KSKomal Shah

Komal Shah

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Komal Shah has a total of 80 co-authors. Here are the top co-authors by number of shared publications:

  • Max Wintermark

    Department of Radiology, The University of Texas Medical Branch, Galveston

    3 shared publications
  • Ahmed Naeem

    From the Department of Neuroradiology (P.I.K., A.N., K.S., M.W.), Division of Diagnostic Imaging, MD Anderson Cancer Center, Houston, TX and Department of Radiology (M.W.), The University of Texas Medical Branch, Galveston, TX

    2 shared publications
  • Kim O. Learned

    Department of Neuroradiology, The University of Texas MD Anderson Cancer Center, Houston, TX, USA

    2 shared publications
  • Melissa M. Chen

    Department of Neuroradiology, The University of Texas MD Anderson Cancer Center, Houston, TX, USA

    2 shared publications
  • Patricia M. de Groot

    Department of Thoracic Imaging, The University of Texas MD Anderson Cancer Center, Houston, TX

    2 shared publications
  • Peter I. Kamel

    From the Department of Neuroradiology (P.I.K., A.N., K.S., M.W.), Division of Diagnostic Imaging, MD Anderson Cancer Center, Houston, TX and Department of Radiology (M.W.), The University of Texas Medical Branch, Galveston, TX.

    2 shared publications

Latest Publications

  • Open Access

    Magnetic Resonance Imaging Simulation of the Skull Base, Oral Cavity, Pharynx, and Larynx: Advanced Imaging Protocol for a Radiation Oncology Center

    by Natalie A. West, Lucas McCullum, Johnny T. Cheruvelil, Vi T. Dinh, Yao Ding, Jihong Wang, Komal Shah, Kim Learned, Melissa Chen, Jack Phan, et al.

    Head and neck radiation oncology, with its complex locality and epidemiology, requires various imaging techniques to achieve sufficient visualization. Through collaboration with medical physicists, radiation therapists, radiation oncologists, and neuroradiologists, our group clinically implemented a standard set of magnetic resonance imaging (MRI) sequences on an MR-simulation scanner for three anatomical regions: skull base, oral cavity, and larynx/pharynx. A total of 41 patients with head and neck cancer and 3 volunteers were scanned on either a 1.5T or 3T MR-simulation scanner. Sequences described include Dixon-based approaches, SPAIR-based approaches, short T2-based approaches, and quantitative imaging methods. We describe and outline the steps to optimize these notable MRI sequences to aid in successful implementation in a radiation oncology clinic.

  • A Systematic Evaluation of Image Preprocessing in Deep Learning Detection and Segmentation of Intracranial Metastatic Disease

    by Peter I. Kamel, Ahmed Naeem, Komal Shah, Max Wintermark

    Image preprocessing is an essential, though often overlooked, part of machine learning, and it is unclear how preprocessing techniques affect metastatic disease segmentation. This is particularly true given the differences between segmentation of primary brain tumors and the detection and segmentation of small metastases. The purpose of this work is to systematically evaluate how preprocessing techniques affect the detection and segmentation of intracranial metastases. Using nnU-Net, we evaluated preprocessing techniques on a training data set of 1,034 MRIs and testing set of 256 MRIs. We examined the effects of label preparation, bias correction, normalization, registration, and denoising, as well as the contribution of each MRI sequence. Model performance was compared to the baseline model using paired t-tests, evaluating global Dice scores, lesion detection rate, and per-lesion and per-exam Dice score, surface Dice score, and 95% Hausdorff distance, stratified by lesion volume. Baseline model performance demonstrated a global Dice score of 0.72 ± 0.28 and a lesion detection rate of 70.8% (1,328/1,877), strongly dependent on lesion size. Including all labels in training, as opposed to only tumor core, resulted in a significant decrease in detection rate to 68.8% (1,292/1,877) (p < 0.001), as well as reduced per-lesion segmentation particularly for smaller lesions. Bias correction and denoising demonstrated no significant effect on performance. Pre-normalization of data led to worse segmentation performance. Image registration reduced both lesion detection and segmentation. Training with fewer MRI sequences, in several cases, improved segmentation. Our findings demonstrate that certain preprocessing steps have a significant impact on model performance while others have negligible effects and may not be necessary. With nnU-Net, we observed highest detection and segmentation scores by reducing training labels and MRI sequences to those focused on tumor core, with additional, less relevant data, reducing performance. Several preprocessing steps showed no measurable benefit and image registration may be detrimental to the detection and segmentation of small metastases.

  • Evaluating the Utility and Limitations of Machine Learning Tumor Segmentation for Automated Longitudinal Response Assessment in Neuro-Oncology Treatment Response Classification

    by Peter Kamel, Ahmed Naeem, Hamza Salim, Sahar Alizada, Ahmed Msherghi, Chloe Barakat, Max Wintermark, Komal Shah

    Machine learning segmentation has emerged in tumor assessment with high performance in volumetric evaluation of brain tumors. It is unclear, however, if such methods can be translated into full end-to-end clinical tools for assessing treatment response. This study aims to assess the ability and current limitations of machine learning of performing fully automated longitudinal Response Assessment in Neuro-Oncology (RANO). An nnU-Net model was trained for segmentation on 4,162 MRIs of pre- and post-treatment brain tumors. A separate dataset of 91 patients was preprocessed, including co-registration for longitudinal lesion tracking. The nnU-Net model was used to segment enhancing tumor on the longitudinal dataset. After excluding patients classified as progression due to T2/FLAIR changes or clinical criteria, 233 pairwise MRI comparisons remained for 60 patients. Using auto-segmented tumor masks, comparisons were classified into Progressive Disease (PD), Partial Response (PR), Stable Disease (SD), and Complete Response (CR) according to the original RANO criteria based on 2D lesion size, measurable criteria, appearance of new lesions, and changes over time. Model results were compared to expert ground truth and evaluated for accuracy in classifying PD and non-PD disease, per class accuracy, and a variety of performance metrics. All discrepant cases were manually reviewed to classify the cause of error. The model was 77.7% (95% CI 70.6%-84.2%) accurate in distinguishing PD and non-PD disease with sensitivity 92.1% (86.2%-97.3%) and specificity 60.4% (43.4%-74.7%). The highest per-class accuracy was PD at 92.1% (86.2%-97.3%), with much lower performance for non-PD categories, worst for PR at 14.3% (0.0%-38.5%). Patient-averaged accuracy was 66.9% (58.2%-75.3%). False positive PD predictions were commonly due to evolving changes around the resection cavity, affecting the morphology and location of enhancement and accounting for 32.8% of SD classifications. Other causes of error included changing morphology but similar volume of enhancement, missed lesions in less typical locations, and radiation necrosis. Automated tumor segmentation is emerging in treatment response assessment, though highly trained segmentation models are still met with limitations in fully automated response assessment due to nuances such as post-surgical enhancement and comparing location and morphology changes between timepoints.

  • Real-World Experience of Tarlatamab in Extensive-Stage Small Cell Lung Carcinoma Showed Promising Systemic and Intracranial Efficacy With Higher Neurologic Toxicity

    by Mitchell Parma, Kaiwen Wang, Ceylan Altintas Taslicay, Komal Shah, Eric K. Singhi, Melvin Rivera, Luisa Solis Soto, Wei-Lien Wang, Lingzhi Hong, Ziping Li, et al.

    PURPOSE: Tarlatamab, a CD3/delta-like ligand 3 bispecific T-cell engager, was approved in 2024 for relapsed extensive-stage small cell lung cancer (ES-SCLC) based on promising clinical trial results. As a T-cell effector therapy, it also has unique side effects, including cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS). In this article, we report a real-world series of patients treated with tarlatamab. MATERIALS AND METHODS: We retrospectively evaluated safety and efficacy in patients with relapsed ES-SCLC receiving tarlatamab from July 1, 2024, to February 15, 2025, at MD Anderson Cancer Center, including clinical response and toxicity and the radiographic response of brain metastases (BM) per Response Assessment in Neuro-Oncology (RANO)-BM. RESULTS: 27 patients were included in our analyses. The median age was 68 (range, 43-83) years, and 15 patients (56%) were women. Before tarlatamab, nine patients (33%) had no BM, and 11 (41%) had active, untreated BM. Of the 22 patients evaluable for response with a median follow-up time of 3 months, four patients (18.2%) had partial response (PR), 71 (31.8%) had stable disease (SD), and 11 (50%) had progressive disease (PD) based on RECIST criteria. Median progression-free survival was 5.9 months (95% CI, 2.6 to not reached). Of the 14 patients with prior BM and serial brain magnetic resonance imaging, based on RANO-BM criteria, two had complete response, one had PR, nine had SD, and only two had PD. Of the four patients without BM, none had new intracranial disease after starting tarlatamab. Of the 27 patients who received cycle 1 day 1 (C1D1) of tarlatamab, 10 (37%) patients developed CRS (1 grade ≥3) and 8 (30%) had ICANS (4 grade ≥3). Of the 25 patients who received C1D8 of tarlatamab, 10 (40%) developed CRS (1 grade ≥3) and 4 (16%) developed ICANS (none grade ≥3). CONCLUSION: Tarlatamab demonstrates both intracranial and extracranial efficacy in our real-world patient cohort. Our cohort's toxicity profile showed similar CRS but higher ICANS rates compared with those reported in clinical trials. More studies are needed to determine biomarkers of efficacy and toxicity.


  • Diagnostic Performance of Ultrasound in Neck Node NI-RADS Category 2

    by Richard Dagher, Alexander Khalaf, Susana Calle, Samir A. Dagher, Komal B. Shah, Amy Juliano, Ashley H. Aiken, Kim O. Learned

    BACKGROUND AND PURPOSE: The NI-RADS scoring system standardized imaging surveillance of head and neck (H&N) cancer with risk classification. A nodal NIRADS score of 2 on contrast-enhanced CT (CECT) of the neck indicates low suspicion for recurrence/persistent disease and close follow-up or addition of PET are recommended. The unclear follow-up imaging findings and/or mild FDG uptake raise patient’s anxiety of potential delay in diagnosis and intervention while adding high imaging cost. Therefore, at our institution, diagnostic US/US-guided fine needle aspiration (US-FNA) is incorporated in our paradigm. We aim to evaluate US performance in nodal NI-RADS 2 on CECT as alternative valuable tool in surveillance imaging guidelines. MATERIALS AND METHODS: We conducted a retrospective database search (2019-2024) for patients with primary H&N cancer (excluding thyroid cancer and melanoma), a single index neck node NI-RADS 2 on surveillance CECT neck, and a neck US/US-FNA performed within 3 months afterwards for evaluation of the NI-RADS 2 node. We categorized US/US-guided FNA results as positive or negative and reviewed clinical and imaging follow-up, management and nodal disease status up to 1 year following US. The incidence of nodal recurrence and US diagnostic performance were evaluated. RESULTS: Of 90 patients, 36 (40%) had normal diagnostic US with no FNA performed and were thus considered negative, and 54 patients (60%) had abnormal US and hence concurrent US-FNA. 18 (33.3%) US-FNAs were positive for tumor; 27 with normal lymphoid tissue and 9 with indeterminate cytology (no viable malignant cells, acellular or atypia) were considered negative (66.7%). All positive US-FNAs resulted in management changes. 2 patients with normal diagnostic US, 1 with negative FNA and 1 with indeterminate FNA developed recurrence in these nodes within 1 year. The incidence of US-detected malignancy was 20% in patients with a nodal NIRADS 2, surpassing the published rate of 14.3%. The sensitivity, accuracy and NPV of US/US-FNA in detecting tumor recurrence/persistence in nodal NI-RADS 2 are 81.8%, 95.6% and 94.4% respectively. CONCLUSIONS: Ultrasound demonstrated good diagnostic performance in the detection of nodal recurrence in patients with NI-RADS 2 on CECT. Its role as an alternative tool in surveillance should be considered. ABBREVIATIONS: CECT = contrast-enhanced CT; CEMR = contrast-enhanced MR; ENE = extranodal extension; FNA = fine-needle aspiration; NI-RADS = Neck Imaging Reporting and Data System; NPV = negative predictive value; PPV = positive predictive value; SCC = squamous cell carcinoma; RVU = relative value units; US = ultrasound.


  • Diagnostic Confidence of Contrast-Enhanced T1-Weighted MRI for the Detection of Brain Metastases: 3D FSE versus 3D GRE–Based Sequences

    by Maria Gule-Monroe, Nathan Chasen, James P. Long, Vinodh A. Kumar, Komal Shah, Melissa Chen, Jason Stafford, Caroline Chung, Max Wintermark, Ping Hou, et al.

    BACKGROUND AND PURPOSE: This retrospective study evaluated the utility of contrast-enhanced (CE) T1-weighted 3D fast spinecho-based SPACE sequences for brain metastasis detection on 3T MRI compared to gradient recalled-echo-based 3D fast low-angle shot (FLASH) sequence. MATERIALS AND METHODS: We identified all patients at a single institution who underwent SPACE and 3D FLASH sequences as part of a practice quality improvement project. Their medical records were retrospectively reviewed. Five certified neuroradiologists reviewed the images, with at least 2 weeks separation between scoring sequences for the same patient. The following parameters were evaluated: number of metastatic lesions, number of indeterminate lesions, lesion margin, contrast-to-noise ratio (CNR), extent of image artifacts, and overall image quality. CNR was also quantified for solidly enhancing lesions > 1 cm. RESULTS: We identified 220 patients who underwent SPACE and 3D FLASH sequences (the order of the sequences was equally distributed). Of these, 79 had brain metastases on imaging, and 7 were excluded; thus, 72 patients were included in the study. Twenty patients were scored by 2 radiologists. Out of the 92 evaluations, SPACE detected more lesions than did 3D FLASH in 35, while 3D FLASH detected more lesions in 10. More indeterminate lesions were seen on 3D FLASH (27) than on SPACE (9). For lesion margin, CNR, and overall image quality on a Likert scale, SPACE performed significantly better than did 3D FLASH, with less image artifacts (P < 0.00001). Higher quantitative CNRs were found on SPACE than on 3D FLASH images, although this result was not statistically significant (median = 22.9 vs. 15.5, respectively, P = 0.134). There was a high inter-reader lesion detection concordance with Krippendorf’s alpha ordinals at 0.962 for SPACE, 0.870 for 3D FLASH, and 0.918 for the two sequences combined. CONCLUSIONS: Compared with 3D FLASH, the SPACE sequence detected more metastatic lesions and was rated higher for image quality, lesion margin, and CNR, with fewer artifacts. Importantly, the SPACE sequence resulted in increased reader confidence, with fewer indeterminate lesions detected. ABBREVIATIONS: FLASH = fast low-angle shot; FSE = fast spin-echo; GRE = gradient-recalled echo; MP-RAGE = magnetization-prepared rapid gradient echo; SPACE = Sampling Perfection with Application-optimized Contrasts using different flip angle Evolution; VIBE = volumetric interpolated breath-hold examination.


  • Open Access

    Peer Learning in Neuroradiology: Not as Easy as It Sounds

    by K. Mani, K. Shah, N. Kadom, D. Seidenwurm, A.J. Nemeth

    Peer Learning in Neuroradiology: Not as Easy as It Sounds Peer Learning (PL) is an engaging activity in which practicing radiologists come together to review cases from which they can learn jointly. The major impetus for PL lies in the overarching goal of improving diagnosis in radiology through a team-based culture of viewing mistakes as an opportunity to learn. 1 In its book Improving Diagnosis in Health Care, the Institute of Medicine (IOM) found that most people will be affected by at least 1 diagnostic error in health care. 2 As a result of the IOM book Improving Diagnosis in Health Care, several external drivers are in place to change the practices of health care providers toward improved diagnosis. Radiologists realized that PL, but not random score-based peer review, best meets the IOM goals of establishing effective teamwork, educating practitioners in the diagnostic process, learning from mistakes, creating a culture that improves diagnostic performance, and establishing a reporting mechanism for discrepancies. 1 Following a 2020 Radiology Peer Learning Summit, the American College of Radiology (ACR) developed a new accreditation pathway that replaces score-based random peer review with PL. 3,4 The American Board of Radiology (ABR) added PL as an alternative participatory activity for meeting Maintenance of Certification (MOC) Part 4 criteria. 5 The Joint Commission (TJC) serves as another external driver of improved practitioner performance, eg, peer review, through Ongoing Professional Practice Evaluation (OPPE) requirements. 6 Assessments of agreement among university neuroradiologists showed up to 12.4% disagreement rates, 7 which are much higher than the reported 2.9% for score-based random peer review. 8 Among errors in neuroradiology are discrepancies regarding vascular, neoplastic, and congenital disorders, as well as artifacts. 7 Moreover, neuroradiology errors can also arise from test-selection errors, protocolling errors, technical errors, and failure to communicate results in a timely fashion. Education can decrease errors, given that neuroradiologists with high participation rates in the Tumor Board have lower diagnostic error rates. 9 Having awareness of "blind spots," for example with complex head and neck anatomy and pathology, may decrease interpretive errors and could conceivably be improved with PL. 10,11 In fact, there is currently some evidence that PL creates learning opportunities, 12,13 but there is still a lack of adoption of PL programs nationwide 14 and a lack of scientific evidence demonstrating its effectiveness. This Perspectives comes from members of the American Society of Neuroradiology (ASNR) Quality, Safety and Value Committee. Here, we describe several challenges faced by neuroradiologists who are interested in serving as PL champions. Among these challenges is an inability to recruit volunteer champions to drive PL programs, lack of resources for running a PL program, unknown effects on required reporting to TJC, and lack of evidence favoring PL over score-based random peer review.


  • Imaging Primer on Chimeric Antigen Receptor T-Cell Therapy for Radiologists

    by Patricia M. de Groot, Octavio Arevalo, Komal Shah, Chad D. Strange, Girish S. Shroff, Jitesh Ahuja, Mylene T. Truong, John F. de Groot, Ioannis Vlahos

    Chimeric antigen receptor T-cell therapy is a recently approved breakthrough in adoptive immunotherapy, with high efficacy in relapsed or refractory hematologic malignancies and promising future ap... Chimeric antigen receptor (CAR) T-cell therapy is a recently approved breakthrough treatment that has become a new paradigm in treatment of recurrent or refractory B-cell lymphomas and pediatric or...