Ajay Gulati
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Ajay Gulati has a total of 44 co-authors. Here are the top co-authors by number of shared publications:
- Niharika Dutta
Postgraduate Institute of Medical Education and Research, Chandigarh, India
5 shared publications - Pankaj Gupta
Department of Radiodiagnosis, Postgraduate Institute of Medical Education and Research, Chandigarh, India
5 shared publications - Naveen Kalra
Post Graduate Institute of Medical Education and Research, Chandigarh, India
3 shared publications - Usha Dutta
Department of Gastroenterology PGIMER Chandigarh India
3 shared publications - Ajay Duseja
Department of Hepatology, PGIMER, Chandigarh, India
2 shared publications - Arka De
Department of Hepatology, PGIMER, Chandigarh, India
2 shared publications
Latest Publications
- 24 November 2025
- Published by Springer Nature in Abdominal Radiology
- Vol. 51 (6) , 2906-2918
Deep learning for non-invasive detection of steatosis and fibrosis in MASLD: a multicenter study with a new fibroscan-labelled ultrasound dataset
Purpose This study aimed to develop and validate deep learning models for non-invasive assessment of hepatic steatosis and fibrosis using conventional B-mode ultrasound images, with Fibroscan-derived measurement as reference standard. Methods We utilized a multi-source approach comprising three distinct ultrasound datasets of patients with MASLD: a private clinical dataset (DS1, n = 111 patients, 1131 images), and two subsets from the public BEHSOF repository—Behbood Clinic subset (DS2, n = 95 patients, 1328 images) for training/validation, and Taleghani Hospital subset (DS3, n = 18 patients, 185 images) for external testing. Additionally, we validated on a temporally independent test set (DS4, n = 23 patients, 155 images). Deep learning models, including EfficientNet-B4 and Vision Transformers, were trained to classify steatosis (S0-S1 vs. S2-S3) and fibrosis (F0-F1 vs. F2-F4), respectively, using 5-fold cross-validation. Performance was assessed and compared with expert radiologists. Results For steatosis classification, our models achieved an AUROC of 0.83 (95% CI: 0.78–0.88) in cross-validation and 0.70 (95% CI: 0.43–0.93) on the external test set. For fibrosis, AUROC reached 0.86 (95% CI: 0.81–0.91) in cross-validation and 0.88 (95% CI: 0.71–1.00) on the external test set. On the temporally independent test set, performance was better with AUROCs of 0.82 (95% CI: 0.66–0.98) for steatosis and 0.91 (95% CI: 0.77–1.00) for fibrosis. The AI models consistently outperformed expert radiologists, whose interobserver agreement was moderate (κ = 0.58–0.65). Conclusion This study demonstrates that deep learning models can accurately identify steatosis and fibrosis from ultrasound images across diverse populations and clinical settings.
- 24 November 2025
- Published by Springer Nature in Abdominal Radiology
- Vol. 51 (6) , 2906-2918
Deep learning for non-invasive detection of steatosis and fibrosis in MASLD: a multicenter study with a new fibroscan-labelled ultrasound dataset
Purpose This study aimed to develop and validate deep learning models for non-invasive assessment of hepatic steatosis and fibrosis using conventional B-mode ultrasound images, with Fibroscan-derived measurement as reference standard. Methods We utilized a multi-source approach comprising three distinct ultrasound datasets of patients with MASLD: a private clinical dataset (DS1, n = 111 patients, 1131 images), and two subsets from the public BEHSOF repository—Behbood Clinic subset (DS2, n = 95 patients, 1328 images) for training/validation, and Taleghani Hospital subset (DS3, n = 18 patients, 185 images) for external testing. Additionally, we validated on a temporally independent test set (DS4, n = 23 patients, 155 images). Deep learning models, including EfficientNet-B4 and Vision Transformers, were trained to classify steatosis (S0-S1 vs. S2-S3) and fibrosis (F0-F1 vs. F2-F4), respectively, using 5-fold cross-validation. Performance was assessed and compared with expert radiologists. Results For steatosis classification, our models achieved an AUROC of 0.83 (95% CI: 0.78–0.88) in cross-validation and 0.70 (95% CI: 0.43–0.93) on the external test set. For fibrosis, AUROC reached 0.86 (95% CI: 0.81–0.91) in cross-validation and 0.88 (95% CI: 0.71–1.00) on the external test set. On the temporally independent test set, performance was better with AUROCs of 0.82 (95% CI: 0.66–0.98) for steatosis and 0.91 (95% CI: 0.77–1.00) for fibrosis. The AI models consistently outperformed expert radiologists, whose interobserver agreement was moderate (κ = 0.58–0.65). Conclusion This study demonstrates that deep learning models can accurately identify steatosis and fibrosis from ultrasound images across diverse populations and clinical settings.
- 6 October 2025
- Published by Springer Nature in Abdominal Radiology
- Vol. 51 (5) , 2353-2367
CT-based multimodal prediction model for chemotherapy response in gallbladder cancer: integration of imaging and clinical biomarkers—a prospective pilot study
Background Gallbladder carcinoma (GBC) is an aggressive malignancy with poor prognosis, often diagnosed at advanced stages. While chemotherapy forms the cornerstone of treatment for locally advanced and metastatic disease, reliable predictors of response remain elusive. This exploratory pilot study aimed to develop a computed tomography (CT)-based prediction model incorporating clinical and biochemical parameters to identify patients likely to respond to chemotherapy. Methods This prospective study enrolled consecutive treatment-naïve patients with cytologically/histologically proven locally advanced or metastatic GBC scheduled for chemotherapy between July 2023 and March 2025. All patients underwent biphasic contrast-enhanced CT within 14 days prior to chemotherapy initiation. Demographic variables, biochemical parameters, and tumor markers (CA 19 − 9, CEA) were systematically recorded. CT-based response assessment was performed using RECIST 1.1 criteria after two cycles of chemotherapy. Three exploratory prediction models were developed using logistic regression with appropriate variable selection methods and penalized regression to address small sample size. Results Of 25 patients (18 females, mean age 50.0 ± 11.6), 11 (44%) achieved partial response while 14 (56%) were non-responders (stable or progressive disease). Univariate analysis identified calculus presence (p = 0.002), largest liver metastasis size (p = 0.001), ascites (p = 0.008), age (p = 0.128), ALT (p = 0.241), and CEA levels (p = 0.123) as potential predictors. After correction for optimism using bootstrap validation, the CT-only model achieved area under the receiver operating characteristic curve (AUC) of 0.82 (95% CI: 0.64–0.92) with 80.0% accuracy, 92.3% sensitivity, and 62.5% specificity (Nagelkerke R²=0.272). The clinical-only model yielded optimism-corrected AUC of 0.74 (95% CI: 0.51–0.88) with 68.0% accuracy, 71.4% sensitivity, and 63.6% specificity (Nagelkerke R²=0.424). The combined model showed better performance with optimism-corrected AUC of 0.85 (95% CI: 0.67–0.95), 84.0% accuracy, 85.7% sensitivity, 81.8% specificity, and highest explained variance (Nagelkerke R²=0.662). Gallstone presence emerged as a potential predictor associated with treatment response across models, while larger liver lesions and elevated CEA showed trends that may be associated with poor response. Conclusions This preliminary hypothesis-generating study suggests that integration of baseline CT imaging features with clinical parameters may potentially allow prediction of chemotherapy response in GBC patients. The identified factors require external validation in larger multi-institutional cohorts before clinical implementation can be considered.
- 6 October 2025
- Published by Springer Nature in Abdominal Radiology
- Vol. 51 (5) , 2353-2367
CT-based multimodal prediction model for chemotherapy response in gallbladder cancer: integration of imaging and clinical biomarkers—a prospective pilot study
Background Gallbladder carcinoma (GBC) is an aggressive malignancy with poor prognosis, often diagnosed at advanced stages. While chemotherapy forms the cornerstone of treatment for locally advanced and metastatic disease, reliable predictors of response remain elusive. This exploratory pilot study aimed to develop a computed tomography (CT)-based prediction model incorporating clinical and biochemical parameters to identify patients likely to respond to chemotherapy. Methods This prospective study enrolled consecutive treatment-naïve patients with cytologically/histologically proven locally advanced or metastatic GBC scheduled for chemotherapy between July 2023 and March 2025. All patients underwent biphasic contrast-enhanced CT within 14 days prior to chemotherapy initiation. Demographic variables, biochemical parameters, and tumor markers (CA 19 − 9, CEA) were systematically recorded. CT-based response assessment was performed using RECIST 1.1 criteria after two cycles of chemotherapy. Three exploratory prediction models were developed using logistic regression with appropriate variable selection methods and penalized regression to address small sample size. Results Of 25 patients (18 females, mean age 50.0 ± 11.6), 11 (44%) achieved partial response while 14 (56%) were non-responders (stable or progressive disease). Univariate analysis identified calculus presence (p = 0.002), largest liver metastasis size (p = 0.001), ascites (p = 0.008), age (p = 0.128), ALT (p = 0.241), and CEA levels (p = 0.123) as potential predictors. After correction for optimism using bootstrap validation, the CT-only model achieved area under the receiver operating characteristic curve (AUC) of 0.82 (95% CI: 0.64–0.92) with 80.0% accuracy, 92.3% sensitivity, and 62.5% specificity (Nagelkerke R²=0.272). The clinical-only model yielded optimism-corrected AUC of 0.74 (95% CI: 0.51–0.88) with 68.0% accuracy, 71.4% sensitivity, and 63.6% specificity (Nagelkerke R²=0.424). The combined model showed better performance with optimism-corrected AUC of 0.85 (95% CI: 0.67–0.95), 84.0% accuracy, 85.7% sensitivity, 81.8% specificity, and highest explained variance (Nagelkerke R²=0.662). Gallstone presence emerged as a potential predictor associated with treatment response across models, while larger liver lesions and elevated CEA showed trends that may be associated with poor response. Conclusions This preliminary hypothesis-generating study suggests that integration of baseline CT imaging features with clinical parameters may potentially allow prediction of chemotherapy response in GBC patients. The identified factors require external validation in larger multi-institutional cohorts before clinical implementation can be considered.
- 19 August 2025
- Published by Springer Nature in Abdominal Radiology
- Vol. 51 (4) , 1770-1782
Advanced liver fibrosis detection using a two-stage deep learning approach on standard T2-weighted MRI
Objectives To develop and validate a deep learning model for automated detection of advanced liver fibrosis using standard T2-weighted MRI. Methods We utilized two datasets: the public CirrMRI600 + dataset (n = 374) containing T2-weighted MRI scans from patients with cirrhosis (n = 318) and healthy subjects (n = 56), and an in-house dataset of chronic liver disease patients (n = 187). A two-stage deep learning pipeline was developed: first, an automated liver segmentation model using nnU-Net architecture trained on CirrMRI600 + and then applied to segment livers in our in-house dataset; second, a Masked Attention ResNet classification model. For classification model training, patients with liver stiffness measurement (LSM) > 12 kPa were classified as advanced fibrosis (n = 104). In contrast, healthy subjects from CirrMRI600 + and patients with LSM ≤ 12 kPa were classified as non-advanced fibrosis (n = 116). Model validation was exclusively performed on a separate test set of 23 patients with histopathological confirmation of the degree of fibrosis (METAVIR ≥ F3 indicating advanced fibrosis). We additionally compared our two-stage approach with direct classification without segmentation, and evaluated alternative architectures including DenseNet121 and SwinTransformer. Results The liver segmentation model performed excellently on the test set (mean Dice score: 0.960 ± 0.009; IoU: 0.923 ± 0.016). On the pathologically confirmed independent test set (n = 23), our two-stage model achieved strong diagnostic performance (sensitivity: 0.778, specificity: 0.800, AUC: 0.811, accuracy: 0.783), significantly outperforming direct classification without segmentation (AUC: 0.743). Classification performance was highly dependent on segmentation quality, with cases having excellent segmentation (Score 1) showing higher accuracy (0.818) than those with poor segmentation (Score 3, accuracy: 0.625). Alternative architectures with masked attention showed comparable but slightly lower performance (DenseNet121: AUC 0.795; SwinTransformer: AUC 0.782). Conclusions Our fully automated deep learning pipeline effectively detects advanced liver fibrosis using standard non-contrast T2-weighted MRI, potentially offering a non-invasive alternative to current diagnostic approaches. The segmentation-first approach provides significant performance gains over direct classification.
- 19 August 2025
- Published by Springer Nature in Abdominal Radiology
- Vol. 51 (4) , 1770-1782
Advanced liver fibrosis detection using a two-stage deep learning approach on standard T2-weighted MRI
Objectives To develop and validate a deep learning model for automated detection of advanced liver fibrosis using standard T2-weighted MRI. Methods We utilized two datasets: the public CirrMRI600 + dataset (n = 374) containing T2-weighted MRI scans from patients with cirrhosis (n = 318) and healthy subjects (n = 56), and an in-house dataset of chronic liver disease patients (n = 187). A two-stage deep learning pipeline was developed: first, an automated liver segmentation model using nnU-Net architecture trained on CirrMRI600 + and then applied to segment livers in our in-house dataset; second, a Masked Attention ResNet classification model. For classification model training, patients with liver stiffness measurement (LSM) > 12 kPa were classified as advanced fibrosis (n = 104). In contrast, healthy subjects from CirrMRI600 + and patients with LSM ≤ 12 kPa were classified as non-advanced fibrosis (n = 116). Model validation was exclusively performed on a separate test set of 23 patients with histopathological confirmation of the degree of fibrosis (METAVIR ≥ F3 indicating advanced fibrosis). We additionally compared our two-stage approach with direct classification without segmentation, and evaluated alternative architectures including DenseNet121 and SwinTransformer. Results The liver segmentation model performed excellently on the test set (mean Dice score: 0.960 ± 0.009; IoU: 0.923 ± 0.016). On the pathologically confirmed independent test set (n = 23), our two-stage model achieved strong diagnostic performance (sensitivity: 0.778, specificity: 0.800, AUC: 0.811, accuracy: 0.783), significantly outperforming direct classification without segmentation (AUC: 0.743). Classification performance was highly dependent on segmentation quality, with cases having excellent segmentation (Score 1) showing higher accuracy (0.818) than those with poor segmentation (Score 3, accuracy: 0.625). Alternative architectures with masked attention showed comparable but slightly lower performance (DenseNet121: AUC 0.795; SwinTransformer: AUC 0.782). Conclusions Our fully automated deep learning pipeline effectively detects advanced liver fibrosis using standard non-contrast T2-weighted MRI, potentially offering a non-invasive alternative to current diagnostic approaches. The segmentation-first approach provides significant performance gains over direct classification.
- 21 July 2025
- Published by Springer Nature in Abdominal Radiology
- Vol. 51 (2) , 787-799
Role of MRI and abbreviated MRI in the evaluation and management of acute pancreatitis: a comprehensive review
Acute pancreatitis (AP) is one of the most common gastrointestinal emergencies, with pancreatic fluid collection (PFC) representing a frequent complication requiring precise evaluation for optimal management. While computed tomography (CT) remains the most widely utilized radiological investigation, it has limitations in correctly identifying and quantifying solid debris within PFCs and evaluating main pancreatic duct (MPD) integrity. Endoscopic ultrasound (EUS) offers improved visualization but is invasive and operator-dependent. Magnetic resonance imaging (MRI) provides superior characterization of PFCs, allowing excellent visualization of necrotic debris, liquid components, degree of encapsulation, and MPD anatomy without ionizing radiation—making it suitable for repeated evaluations. Abbreviated MRI (AMRI) protocols, involving acquisition of a few selected sequences while maintaining comprehensive evaluation capability, have emerged as promising tools for efficient assessment of AP patients. This review evaluates the utility of MRI and AMRI in characterizing PFCs and its implications for management decisions.
- 21 July 2025
- Published by Springer Nature in Abdominal Radiology
- Vol. 51 (2) , 787-799
Role of MRI and abbreviated MRI in the evaluation and management of acute pancreatitis: a comprehensive review
Acute pancreatitis (AP) is one of the most common gastrointestinal emergencies, with pancreatic fluid collection (PFC) representing a frequent complication requiring precise evaluation for optimal management. While computed tomography (CT) remains the most widely utilized radiological investigation, it has limitations in correctly identifying and quantifying solid debris within PFCs and evaluating main pancreatic duct (MPD) integrity. Endoscopic ultrasound (EUS) offers improved visualization but is invasive and operator-dependent. Magnetic resonance imaging (MRI) provides superior characterization of PFCs, allowing excellent visualization of necrotic debris, liquid components, degree of encapsulation, and MPD anatomy without ionizing radiation—making it suitable for repeated evaluations. Abbreviated MRI (AMRI) protocols, involving acquisition of a few selected sequences while maintaining comprehensive evaluation capability, have emerged as promising tools for efficient assessment of AP patients. This review evaluates the utility of MRI and AMRI in characterizing PFCs and its implications for management decisions.
- 3 July 2025
- Published by Springer Nature in Abdominal Radiology
- Vol. 51 (2) , 660-672
Radiological signature of HER2-Positive gallbladder cancer: analysis of CT features
Objective This study aimed to identify distinctive computed tomography (CT) features associated with human epidermal growth factor receptor 2 (HER2) status in gallbladder cancer (GBC) that could serve as noninvasive imaging biomarkers. Materials and methods This study included 213 patients with pathologically confirmed GBCs with availability of HER2 status (171 HER2-negative, 42 HER2-positive). Pre-treatment contrast-enhanced CT scans were evaluated by two radiologists blinded to HER2 status. Multivariate analysis was performed using logistic regression with L2 regularization. Model discrimination was assessed using receiver operating characteristic (ROC) analysis, and internal validation was performed using bootstrap resampling (1,000 iterations) to correct for optimism. Results HER2-positive tumors exhibited larger lymph nodes (1.93 ± 0.79 cm vs. 1.61 ± 0.61 cm, p = 0.015), less frequent gallstones (14.3% vs. 35.7%, p = 0.013), arterial phase hyperenhancement (20.0% vs. 44.1%, p = 0.026), mass-like morphology (35.7% vs. 55.6%, p = 0.033), and more frequent biliary compression by lymph nodes (19.0% vs. 4.1%, p = 0.002). Multivariate analysis identified biliary compression by lymph nodes as the strongest positive predictor of HER2 positivity [odds ratio (OR) 2.99, 95% CI: 1.25–7.04], while arterial phase hyperenhancement (OR 0.40, 95% CI: 0.19–0.75), gallstone presence (OR 0.40, 95% CI: 0.18–0.75), and mass-like morphology (OR 0.54, 95% CI: 0.29–0.95) were significant negative predictors. The model demonstrated good discrimination (area under the ROC curve 0.782) with sensitivity 75%, specificity 79.4%, and negative predictive value 96.2%. Conclusion HER2-positive GBCs display characteristic CT findings that can be utilized for noninvasive diagnosis with robust predictive performance.
- 3 July 2025
- Published by Springer Nature in Abdominal Radiology
- Vol. 51 (2) , 660-672
Radiological signature of HER2-Positive gallbladder cancer: analysis of CT features
Objective This study aimed to identify distinctive computed tomography (CT) features associated with human epidermal growth factor receptor 2 (HER2) status in gallbladder cancer (GBC) that could serve as noninvasive imaging biomarkers. Materials and methods This study included 213 patients with pathologically confirmed GBCs with availability of HER2 status (171 HER2-negative, 42 HER2-positive). Pre-treatment contrast-enhanced CT scans were evaluated by two radiologists blinded to HER2 status. Multivariate analysis was performed using logistic regression with L2 regularization. Model discrimination was assessed using receiver operating characteristic (ROC) analysis, and internal validation was performed using bootstrap resampling (1,000 iterations) to correct for optimism. Results HER2-positive tumors exhibited larger lymph nodes (1.93 ± 0.79 cm vs. 1.61 ± 0.61 cm, p = 0.015), less frequent gallstones (14.3% vs. 35.7%, p = 0.013), arterial phase hyperenhancement (20.0% vs. 44.1%, p = 0.026), mass-like morphology (35.7% vs. 55.6%, p = 0.033), and more frequent biliary compression by lymph nodes (19.0% vs. 4.1%, p = 0.002). Multivariate analysis identified biliary compression by lymph nodes as the strongest positive predictor of HER2 positivity [odds ratio (OR) 2.99, 95% CI: 1.25–7.04], while arterial phase hyperenhancement (OR 0.40, 95% CI: 0.19–0.75), gallstone presence (OR 0.40, 95% CI: 0.18–0.75), and mass-like morphology (OR 0.54, 95% CI: 0.29–0.95) were significant negative predictors. The model demonstrated good discrimination (area under the ROC curve 0.782) with sensitivity 75%, specificity 79.4%, and negative predictive value 96.2%. Conclusion HER2-positive GBCs display characteristic CT findings that can be utilized for noninvasive diagnosis with robust predictive performance.
- 17 May 2018
- Published by Elsevier in European Journal of Radiology
- Vol. 105, 25-31
Urolithiasis: Comparison of diagnostic performance of digital tomosynthesis and ultrasound. Which one to choose and when?
- 17 May 2018
- Published by Elsevier in European Journal of Radiology
- Vol. 105, 25-31
Urolithiasis: Comparison of diagnostic performance of digital tomosynthesis and ultrasound. Which one to choose and when?