Shobhit Mathur
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Shobhit Mathur has a total of 239 co-authors. Here are the top co-authors by number of shared publications:
- Aditya Bharatha
Division of Neuroradiology, Department of Medical Imaging, Unity Health Toronto–St. Michael’s Hospital, University of Toronto, 30 Bond St, Toronto, ON, Canada M5B 1W8
7 shared publications Yingming Amy Chen
Department of Medical Imaging, St. Michael's Hospital, Toronto, Canada
6 shared publicationsAditya Bharatha
Department of Medical Imaging, St. Michael’s Hospital, University of Toronto, ON, Canada
5 shared publications- Errol Colak
Department of Medical Imaging, St. Michael’s Hospital, Unity Health Toronto, Toronto, Canada
4 shared publications - Hui Ming Lin
Department of Medical Imaging, St. Michael’s Hospital, Unity Health Toronto, University of Toronto, Toronto, Canada
4 shared publications - Noah Ditkofsky
Department of Medical Imaging, St. Michael’s Hospital, Unity Health Toronto, University of Toronto, Toronto, Canada
4 shared publications
Latest Publications
- Open Access
- 24 August 2026
- Published by Springer Nature in Neuroradiology
- Vol. 68 (9) , 3083-3091
Brain cone-beam CT image quality improvement using a deep-learning-based denoising method: a multicenter retrospective study
Purpose: Deep learning (DL) denoising may improve cone-beam CT (CBCT) image quality for point-of-care stroke assessment in the interventional suite. The purpose of this study was to evaluate the impact of a DL-based denoising algorithm on objective and subjective image quality in brain CBCT using both standard circular and advanced dual-axis trajectories. Methods: We retrospectively analyzed 20 noncontrast brain CBCT acquisitions (Karolinska: 10 standard circular; St Michael's: 10 dual-axis). A DL-based denoising algorithm was applied at three strengths (Minimal, Medium, High) and compared to standard images with no additional denoising. Objective metrics (noise, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), artifact indices) were measured using standardized ROIs. Six experts rated subjective image quality on 5-point Likert scales. Paired tests with Bonferroni correction were used for comparisons. Results: High-level DL denoising significantly improved all objective metrics (noise, SNR, CNR, artifact indices) for both thin and thick slices (all p <.001). It doubled gray–white matter CNR (thin slices: 2.31 vs. 1.08), reduced noise, and improved subcalvarial and posterior fossa artifact indices. Subjectively, high-level denoising yielded higher median ratings for noise, texture, sharpness, brain parenchyma visualization, CSF spaces, and confidence in assessing ischemia and hemorrhage (all p <.001). Improvements were consistent for both acquisition techniques, and perceived artifact severity did not differ ( p >.99). Inter-reader agreement was substantial. Conclusion: The DL-based denoising algorithm significantly improved objective and subjective brain CBCT image-quality; no difference in perceived artifact severity was detected. These findings support further evaluation of deep learning-enhanced CBCT denoising for brain imaging in the interventional suite.
- 24 August 2026
- Published by Springer Nature in Neuroradiology
- Vol. 68 (9) , 3083-3091
Open AccessBrain cone-beam CT image quality improvement using a deep-learning-based denoising method: a multicenter retrospective study
Purpose: Deep learning (DL) denoising may improve cone-beam CT (CBCT) image quality for point-of-care stroke assessment in the interventional suite. The purpose of this study was to evaluate the impact of a DL-based denoising algorithm on objective and subjective image quality in brain CBCT using both standard circular and advanced dual-axis trajectories. Methods: We retrospectively analyzed 20 noncontrast brain CBCT acquisitions (Karolinska: 10 standard circular; St Michael's: 10 dual-axis). A DL-based denoising algorithm was applied at three strengths (Minimal, Medium, High) and compared to standard images with no additional denoising. Objective metrics (noise, signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), artifact indices) were measured using standardized ROIs. Six experts rated subjective image quality on 5-point Likert scales. Paired tests with Bonferroni correction were used for comparisons. Results: High-level DL denoising significantly improved all objective metrics (noise, SNR, CNR, artifact indices) for both thin and thick slices (all p <.001). It doubled gray–white matter CNR (thin slices: 2.31 vs. 1.08), reduced noise, and improved subcalvarial and posterior fossa artifact indices. Subjectively, high-level denoising yielded higher median ratings for noise, texture, sharpness, brain parenchyma visualization, CSF spaces, and confidence in assessing ischemia and hemorrhage (all p <.001). Improvements were consistent for both acquisition techniques, and perceived artifact severity did not differ ( p >.99). Inter-reader agreement was substantial. Conclusion: The DL-based denoising algorithm significantly improved objective and subjective brain CBCT image-quality; no difference in perceived artifact severity was detected. These findings support further evaluation of deep learning-enhanced CBCT denoising for brain imaging in the interventional suite.
- 1 April 2026
- Published by Radiological Society of North America (RSNA) in RadioGraphics
- Vol. 46 (4) , e250153
Imaging the Atypical and Equivocal Appendix
This slide presentation reviews challenging CT and US findings of acute appendicitis, its mimics and atypical manifestations, and appendiceal neoplasms and provides radiologists with practical strategies to support accurate diagnosis.
- 1 April 2026
- Published by Radiological Society of North America (RSNA) in RadioGraphics
- Vol. 46 (4) , e250153
Imaging the Atypical and Equivocal Appendix
This slide presentation reviews challenging CT and US findings of acute appendicitis, its mimics and atypical manifestations, and appendiceal neoplasms and provides radiologists with practical strategies to support accurate diagnosis.
- 4 March 2026
- Published by Springer Nature in Emergency Radiology
- Vol. 33 (2) , 235-243
CT evaluation of traumatic diaphragmatic injury: a large retrospective cohort study of blunt and penetrating trauma
- 4 March 2026
- Published by Springer Nature in Emergency Radiology
- Vol. 33 (2) , 235-243
CT evaluation of traumatic diaphragmatic injury: a large retrospective cohort study of blunt and penetrating trauma
- 19 February 2026
- Published by Springer Nature in Neuroradiology
- Vol. 68 (8) , 2343-2352
Optimizing virtual monoenergetic imaging for dual‑energy CT venography assessment of intracranial venous sinus stents
To determine a practical virtual monoenergetic imaging (VMI) energy range that improves in‑stent lumen assessment and reduces metal‑related artifacts on dual‑energy CT venography (DECT‑CTV) after intracranial venous sinus stenting. Retrospective single‑center study including 10 patients (13 stents) who underwent DECT‑CTV after venous sinus stenting for idiopathic intracranial hypertension or pulsatile tinnitus. VMI series were reconstructed from 40 to 140 keV in 10‑keV increments. Two neuroradiologists independently scored in‑stent lumen visibility and beam‑hardening/streak artifacts (5‑point Likert). Quantitative analysis used standardized ROI measurements (3 in‑stent and 2 adjacent out‑of‑stent ROIs per stent per keV) and derived metrics including in‑stent attenuation, background noise, CNR, SNR, and dCT_in-out (in‑stent minus out‑of‑stent attenuation). Qualitatively, both readers preferred intermediate‑high energies (90-100 keV) for overall interpretability. Quantitatively, in‑stent attenuation decreased from 1180.7 ± 245.5 HU (40 keV) to 112.5 ± 15.7 HU (140 keV). Noise and dCT_in-out decreased with increasing energy, whereas CNR decreased with increasing energy. SNR showed no statistically significant differences across energies. For intracranial venous sinus stents on DECT‑CTV, VMI around 90-100 keV provided the best perceived balance between lumen accessibility and artifact suppression. Findings require validation in larger cohorts with device‑specific subgroup analyses and complementary artifact indices.
- 19 February 2026
- Published by Springer Nature in Neuroradiology
- Vol. 68 (8) , 2343-2352
Optimizing virtual monoenergetic imaging for dual‑energy CT venography assessment of intracranial venous sinus stents
To determine a practical virtual monoenergetic imaging (VMI) energy range that improves in‑stent lumen assessment and reduces metal‑related artifacts on dual‑energy CT venography (DECT‑CTV) after intracranial venous sinus stenting. Retrospective single‑center study including 10 patients (13 stents) who underwent DECT‑CTV after venous sinus stenting for idiopathic intracranial hypertension or pulsatile tinnitus. VMI series were reconstructed from 40 to 140 keV in 10‑keV increments. Two neuroradiologists independently scored in‑stent lumen visibility and beam‑hardening/streak artifacts (5‑point Likert). Quantitative analysis used standardized ROI measurements (3 in‑stent and 2 adjacent out‑of‑stent ROIs per stent per keV) and derived metrics including in‑stent attenuation, background noise, CNR, SNR, and dCT_in-out (in‑stent minus out‑of‑stent attenuation). Qualitatively, both readers preferred intermediate‑high energies (90-100 keV) for overall interpretability. Quantitatively, in‑stent attenuation decreased from 1180.7 ± 245.5 HU (40 keV) to 112.5 ± 15.7 HU (140 keV). Noise and dCT_in-out decreased with increasing energy, whereas CNR decreased with increasing energy. SNR showed no statistically significant differences across energies. For intracranial venous sinus stents on DECT‑CTV, VMI around 90-100 keV provided the best perceived balance between lumen accessibility and artifact suppression. Findings require validation in larger cohorts with device‑specific subgroup analyses and complementary artifact indices.
- 19 January 2026
- Published by Springer Nature in Neuroradiology
- Vol. 68 (4) , 831-841
Comparison of deep learning reconstruction and adaptive statistical iterative reconstruction for head CT in acute stroke
- 19 January 2026
- Published by Springer Nature in Neuroradiology
- Vol. 68 (4) , 831-841
Comparison of deep learning reconstruction and adaptive statistical iterative reconstruction for head CT in acute stroke
- Open Access
- 9 July 2025
- Published by SAGE Publications in Canadian Association of Radiologists Journal
- Vol. 77 (1) , 212-217
The ‘FLAIR Motor Sign’: FLAIR Signal Abnormality in Precentral Cortex is Useful to Diagnose Adult Global Hypoxic-Ischemic Brain Injury Following Cardiac Arrest
Purpose: The precentral cortex normally demonstrates lower signal intensity compared to remainder of the neocortex on 2D fluid attenuated inversion recovery (FLAIR) images. Loss of this normal hypointensity bilaterally can be seen in patients with adult hypoxic-ischemic brain injury (HIBI). We have named this the ‘FLAIR motor’ sign (FMS). The performance of this sign for detection of HIBI is evaluated in this case-control study. Methods: MRI studies of 74 consecutive patients with clinical evidence of HIBI following cardiac arrest formed the ‘case’ group. Controls comprised of normal MRI studies of an equal number of age and gender matched patients. Two fellowship-trained neuro-radiologists reviewed the MRI studies in a blinded randomized fashion and recorded the presence or absence of ‘FLAIR motor’ sign. Results: Average time from cardiac arrest to MRI was 7.12 days (range: 1-25 days). The average sensitivity and specificity of ‘FLAIR motor’ sign for HIBI was 86.49% and 100% respectively. The sign demonstrated excellent inter-reader agreement (kappa >0.8). Conclusion: The loss of the normal hypointensity in bilateral pre-central cortex on 2D-FLAIR images is a specific and reliable MRI sign of HIBI in the subacute phase following cardiac arrest in adults.
- 9 July 2025
- Published by SAGE Publications in Canadian Association of Radiologists Journal
- Vol. 77 (1) , 212-217
Open AccessThe ‘FLAIR Motor Sign’: FLAIR Signal Abnormality in Precentral Cortex is Useful to Diagnose Adult Global Hypoxic-Ischemic Brain Injury Following Cardiac Arrest
Purpose: The precentral cortex normally demonstrates lower signal intensity compared to remainder of the neocortex on 2D fluid attenuated inversion recovery (FLAIR) images. Loss of this normal hypointensity bilaterally can be seen in patients with adult hypoxic-ischemic brain injury (HIBI). We have named this the ‘FLAIR motor’ sign (FMS). The performance of this sign for detection of HIBI is evaluated in this case-control study. Methods: MRI studies of 74 consecutive patients with clinical evidence of HIBI following cardiac arrest formed the ‘case’ group. Controls comprised of normal MRI studies of an equal number of age and gender matched patients. Two fellowship-trained neuro-radiologists reviewed the MRI studies in a blinded randomized fashion and recorded the presence or absence of ‘FLAIR motor’ sign. Results: Average time from cardiac arrest to MRI was 7.12 days (range: 1-25 days). The average sensitivity and specificity of ‘FLAIR motor’ sign for HIBI was 86.49% and 100% respectively. The sign demonstrated excellent inter-reader agreement (kappa >0.8). Conclusion: The loss of the normal hypointensity in bilateral pre-central cortex on 2D-FLAIR images is a specific and reliable MRI sign of HIBI in the subacute phase following cardiac arrest in adults.
- 2 May 2025
- Published by Springer Nature in Abdominal Radiology
- Vol. 50 (11) , 5581-5590
External validation of an RSNA 2023 Abdominal Trauma AI Challenge high performing machine learning model in the detection and grading of splenic injuries on CT
Purpose This study aims to validate the performance of an award-winning machine learning (ML) model from the Radiological Society of North America (RSNA) 2023 Abdominal Trauma AI Challenge in detecting splenic injuries on CT scans using a large, geographically and temporally distinct external dataset. Method A single-center retrospective study was conducted using an external dataset comprising 1216 CT scans (608 positive and 608 negative for splenic injuries). The ML model, trained on the RSNA Abdominal Traumatic Injury CT (RATIC) dataset, employs a multi-component pipeline including 2D MaxVit, 2.5D CoatNet with LSTM for study-level predictions. Model performance was evaluated using sensitivity, specificity, PPV, NPV, accuracy, F1 score, and AUC. Results The ML model achieved an AUC of 0.931 (95% CI: 0.917, 0.945) for binary classification of splenic injuries, with an accuracy of 0.849 (95% CI: 0.827, 0.868), sensitivity of 0.747 (95% CI: 0.711, 0.780), and specificity of 0.951 (95% CI: 0.930, 0.965). For high-grade splenic injuries, the model achieved an AUC of 0.950 (95% CI: 0.932, 0.968), accuracy of 0.928 (95% CI: 0.912, 0.941), sensitivity of 0.719 (95% CI: 0.643, 0.784), and specificity of 0.958 (95% CI: 0.944, 0.968). Conclusion The ML model shows strong, reliable performance and generalizability in detecting and grading splenic injuries on CT scans. This supports its potential clinical application, particularly for quick and accurate diagnosis in splenic trauma patients, and highlights the value of RSNA AI challenges in advancing clinical research and applications in medical imaging. Graphical abstract
- 2 May 2025
- Published by Springer Nature in Abdominal Radiology
- Vol. 50 (11) , 5581-5590
External validation of an RSNA 2023 Abdominal Trauma AI Challenge high performing machine learning model in the detection and grading of splenic injuries on CT
Purpose This study aims to validate the performance of an award-winning machine learning (ML) model from the Radiological Society of North America (RSNA) 2023 Abdominal Trauma AI Challenge in detecting splenic injuries on CT scans using a large, geographically and temporally distinct external dataset. Method A single-center retrospective study was conducted using an external dataset comprising 1216 CT scans (608 positive and 608 negative for splenic injuries). The ML model, trained on the RSNA Abdominal Traumatic Injury CT (RATIC) dataset, employs a multi-component pipeline including 2D MaxVit, 2.5D CoatNet with LSTM for study-level predictions. Model performance was evaluated using sensitivity, specificity, PPV, NPV, accuracy, F1 score, and AUC. Results The ML model achieved an AUC of 0.931 (95% CI: 0.917, 0.945) for binary classification of splenic injuries, with an accuracy of 0.849 (95% CI: 0.827, 0.868), sensitivity of 0.747 (95% CI: 0.711, 0.780), and specificity of 0.951 (95% CI: 0.930, 0.965). For high-grade splenic injuries, the model achieved an AUC of 0.950 (95% CI: 0.932, 0.968), accuracy of 0.928 (95% CI: 0.912, 0.941), sensitivity of 0.719 (95% CI: 0.643, 0.784), and specificity of 0.958 (95% CI: 0.944, 0.968). Conclusion The ML model shows strong, reliable performance and generalizability in detecting and grading splenic injuries on CT scans. This supports its potential clinical application, particularly for quick and accurate diagnosis in splenic trauma patients, and highlights the value of RSNA AI challenges in advancing clinical research and applications in medical imaging. Graphical abstract
- 22 April 2025
- Published by American Society of Neuroradiology (ASNR) in American Journal of Neuroradiology
- Vol. 46 (10) , 2132-2138
Portable Bedside Low-Field MRI for Assessment of Ventricular Size
BACKGROUND AND PURPOSE: Low-field portable MRI (pMRI) has been shown to be a useful alternative neuroimaging tool in the emergency department (ED) and intensive care unit (ICU), potentially addressing challenges associated with the transport of critically ill patients. We aimed to evaluate the intermodality reliability between low-field pMRI and conventional neuroimaging (CN) for assessment of ventricular size and hydrocephalus. MATERIALS AND METHODS: This retrospective study included all patients who underwent point-of-care 64 mT pMRI at a single tertiary hospital from March 30, 2022 to January 4, 2024, and had a follow-up CN, either CT or MRI, performed within 24 hours of the pMRI scan. Two raters independently evaluated pMRI for presence of hydrocephalus while blinded to CN. Bifrontal diameter, maximum skull, and Evans index were recorded. Interrater and intermodality agreement between pMRI and CN were evaluated by using the intraclass coefficient (ICC) and Cohen κ. RESULTS: Fifty-six patients (mean age of 53.5 [±14.6] years, 61% men) were included in this study. Hydrocephalus was identified in 12 (21%) on pMRI and 13 (23%) on CN. Interrater agreement on pMRI was almost perfect for bifrontal diameter (ICC, 0.94; 95% CI, 0.89–0.97) and Evans index (ICC, 0.92; 95% CI, 0.86–0.95) and substantial for determination of hydrocephalus (κ = 0.72), all P < .01. Intermodality agreement between pMRI and CN was also near perfect for averaged measurements of bifrontal diameter (ICC, 0.94; 95% CI, 0.88–0.97), Evans index (ICC, 0.95; 95% CI, 0.92–0.97), and determination of hydrocephalus (κ = 0.95), all P < .01. Using CN as reference standard, pMRI had sensitivity and specificity of 92% (95% CI, 0.85–0.99) and 100% (95% CI, 1.0–1.0), respectively on qualitative visual assessment, and 80% (95% CI, 0.70–0.90) and 98% (95% CI, 0.94–1.0), respectively, on quantitative assessment by using an Evans index cutoff of 0.3. CONCLUSIONS: Low-field pMRI demonstrated excellent interrater agreement and strong concordance with CN in assessing ventricular size, highlighting its potential as an effective point-of-care tool for neurocritical care applications. Graphical Abstract
- 22 April 2025
- Published by American Society of Neuroradiology (ASNR) in American Journal of Neuroradiology
- Vol. 46 (10) , 2132-2138
Portable Bedside Low-Field MRI for Assessment of Ventricular Size
BACKGROUND AND PURPOSE: Low-field portable MRI (pMRI) has been shown to be a useful alternative neuroimaging tool in the emergency department (ED) and intensive care unit (ICU), potentially addressing challenges associated with the transport of critically ill patients. We aimed to evaluate the intermodality reliability between low-field pMRI and conventional neuroimaging (CN) for assessment of ventricular size and hydrocephalus. MATERIALS AND METHODS: This retrospective study included all patients who underwent point-of-care 64 mT pMRI at a single tertiary hospital from March 30, 2022 to January 4, 2024, and had a follow-up CN, either CT or MRI, performed within 24 hours of the pMRI scan. Two raters independently evaluated pMRI for presence of hydrocephalus while blinded to CN. Bifrontal diameter, maximum skull, and Evans index were recorded. Interrater and intermodality agreement between pMRI and CN were evaluated by using the intraclass coefficient (ICC) and Cohen κ. RESULTS: Fifty-six patients (mean age of 53.5 [±14.6] years, 61% men) were included in this study. Hydrocephalus was identified in 12 (21%) on pMRI and 13 (23%) on CN. Interrater agreement on pMRI was almost perfect for bifrontal diameter (ICC, 0.94; 95% CI, 0.89–0.97) and Evans index (ICC, 0.92; 95% CI, 0.86–0.95) and substantial for determination of hydrocephalus (κ = 0.72), all P < .01. Intermodality agreement between pMRI and CN was also near perfect for averaged measurements of bifrontal diameter (ICC, 0.94; 95% CI, 0.88–0.97), Evans index (ICC, 0.95; 95% CI, 0.92–0.97), and determination of hydrocephalus (κ = 0.95), all P < .01. Using CN as reference standard, pMRI had sensitivity and specificity of 92% (95% CI, 0.85–0.99) and 100% (95% CI, 1.0–1.0), respectively on qualitative visual assessment, and 80% (95% CI, 0.70–0.90) and 98% (95% CI, 0.94–1.0), respectively, on quantitative assessment by using an Evans index cutoff of 0.3. CONCLUSIONS: Low-field pMRI demonstrated excellent interrater agreement and strong concordance with CN in assessing ventricular size, highlighting its potential as an effective point-of-care tool for neurocritical care applications. Graphical Abstract
- 1 November 2024
- Published by Radiological Society of North America (RSNA) in Radiology: Artificial Intelligence
- Vol. 6 (6) , e240101
The RSNA Abdominal Traumatic Injury CT (RATIC) Dataset
The RSNA Abdominal Traumatic Injury CT (ie, RATIC) dataset contains 4274 abdominal CT studies with annotations related to traumatic injuries and is available at https://www.kaggle.com/competitions/rsna-2023-abdominal-trauma-detection and https://imaging.rsna.org/dataset/5 .
- 1 November 2024
- Published by Radiological Society of North America (RSNA) in Radiology: Artificial Intelligence
- Vol. 6 (6) , e240101
The RSNA Abdominal Traumatic Injury CT (RATIC) Dataset
The RSNA Abdominal Traumatic Injury CT (ie, RATIC) dataset contains 4274 abdominal CT studies with annotations related to traumatic injuries and is available at https://www.kaggle.com/competitions/rsna-2023-abdominal-trauma-detection and https://imaging.rsna.org/dataset/5 .
- 1 November 2024
- Published by Radiological Society of North America (RSNA) in Radiology: Artificial Intelligence
- Vol. 6 (6) , e230550
Assessing the Performance of Models from the 2022 RSNA Cervical Spine Fracture Detection Competition at a Level I Trauma Center
Winning machine learning models from the RSNA 2022 Cervical Spine Fracture Detection competition demonstrated high performance on a large clinical test set of emergency department cervical spine CT scans from a level I trauma center.
- 1 November 2024
- Published by Radiological Society of North America (RSNA) in Radiology: Artificial Intelligence
- Vol. 6 (6) , e230550
Assessing the Performance of Models from the 2022 RSNA Cervical Spine Fracture Detection Competition at a Level I Trauma Center
Winning machine learning models from the RSNA 2022 Cervical Spine Fracture Detection competition demonstrated high performance on a large clinical test set of emergency department cervical spine CT scans from a level I trauma center.