective analysis of prospectively enrolled participants from the National Registry of Coronary Anomalies (NARCO) who underwent invasive FFRAdenosine and FFRDobutamine assessment. We calculated multiple CCTA-derived geometric parameters organized into two categories: (1) conventional geometric parameters including cross-sectional area (A), perimeter (P), minor and major axes at the ostium and intramural lumen area (ILA; i.e., the point of maximal stenosis), and effective diameter (Deff = (4A/π)); and conventional stenosis ratios including the area stenosis ratio (ASR) and effective diameter stenosis ratio (EDSR); (2) fluid dynamics-informed parameters comprising shape descriptors: including hydraulic diameter (Dh = 4A/P), elliptic ratio (ER = major/minor axis), circularity (C = 4πA/P2), and hydraulic diameter stenosis ratio (HDSR); and resistance-based indices including resistance index (RI = L × P4/A4), ostial angulation penalty (OAP = Aostial/Aref × [1 + tan(θ)]), and comprehensive stenosis score (CSS = RI × ER × [1 + sin(θ)]) integrating resistance, shape-dependent losses, and take-off angle (θ) effects. The invasive reference was hemodynamic relevance (i.e. FFR≤0.80) in FFRAdenosine and FFRDobutamine. We assessed Pearson correlation with FFR, univariable linear regression (R2), logistic regression with odds ratios (OR), and ROC AUC for classification (FFR≤0.80). A total of 81 R-AAOCA patients were included with a mean age of 52.3 ± 12.0 years and a mean body mass index of 27.0 ± 5.1 kg/m2; 27 patients (33.3%) were female. FFRAdenosine≤0.80 was observed in 5 (6.2%) patients and FFRDobutamine≤0.80 in 16 (19.8%) patients. The highest discriminatory power to predict abnormal FFRAdenosine was observed for the RI (AUC = 0.97, 95% CI 0.92-1.00), followed by the OAP (AUC = 0.97, 95% CI 0.92-1.00), ostial area (AUC = 0.96, 95% CI 0.90-1.00), CSS (AUC = 0.95, 95% CI 0.89-0.99), and ostial minor diameter (AUC = 0.95, 95% CI 0.90-0.99). The highest discriminatory power to predict abnormal FFRDobutamine was observed for ostial minor diameter (AUC = 0.85, 95% CI 0.74-0.94), followed by RI (AUC = 0.83, 95% CI 0.68-0.95) and ILA minor diameter (AUC = 0.81, 95% CI 0.69-0.92). RI explained 45% of the variance in FFRAdenosine and 43% of the variance in FFRDobutamine. At optimal thresholds, RI achieved 100% sensitivity and 95% specificity for predicting abnormal FFRAdenosine, whereas ostial minor diameter achieved 100% sensitivity and 57% specificity for predicting abnormal FFRDobutamine. In patients with R-AAOCA, CCTA-derived fluid dynamics-informed metrics showed high diagnostic performance for identifying hemodynamically relevant fixed stenosis (i.e., defined by abnormal FFRAdenosine), with discriminatory ability comparable to conventional geometric parameters. For combined fixed and stress-induced dynamic stenosis (i.e., defined by abnormal FFRDobutamine), both approaches performed similarly but with lower accuracy, underscoring the need for more advanced noninvasive modeling approaches that explicitly capture the stress-induced dynamic component.