Mohammadreza Elhaie
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Mohammadreza Elhaie has a total of 112 co-authors. Here are the top co-authors by number of shared publications:
Abolfazl Koozari
18 shared publications- Abolfazl Koozari
Department of Medical Physics, School of Medicine Ahvaz Jundishapur University of Medical Sciences Ahvaz Iran
5 shared publications Daryoush Shahbazi-Gahrouei
4 shared publicationsIraj Abedi
Department of Medical Physics, School of Medicine, Isfahan University of Medical Sciences, Isfahan, Iran
4 shared publicationsSomayeh Hajiahmadi
Department of Radiology, School of Medicine, Isfahan University of Medical Sciences, Isfahan, Iran
4 shared publicationsHossein Koohi
School of Medicine, Isfahan University of Medical Sciences, Isfahan, Iran
3 shared publications
Latest Publications
- Open Access
- 14 November 2026
- Published by Taylor & Francis in Journal of Obstetrics and Gynaecology
- Vol. 46 (1)
Endometrial ultrasound features associated with methotrexate success in tubal ectopic pregnancy: a prospective case series
Ectopic pregnancy, defined as implantation of a fertilised ovum outside the uterine cavity, remains an important gynaecological emergency and a major cause of first-trimester maternal morbidity and...
- 14 November 2026
- Published by Taylor & Francis in Journal of Obstetrics and Gynaecology
- Vol. 46 (1)
Open AccessEndometrial ultrasound features associated with methotrexate success in tubal ectopic pregnancy: a prospective case series
Ectopic pregnancy, defined as implantation of a fertilised ovum outside the uterine cavity, remains an important gynaecological emergency and a major cause of first-trimester maternal morbidity and...
- 9 November 2026
- Published by Wiley in Journal of Clinical Ultrasound
Transabdominal Ultrasound Measurement of Fetal Optic Chiasm Width Across Gestational Age in an Iranian Cohort
Introduction: Prenatal ultrasound assessment of the fetal optic chiasm (OC) may contribute to the evaluation of midline brain structures. Regional descriptive data on fetal OC width remain limited. Methods: This cross‐sectional study included 170 singleton fetuses examined between 17 weeks 6 days and 39 weeks 1 day at two tertiary hospitals. OC width was measured using the standardized three‐step transabdominal technique, and interocular distance (IOD) was recorded. Measurements were summarized descriptively across four gestational‐age groups. Associations were assessed using Pearson and Spearman correlations, linear regression, and Kruskal–Wallis tests. Results: Mean OC width was 6.46 ± 1.07 mm and was positively associated with gestational age (Pearson r = 0.582, p < 0.001). The regression slope was 0.157 mm per gestational week (95% CI, 0.123–0.190). Mean IOD was 16.46 ± 2.84 mm and was also associated with gestational age ( r = 0.630, p < 0.001). Both measurements differed across gestational‐age groups ( p < 0.001). The cohort was concentrated at 32–36 weeks 6 days (54.1%). Conclusion: OC width was positively associated with gestational age in this eligible ultrasound cohort. The findings provide descriptive biometric data, particularly for later gestation. Larger, more evenly distributed samples and reproducibility assessment are needed to establish clinically applicable reference values.
- 9 November 2026
- Published by Wiley in Journal of Clinical Ultrasound
Transabdominal Ultrasound Measurement of Fetal Optic Chiasm Width Across Gestational Age in an Iranian Cohort
Introduction: Prenatal ultrasound assessment of the fetal optic chiasm (OC) may contribute to the evaluation of midline brain structures. Regional descriptive data on fetal OC width remain limited. Methods: This cross‐sectional study included 170 singleton fetuses examined between 17 weeks 6 days and 39 weeks 1 day at two tertiary hospitals. OC width was measured using the standardized three‐step transabdominal technique, and interocular distance (IOD) was recorded. Measurements were summarized descriptively across four gestational‐age groups. Associations were assessed using Pearson and Spearman correlations, linear regression, and Kruskal–Wallis tests. Results: Mean OC width was 6.46 ± 1.07 mm and was positively associated with gestational age (Pearson r = 0.582, p < 0.001). The regression slope was 0.157 mm per gestational week (95% CI, 0.123–0.190). Mean IOD was 16.46 ± 2.84 mm and was also associated with gestational age ( r = 0.630, p < 0.001). Both measurements differed across gestational‐age groups ( p < 0.001). The cohort was concentrated at 32–36 weeks 6 days (54.1%). Conclusion: OC width was positively associated with gestational age in this eligible ultrasound cohort. The findings provide descriptive biometric data, particularly for later gestation. Larger, more evenly distributed samples and reproducibility assessment are needed to establish clinically applicable reference values.
- 3 November 2026
- Published by Elsevier in Journal of Medical Imaging and Radiation Sciences
- Vol. 58 (1)
Assessing the perspectives of radiation therapists on managing dementia in radiotherapy: An international qualitative study
- 3 November 2026
- Published by Elsevier in Journal of Medical Imaging and Radiation Sciences
- Vol. 58 (1)
Assessing the perspectives of radiation therapists on managing dementia in radiotherapy: An international qualitative study
- 22 October 2026
- Published by Wiley in Oral Science International
- Vol. 24 (1)
Cone Beam CT Evaluation of the Three‐Dimensional Relationship Between Impacted Mandibular Third Molars and the Mandibular Canal
Introduction: Impacted mandibular third molars are often removed, but proximity between mandibular third molar roots and the mandibular canal is clinically important because it may influence surgical planning and the assessment of anatomical relationships before extraction. Cone beam computed tomography (CBCT) enables three‐dimensional assessment of the mandibular canal–root relationship for surgical planning. Methods: This retrospective cross‐sectional study was approved by the Ethics Committee of Islamic Azad University, Isfahan (approval no. IR.IAU.KHUISF.REC.1399.0113). Included mandibular third molars were impacted/semi‐impacted with complete root development and radiographic proximity to the canal; inadequate images and cases with anomalies, lesions, fractures, tumors, infection, or other pathology were excluded. Canal position relative to roots (apical, buccal, lingual, inter‐radicular), impaction angulation (vertical, horizontal, mesioangular, distoangular), and canal–root relationship (no contact, cortical contact, penetration) were recorded. Associations were tested using chi‐square/Fisher's exact tests ( α = 0.05). Results: In 164 teeth, mesioangular impaction was most common (44.5%), followed by vertical (33.5%), horizontal (20.7%), and distoangular (1.2%). Canal position was most often apical (43.3%), then buccal (27.4%), lingual (23.8%), and between roots (5.5%). Impaction type was not significantly associated with canal position ( p = 0.153) or contact variables (contact type p = 0.328). Among 162 teeth with available contact classification, contact type differed by canal position ( p < 0.001): Penetration was highest with lingual (66.7%) and apical (54.9%) canals, while no contact predominated with buccal canals (48.9%). Conclusion: CBCT indicated that apical canal positioning and mesioangular impaction predominated. Canal position, rather than impaction type, showed the strongest association with radiographic penetration and should be considered during preoperative evaluation of mandibular canal–root relationships and surgical planning.
- 22 October 2026
- Published by Wiley in Oral Science International
- Vol. 24 (1)
Cone Beam CT Evaluation of the Three‐Dimensional Relationship Between Impacted Mandibular Third Molars and the Mandibular Canal
Introduction: Impacted mandibular third molars are often removed, but proximity between mandibular third molar roots and the mandibular canal is clinically important because it may influence surgical planning and the assessment of anatomical relationships before extraction. Cone beam computed tomography (CBCT) enables three‐dimensional assessment of the mandibular canal–root relationship for surgical planning. Methods: This retrospective cross‐sectional study was approved by the Ethics Committee of Islamic Azad University, Isfahan (approval no. IR.IAU.KHUISF.REC.1399.0113). Included mandibular third molars were impacted/semi‐impacted with complete root development and radiographic proximity to the canal; inadequate images and cases with anomalies, lesions, fractures, tumors, infection, or other pathology were excluded. Canal position relative to roots (apical, buccal, lingual, inter‐radicular), impaction angulation (vertical, horizontal, mesioangular, distoangular), and canal–root relationship (no contact, cortical contact, penetration) were recorded. Associations were tested using chi‐square/Fisher's exact tests ( α = 0.05). Results: In 164 teeth, mesioangular impaction was most common (44.5%), followed by vertical (33.5%), horizontal (20.7%), and distoangular (1.2%). Canal position was most often apical (43.3%), then buccal (27.4%), lingual (23.8%), and between roots (5.5%). Impaction type was not significantly associated with canal position ( p = 0.153) or contact variables (contact type p = 0.328). Among 162 teeth with available contact classification, contact type differed by canal position ( p < 0.001): Penetration was highest with lingual (66.7%) and apical (54.9%) canals, while no contact predominated with buccal canals (48.9%). Conclusion: CBCT indicated that apical canal positioning and mesioangular impaction predominated. Canal position, rather than impaction type, showed the strongest association with radiographic penetration and should be considered during preoperative evaluation of mandibular canal–root relationships and surgical planning.
- Open Access
- 13 October 2026
- Published by Elsevier in Intelligence-Based Medicine
- Vol. 16
Deep learning for cerebrovascular feature detection in susceptibility-weighted MRI: A systematic review and implications for clinical decision-making
- 13 October 2026
- Published by Elsevier in Intelligence-Based Medicine
- Vol. 16
Open AccessDeep learning for cerebrovascular feature detection in susceptibility-weighted MRI: A systematic review and implications for clinical decision-making
- Open Access
- 2 October 2026
- Published by Elsevier in SSRN Electronic Journal
Deep Learning-Derived Features Outperform Radiomics for Multi-Class BI-RADS Classification in Multiview Mammography: A Large-Scale Benchmark Using 15 Machine Learning Algorithms
IntroductionThis study compared radiomics-based handcrafted features and deep learning (DL)–derived features for multi-class BI-RADS (1–5) classification using multiview full-field digital mammography. The VinDr-Mammo dataset of 9,999 breast-level studies exhibited severe class imbalance (BI-RADS 1 ≈ 67%, BI-RADS 5 ≈ 1%).MethodsThe VinDr-Mammo dataset included 9,999 breast-level studies and showed severe class imbalance, with BI-RADS 1 accounting for approximately 67% of studies and BI-RADS 5 for approximately 1%. The original training and validation partitions were combined into a development cohort, while the independent test set was reserved exclusively for final evaluation. Radiomics features were extracted using PyRadiomics, whereas deep-learning-derived features were obtained from ImageNet-pretrained VGG16 and ResNet-18 networks used as fixed feature extractors without mammography-specific fine-tuning. Fifteen machine-learning classifiers were evaluated using stratified five-fold group cross-validation.ResultsBased on the prespecified primary endpoint of cross-validation macro F1-score, ResNet-18 features combined with SVC-RBF ranked first overall, achieving 0.305 ± 0.051. ResNet-18 with shrinkage LDA ranked second at 0.279 ± 0.045, followed by VGG16 with SVC-RBF at 0.270 ± 0.018. The best radiomics configuration was Extra Trees, with a cross-validation macro F1-score of 0.223 ± 0.030. The leading deep-learning-derived configurations therefore outperformed the leading radiomics configurations. Independent test-set performance was evaluated separately and was not used for model selection or ranking.ConclusionWithin the evaluated benchmark, deep-learning-derived feature representations extracted from ResNet-18 and VGG16 outperformed the evaluated radiomics features. These findings should not be generalized to all deep-learning architectures or all radiomics pipelines.
- 2 October 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessDeep Learning-Derived Features Outperform Radiomics for Multi-Class BI-RADS Classification in Multiview Mammography: A Large-Scale Benchmark Using 15 Machine Learning Algorithms
IntroductionThis study compared radiomics-based handcrafted features and deep learning (DL)–derived features for multi-class BI-RADS (1–5) classification using multiview full-field digital mammography. The VinDr-Mammo dataset of 9,999 breast-level studies exhibited severe class imbalance (BI-RADS 1 ≈ 67%, BI-RADS 5 ≈ 1%).MethodsThe VinDr-Mammo dataset included 9,999 breast-level studies and showed severe class imbalance, with BI-RADS 1 accounting for approximately 67% of studies and BI-RADS 5 for approximately 1%. The original training and validation partitions were combined into a development cohort, while the independent test set was reserved exclusively for final evaluation. Radiomics features were extracted using PyRadiomics, whereas deep-learning-derived features were obtained from ImageNet-pretrained VGG16 and ResNet-18 networks used as fixed feature extractors without mammography-specific fine-tuning. Fifteen machine-learning classifiers were evaluated using stratified five-fold group cross-validation.ResultsBased on the prespecified primary endpoint of cross-validation macro F1-score, ResNet-18 features combined with SVC-RBF ranked first overall, achieving 0.305 ± 0.051. ResNet-18 with shrinkage LDA ranked second at 0.279 ± 0.045, followed by VGG16 with SVC-RBF at 0.270 ± 0.018. The best radiomics configuration was Extra Trees, with a cross-validation macro F1-score of 0.223 ± 0.030. The leading deep-learning-derived configurations therefore outperformed the leading radiomics configurations. Independent test-set performance was evaluated separately and was not used for model selection or ranking.ConclusionWithin the evaluated benchmark, deep-learning-derived feature representations extracted from ResNet-18 and VGG16 outperformed the evaluated radiomics features. These findings should not be generalized to all deep-learning architectures or all radiomics pipelines.
- Open Access
- 27 September 2026
- Published by Wiley in Journal of Medical Radiation Sciences
Deep Learning and Machine Learning Algorithms for Cervical Cancer Segmentation on MRI : A Systematic Review
Introduction: Cervical cancer management relies heavily on magnetic resonance imaging (MRI) for staging and treatment planning; however, manual segmentation is time‐consuming and prone to variability. This systematic review aimed to evaluate the performance and clinical potential of machine learning (ML) and deep learning (DL) algorithms for automated cervical cancer segmentation on MRI. Methods: A systematic review guided by the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) (PROSPERO: CRD420251247441) was conducted by searching PubMed/MEDLINE, Embase, Scopus, Web of Science, IEEE Xplore, and Google Scholar until December 2025. Two reviewers independently screened studies, extracted data, and assessed risk of bias using the Quality Assessment of Diagnostic Accuracy Studies‐2 (QUADAS‐2). MRI‐based segmentation studies were primarily included, while computed tomography (CT)‐ and positron emission tomography (PET)‐based multimodal studies were analysed as complementary evidence. Due to substantial heterogeneity, findings were synthesized narratively, and certainty of evidence was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework. Results: Thirty studies were included, mainly evaluating segmentation of tumours, high‐risk clinical target volumes, and organs at risk. DL models, particularly U‐Net variants, convolutional neural networks, and transformer‐based architectures, demonstrated promising performance, with reported Dice similarity coefficients ranging from 0.60 to 0.93. However, performance varied according to target structure, tumour characteristics, and imaging protocols. Overall risk of bias showed “some concerns,” mainly due to patient selection, reference standards, and limited external validation. GRADE assessment indicated low certainty of evidence due to heterogeneity, methodological limitations, and imprecision. Conclusion: ML and DL algorithms show promising potential for automated cervical cancer segmentation on MRI, with potential benefits for efficiency and reproducibility. However, further multicenter studies with standardized protocols and external validation are required before routine clinical implementation. Trial Registration: PRISMA guidelines and was prospectively registered in PROSPERO: CRD420251247441
- 27 September 2026
- Published by Wiley in Journal of Medical Radiation Sciences
Open AccessDeep Learning and Machine Learning Algorithms for Cervical Cancer Segmentation on MRI : A Systematic Review
Introduction: Cervical cancer management relies heavily on magnetic resonance imaging (MRI) for staging and treatment planning; however, manual segmentation is time‐consuming and prone to variability. This systematic review aimed to evaluate the performance and clinical potential of machine learning (ML) and deep learning (DL) algorithms for automated cervical cancer segmentation on MRI. Methods: A systematic review guided by the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) (PROSPERO: CRD420251247441) was conducted by searching PubMed/MEDLINE, Embase, Scopus, Web of Science, IEEE Xplore, and Google Scholar until December 2025. Two reviewers independently screened studies, extracted data, and assessed risk of bias using the Quality Assessment of Diagnostic Accuracy Studies‐2 (QUADAS‐2). MRI‐based segmentation studies were primarily included, while computed tomography (CT)‐ and positron emission tomography (PET)‐based multimodal studies were analysed as complementary evidence. Due to substantial heterogeneity, findings were synthesized narratively, and certainty of evidence was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework. Results: Thirty studies were included, mainly evaluating segmentation of tumours, high‐risk clinical target volumes, and organs at risk. DL models, particularly U‐Net variants, convolutional neural networks, and transformer‐based architectures, demonstrated promising performance, with reported Dice similarity coefficients ranging from 0.60 to 0.93. However, performance varied according to target structure, tumour characteristics, and imaging protocols. Overall risk of bias showed “some concerns,” mainly due to patient selection, reference standards, and limited external validation. GRADE assessment indicated low certainty of evidence due to heterogeneity, methodological limitations, and imprecision. Conclusion: ML and DL algorithms show promising potential for automated cervical cancer segmentation on MRI, with potential benefits for efficiency and reproducibility. However, further multicenter studies with standardized protocols and external validation are required before routine clinical implementation. Trial Registration: PRISMA guidelines and was prospectively registered in PROSPERO: CRD420251247441
- Open Access
- 12 September 2026
- Published by Elsevier in European Journal of Obstetrics & Gynecology and Reproductive Biology: X
- Vol. 32
Twelve-month pelvic pain after ovarian-preserving laparoscopic hysterectomy: Secondary analysis of a randomized trial of hyaluronic acid gel
- 12 September 2026
- Published by Elsevier in European Journal of Obstetrics & Gynecology and Reproductive Biology: X
- Vol. 32
Open AccessTwelve-month pelvic pain after ovarian-preserving laparoscopic hysterectomy: Secondary analysis of a randomized trial of hyaluronic acid gel
- 3 September 2026
- Published by Wiley in Journal of Clinical Ultrasound
Ultrasonographic Evaluation of Cavum Septum Pellucidum Dimensions and Morphology in the Second Half of Pregnancy: A Prospective Cross‐Sectional Study
Objective: To determine normative ultrasonographic dimensions, morphology, and visualization rate of the cavum septum pellucidum (CSP) in structurally normal singleton fetuses between 19 and 40 weeks of gestation and to assess correlations with gestational age and fetal biometric parameters. Methods: This prospective cross‐sectional study enrolled 600 consecutive singleton pregnancies with normal fetal anatomy at Beheshti Hospital, Isfahan, Iran (2024–2025). Transabdominal ultrasound was performed in standard ISUOG‐recommended axial planes. CSP visualization status, morphology (rectangular, triangular, oval, square, or circular), length, and width were recorded together with biparietal diameter (BPD) and head circumference (HC). Data were analyzed using descriptive statistics, Pearson's and Spearman's correlations, and chi‐squared tests. Results: The CSP was visualized in 552 of 600 fetuses (92.0%). Mean CSP length was 8.27 ± 2.94 mm and mean width was 6.60 ± 2.74 mm, with rectangular morphology predominating (71.7%). Mean length increased modestly from 7.01 ± 2.62 mm (19–23 weeks) to 8.67 ± 2.98 mm (32–35 weeks) then stabilized, whereas width remained relatively stable. Weak positive correlations were observed between CSP length and gestational age ( r = 0.132, p = 0.002) and HC ( r = 0.156, p < 0.001); however, the small effect sizes indicate limited explanatory and clinical significance. No significant correlations were observed for width or morphology with biometric parameters. Nonvisualization (8.0%) occurred exclusively at or after 19 weeks. Conclusion: This study provides contemporary, population‐specific normative data for fetal CSP in the second and third trimesters in an Iranian cohort. CSP length shows modest gestation‐ and HC‐dependent growth, while width and morphology remain largely independent of standard biometry.
- 3 September 2026
- Published by Wiley in Journal of Clinical Ultrasound
Ultrasonographic Evaluation of Cavum Septum Pellucidum Dimensions and Morphology in the Second Half of Pregnancy: A Prospective Cross‐Sectional Study
Objective: To determine normative ultrasonographic dimensions, morphology, and visualization rate of the cavum septum pellucidum (CSP) in structurally normal singleton fetuses between 19 and 40 weeks of gestation and to assess correlations with gestational age and fetal biometric parameters. Methods: This prospective cross‐sectional study enrolled 600 consecutive singleton pregnancies with normal fetal anatomy at Beheshti Hospital, Isfahan, Iran (2024–2025). Transabdominal ultrasound was performed in standard ISUOG‐recommended axial planes. CSP visualization status, morphology (rectangular, triangular, oval, square, or circular), length, and width were recorded together with biparietal diameter (BPD) and head circumference (HC). Data were analyzed using descriptive statistics, Pearson's and Spearman's correlations, and chi‐squared tests. Results: The CSP was visualized in 552 of 600 fetuses (92.0%). Mean CSP length was 8.27 ± 2.94 mm and mean width was 6.60 ± 2.74 mm, with rectangular morphology predominating (71.7%). Mean length increased modestly from 7.01 ± 2.62 mm (19–23 weeks) to 8.67 ± 2.98 mm (32–35 weeks) then stabilized, whereas width remained relatively stable. Weak positive correlations were observed between CSP length and gestational age ( r = 0.132, p = 0.002) and HC ( r = 0.156, p < 0.001); however, the small effect sizes indicate limited explanatory and clinical significance. No significant correlations were observed for width or morphology with biometric parameters. Nonvisualization (8.0%) occurred exclusively at or after 19 weeks. Conclusion: This study provides contemporary, population‐specific normative data for fetal CSP in the second and third trimesters in an Iranian cohort. CSP length shows modest gestation‐ and HC‐dependent growth, while width and morphology remain largely independent of standard biometry.
- Open Access
- 21 August 2026
- Published by Elsevier in Cancer Treatment and Research Communications
- Vol. 49, 101382
Comparative evaluation of feature-selection strategies for machine learning–based histological grading of breast cancer using tumor and fibroglandular-tissue DCE-MRI features
- 21 August 2026
- Published by Elsevier in Cancer Treatment and Research Communications
- Vol. 49, 101382
Open AccessComparative evaluation of feature-selection strategies for machine learning–based histological grading of breast cancer using tumor and fibroglandular-tissue DCE-MRI features