Joshua Vic Chen

All Publications

• Automated neonatal nnU-Net brain MRI extractor trained on a large multi-institutional dataset. Scientific reports Chen, J. V., Li, Y., Tang, F., Chaudhari, G., Lew, C., Lee, A., Rauschecker, A. M., Haskell-Mendoza, A. P., Wu, Y. W., Calabrese, E. 2024; 14 (1): 4583

  Abstract

  Brain extraction, or skull-stripping, is an essential data preprocessing step for machine learning approaches to brain MRI analysis. Currently, there are limited extraction algorithms for the neonatal brain. We aim to adapt an established deep learning algorithm for the automatic segmentation of neonatal brains from MRI, trained on a large multi-institutional dataset for improved generalizability across image acquisition parameters. Our model, ANUBEX (automated neonatal nnU-Net brain MRI extractor), was designed using nnU-Net and was trained on a subset of participants (N = 433) enrolled in the High-dose Erythropoietin for Asphyxia and Encephalopathy (HEAL) study. We compared the performance of our model to five publicly available models (BET, BSE, CABINET, iBEATv2, ROBEX) across conventional and machine learning methods, tested on two public datasets (NIH and dHCP). We found that our model had a significantly higher Dice score on the aggregate of both data sets and comparable or significantly higher Dice scores on the NIH (low-resolution) and dHCP (high-resolution) datasets independently. ANUBEX performs similarly when trained on sequence-agnostic or motion-degraded MRI, but slightly worse on preterm brains. In conclusion, we created an automatic deep learning-based neonatal brain extraction algorithm that demonstrates accurate performance with both high- and low-resolution MRIs with fast computation time.

  View details for DOI 10.1038/s41598-024-54436-8

  View details for PubMedID 38403673

  View details for PubMedCentralID PMC10894871

• Factors and Labor Cost Savings Associated with Successful Pediatric Imaging without Anesthesia: a Single-Institution Study. Academic radiology Chen, J. V., Zapala, M. A., Zhou, A., Vu, N., Meyer, L., Smith, M. D., Kelleher, C., Glenn, O. A., Courtier, J., Li, Y. 2023; 30 (9): 1979-1988

  Abstract

  In pediatric imaging, sedation is often necessary to obtain diagnostic quality imaging. We aim to quantify patient and imaging-specific factors associated with successful pediatric scans without anesthesia and to evaluate labor cost savings associated with our institutional Scan Without Anesthesia Program (SWAP).Patients who participated in SWAP between 2019-2022 were identified. Chart review was conducted to obtain sociodemographic and clinical information. Radiology database was used to obtain scan duration, modality/body part of examination, and administration of contrast. Mann-Whitney U and Chi-Square tests were used for univariate analysis of factors associated with success. Multivariate logistic regression was used to evaluate independent contributions to success. Associated hospital labor cost savings were estimated using salary information obtained through publicly available resources.Of 731 patients, 698 had successful and 33 had unsuccessful scans (95% success rate). In univariate analysis, older age, female sex, absence of developmental delay, and administration of contrast were significantly associated with successful scans. Multivariate analyses revealed that older age, female sex, and absence of developmental delay were significant independent factors lending toward success. Imaging-related factors were not associated with outcome in multivariate analysis. Estimated labor cost savings were $139,367.80 per year for the medical center.SWAP had an overall success rate of 95%. Older age, absence of developmental delay, and female sex were independently significantly associated with successful outcome. Cost analysis reveals substantial labor cost savings to the institution compared with imaging under anesthesia.

  View details for DOI 10.1016/j.acra.2022.12.041

  View details for PubMedID 36641347

• MR-guided focused ultrasound therapy of extra-abdominal desmoid tumors: a multicenter retrospective study of 105 patients. European radiology Düx, D. M., Baal, J. D., Bitton, R., Chen, J., Brunsing, R. L., Sheth, V. R., Rosenberg, J., Kim, K., Ozhinsky, E., Avedian, R., Ganjoo, K., Bucknor, M., Dobrotwir, A., Ghanouni, P. 2023

  Abstract

  To assess the safety and efficacy of magnetic resonance-guided focused ultrasound (MRgFUS) for the treatment extra-abdominal desmoids.A total of 105 patients with desmoid fibromatosis (79 females, 26 males; 35 ± 14 years) were treated with MRgFUS between 2011 and 2021 in three centers. Total and viable tumors were evaluated per patient at last follow-up after treatment. Response and progression-free survival (PFS) were assessed with (modified) response evaluation criteria in solid tumors (RECIST v.1.1 and mRECIST). Change in Numerical Rating Scale (NRS) pain and 36-item Short Form Health Survey (SF-36) scores were compared. Treatment-related adverse events were recorded.The median initial tumor volume was 114 mL (IQR 314 mL). After MRgFUS, median total and viable tumor volume decreased to 51 mL (95% CI: 30-71 mL, n = 101, p < 0.0001) and 29 mL (95% CI: 17-57 mL, n = 88, p < 0.0001), respectively, at last follow-up (median: 15 months, 95% CI: 11-20 months). Based on total tumor measurements (RECIST), 86% (95% CI: 75-93%) had at least stable disease or better at last follow-up, but 50% (95% CI: 38-62%) of remaining viable nodules (mRECIST) progressed within the tumor. Median PFS was reached at 17 and 13 months for total and viable tumors, respectively. NRS decreased from 6 (IQR 3) to 3 (IQR 4) (p < 0.001). SF-36 scores improved (physical health (41 (IQR 15) to 46 (IQR 12); p = 0.05, and mental health (49 (IQR 17) to 53 (IQR 9); p = 0.02)). Complications occurred in 36%, most commonly 1st/2nd degree skin burns.MRgFUS reduced tumor volume, reduced pain, and improved quality of life in this series of 105 patients with extra-abdominal desmoid fibromatosis.Imaging-guided ablation is being increasingly used as an alternative to surgery, radiation, and medical therapy for the treatment of desmoid fibromatosis. MR-guided high-intensity focused ultrasound is an incisionless ablation technique that can be used to reduce tumor burden effectively and safely.• Desmoid fibromatosis was treated with MR-guided high-intensity focused ultrasound in 105 patients. • MR-guided focused ultrasound ablation reduced tumor volume and pain and improved quality of life. • MR-guided focused ultrasound is a treatment option for patients with extra-abdominal desmoid tumors.

  View details for DOI 10.1007/s00330-023-10073-9

  View details for PubMedID 37615768

• MR Imaging Findings in a Large Population of Autoimmune Encephalitis. AJNR. American journal of neuroradiology Gillon, S., Chan, M., Chen, J., Guterman, E. L., Wu, X., Glastonbury, C. M., Li, Y. 2023; 44 (7): 799-806

  Abstract

  Autoimmune encephalitis is a rare condition in which autoantibodies attack neuronal tissue, causing neuropsychiatric disturbances. This study sought to evaluate MR imaging findings associated with subtypes and categories of autoimmune encephalitis.Cases of autoimmune encephalitis with specific autoantibodies were identified from the medical record (2009-2019). Cases were excluded if no MR imaging of the brain was available, antibodies were associated with demyelinating disease, or >1 concurrent antibody was present. Demographics, CSF profile, antibody subtype and group (group 1 intracellular antigen or group 2 extracellular antigen), and MR imaging features at symptom onset were reviewed. Imaging and clinical features were compared across antibody groups using χ2 and Wilcoxon rank-sum tests.Eighty-five cases of autoimmune encephalitis constituting 16 distinct antibodies were reviewed. The most common antibodies were anti-N-methyl-D-aspartate (n = 41), anti-glutamic acid decarboxylase (n = 7), and anti-voltage-gated potassium channel (n = 6). Eighteen of 85 (21%) were group 1; and 67/85 (79%) were group 2. The median time between MR imaging and antibody diagnosis was 14 days (interquartile range, 4-26 days). MR imaging had normal findings in 33/85 (39%), and 20/33 (61%) patients with normal MRIs had anti-N-methyl-D-aspartate receptor antibodies. Signal abnormality was most common in the limbic system (28/85, 33%); 1/68 (1.5%) had susceptibility artifacts. Brainstem and cerebellar involvement were more common in group 1, while leptomeningeal enhancement was more common in group 2.Sixty-one percent of patients with autoimmune encephalitis had abnormal brain MR imaging findings at symptom onset, most commonly involving the limbic system. Susceptibility artifact is rare and makes autoimmune encephalitis less likely as a diagnosis. Brainstem and cerebellar involvement were more common in group 1, while leptomeningeal enhancement was more common in group 2.

  View details for DOI 10.3174/ajnr.A7907

  View details for PubMedID 37385678

  View details for PubMedCentralID PMC10337613

• Deep Learning to Predict Neonatal and Infant Brain Age from Myelination on Brain MRI Scans. Radiology Chen, J. V., Chaudhari, G., Hess, C. P., Glenn, O. A., Sugrue, L. P., Rauschecker, A. M., Li, Y. 2022; 305 (3): 678-687

  Abstract

  Background Assessment of appropriate brain myelination on T1- and T2-weighted MRI scans is based on gestationally corrected age (GCA) and requires subjective visual inspection of the brain with knowledge of normal myelination milestones. Purpose To develop a convolutional neural network (CNN) capable of estimating neonatal and infant GCA based on brain myelination on MRI scans. Materials and methods In this retrospective study from one academic medical center, brain MRI scans of patients aged 0-25 months with reported normal myelination were consecutively collected between January 1995 and June 2019. The GCA at MRI was manually calculated. After exclusion criteria were applied, T1- and T2-weighted MRI scans were preprocessed with skull stripping, linear registration, z scoring for normalization, and downsampling. A three-dimensional regression CNN was trained to predict GCA using mean absolute error (MAE) as its loss function. Attention maps were calculated using layer-wise relevance propagation. Models were validated on an external test set from the National Institutes of Health (NIH). Model MAEs were compared using Kruskal-Wallis and Mann-Whitney tests. Results A total of 518 neonates and infants (mean GCA, 67 weeks ± 33 [SD], 56% male) was included, comprising 469 T1-, 438 T2-, and 389 T1- and T2-weighted studies. Across 10 runs, MAEs of T1-, T2-, and T1- and T2-weighted networks were 9.8 ± 2.3, 9.1 ± 1.9, and 7.7 ± 1.7 weeks, respectively. Attention map analysis demonstrated increased network attention to the cerebellum, posterior white matter, and basal ganglia signal in neonates with GCA of less than 40 weeks and the anterior white matter signal in infants with GCA of more than 120 weeks, corresponding to the known progression of myelination. The T1- and T2-weighted network tested on the external NIH test set had an MAE of 9.1 weeks, which was reduced to 5.9 weeks with further training using half the external test set (P < .001). Conclusion A three-dimensional convolutional neural network can predict the gestationally corrected age of neonates and infants aged 0-25 months based on brain myelination patterns on T1- and T2-weighted MRI scans. © RSNA, 2022 Online supplemental material is available for this article.

  View details for DOI 10.1148/radiol.211860

  View details for PubMedID 35852429

• Diagnosis of acute mesenteric ischemia assisted by dual-energy CT: a case report. Emergency radiology Chen, J. V., Ohliger, M. A., Chung, M., Sugi, M. D., Ye, J., Nishimura, S. L., Choi, H. H. 2022; 29 (3): 611-614

  Abstract

  In this case report, dual-energy CT was critical in the diagnosis of acute mesenteric ischemia by differentiating normal contrast-enhanced bowel and hemorrhagic necrosis. Iodine map showed a segment of small bowel with minimal contrast enhancement, and virtual non-contrast imaging revealed hyperattenuating bowel. This finding changed management for the patient and prevented complications from impending bowel perforation. Histopathological analysis confirmed hemorrhagic necrosis of the bowel segment. In cases of suspected bowel ischemia, dual-energy CT can distinguish bowel wall hemorrhage from contrast enhancement and allow for accurate diagnosis.

  View details for DOI 10.1007/s10140-022-02029-1

  View details for PubMedID 35128620

  View details for PubMedCentralID 5036332

• Readability of Patient Discharge Instructions. Journal of general internal medicine Burns, S. T., Amobi, N., Chen, J. V., O'Brien, M., Haber, L. A. 2022; 37 (7): 1797-1798

  View details for DOI 10.1007/s11606-021-06988-y

  View details for PubMedID 34240287

  View details for PubMedCentralID PMC9130361

• Multi-institutional Experience with Patient Image Access Through Electronic Health Record Patient Portals. Journal of digital imaging Choi, H. H., Kotsenas, A. L., Chen, J. V., Bronsky, C., Roth, C. J., Kohli, M. D. 2022; 35 (2): 320-326

  Abstract

  The objective is to determine patients' utilization rate of radiology image viewing through an online patient portal and to understand its impact on radiologists. IRB approval was waived. In this two-part, multi-institutional study, patients' image viewing rate was retrospectively assessed, and radiologists were anonymously surveyed for the impact of patient imaging access on their workflow. Patient access to web-based image viewing via electronic patient portals was enabled at 3 institutions (all had open radiology reports) within the past 5 years. The number of exams viewed online was compared against the total number of viewable imaging studies. An anonymized survey was distributed to radiologists at the 3 institutions, and responses were collected over 2 months. Patients viewed 14.2% of available exams - monthly open rate varied from 7.3 to 41.0%. A total of 254 radiologists responded to the survey (response rate 32.8%); 204 were aware that patients could view images. The majority (155/204; 76.0%) felt no impact on their role as radiologists; 11.8% felt negative and 9.3% positive. The majority (63.8%) were never approached by patients. Of the 86 who were contacted, 46.5% were contacted once or twice, 46.5% 3-4 times a year, and 4.7% 3-4 times a month. Free text comments included support for healthcare transparency (71), concern for patient confusion and anxiety (45), and need for attention to radiology reports and image annotations (15). A small proportion of patients viewed their radiology images. Overall, patients' image viewing had minimal impact on radiologists. Radiologists were seldom contacted by patients. While many radiologists feel supportive, some are concerned about causing patient confusion and suggest minor workflow modifications.

  View details for DOI 10.1007/s10278-021-00565-9

  View details for PubMedID 35022926

  View details for PubMedCentralID PMC8921401

• Cervical Lymph Node Features Predictive of Suboptimal Adequacy During Ultrasound-Guided Fine-Needle Aspiration in Thyroid Cancer Patients. Journal of ultrasound in medicine : official journal of the American Institute of Ultrasound in Medicine Chen, J. V., Morgan, T. A., Liu, C., Khanafshar, E., Choi, H. H. 2022; 41 (1): 135-145

  Abstract

  To determine the rate of cytologic and diagnostic adequacy and identify features associated with suboptimal tissue sampling in ultrasound-guided fine-needle aspiration (US-FNA) of suspected nodal disease in thyroid cancer patients.A single-institution pathology database was queried for lymph node FNA reports in thyroid cancer patients from 2014 to 2019. Charts were reviewed for demographics, body mass index (BMI), prior thyroidectomy, cancer type, and subsequent surgery. Ultrasound images were retrospectively reviewed for location, size, depth from skin, cystic components, macrocalcification, echogenic foci, and internal vascularity score. Pathology reports were categorized as cellular and diagnostic, hypocellular/acellular but diagnostic with abnormal cells or thyroglobulin levels, or hypocellular and nondiagnostic. Correlation and multivariate regression analyses were performed.Initial query yielded 552 lesions in 343 subjects. Following exclusion, 377 lesions in 255 subjects were included. Mean patient age was 48.5 years (14-90), BMI 28.5, and 66.7% female and 33.3% male. The majority (95.3%) had papillary thyroid carcinoma (PTC); and 65.5% had prior thyroidectomy. 17.7% of lesions were hypocellular/acellular (suboptimal), and 5.6% nondiagnostic. Patient factors had no association (P >.05). Right-sidedness and hypovascularity were associated with hypocellularity (P <.05). Higher long/short-axis ratio and cystic foci were weakly associated. On multivariate analysis, right-sidedness (odds ratio [OR] 1.99; confidence interval [CI] 1.10-3.57) and lower vascularity score (OR 0.54; CI 0.39-0.73) were predictive of suboptimal sampling.US-FNA has high diagnostic yield and cellular sample rate. Lesion size had no effect. Right-sidedness and lower vascularity scores were predictive of suboptimal tissue. Identifying these features and expected sample adequacy rates can inform management decisions for thyroid cancer patients with cervical lymphadenopathy.

  View details for DOI 10.1002/jum.15688

  View details for PubMedID 33665844

• Locally Delivered Ascorbic Acid and β-Glycerophosphate Augment Local Bone Graft in a Murine Model of 2-Level Posterior Spinal Fusion. International journal of spine surgery Chen, J. V., Lee, K., Tillinghast, K., Halloran, B., Dang, A. B. 2021; 15 (5): 921-928

  Abstract

  Ascorbic acid is involved in collagen biosynthesis and upregulates alkaline phosphatase, potentially alleviating cell senescence and stimulating mesenchymal stem cell proliferation and differentiation into osteoblasts. We hypothesized locally delivered ascorbic acid and β-glycerophosphate act as a bone graft extender to increase the volume of new bone formed in a murine model of posterior lumbar fusion.Collagen sponges were used as delivery vehicles. Sponges were prepared with primary media alone or with the addition of ascorbic acid and β-glycerophosphate. Fresh morselized bone graft from 12 donor mice was used. Twenty-four healthy male C57BL/6 mice underwent an uninstrumented posterior L3-L5 lumbar fusion. One control group received morselized bone only. A second "sponge control" group received morselized bone with the control collagen sponge. The third group received morselized bone and a collagen sponge with ascorbic acid and β-glycerophosphate. Three months postoperatively, the lumbar spine underwent high-resolution micro-computed tomography for analysis of bone formation, density, and bridging fusion.Animals receiving ascorbic acid and β-glycerophosphate had a statistically significant increase in corrected bone volume compared with control and sponge groups, with a 56.3% and 25.4% increase, respectively. Mineralized bone fraction was statistically significantly decreased for animals in the ascorbic acid group compared with control and sponge groups. There was no significant difference in fusion rate between test groups.Locally delivered ascorbic acid and β-glycerophosphate in a murine model of posterior spinal fusion yielded statistically significant increases in new bone formation in the lumbar spine but statistically significant decreases in mineralized bone fraction. Differences in fusion rate were not statistically significant.This study provides early data suggesting that delivery of ascorbic acid to a spinal fusion site may be beneficial but does not yet establish an indication for clinical use. Further studies are needed to determine optimal dose and delivery of ascorbic acid.

  View details for DOI 10.14444/8120

  View details for PubMedID 34649952

  View details for PubMedCentralID PMC8651186

• Comparing cost and print time estimates for six commercially-available 3D printers obtained through slicing software for clinically relevant anatomical models. 3D printing in medicine Chen, J. V., Dang, A. B., Dang, A. 2021; 7 (1): 1

  Abstract

  3D printed patient-specific anatomical models have been applied clinically to orthopaedic care for surgical planning and patient education. The estimated cost and print time per model for 3D printers have not yet been compared with clinically representative models across multiple printing technologies. This study investigates six commercially-available 3D printers: Prusa i3 MK3S, Formlabs Form 2, Formlabs Form 3, LulzBot TAZ 6, Stratasys F370, and Stratasys J750 Digital Anatomy.Seven representative orthopaedic standard tessellation models derived from CT scans were imported into the respective slicing software for each 3D printer. For each printer and corresponding print setting, the slicing software provides a print time and material use estimate. Material quantity was used to calculate estimated model cost. Print settings investigated were infill percentage, layer height, and model orientation on the print bed. The slicing software investigated are Cura LulzBot Edition 3.6.20, GrabCAD Print 1.43, PreForm 3.4.6, and PrusaSlicer 2.2.0.The effect of changing infill between 15% and 20% on estimated print time and material use was negligible. Orientation of the model has considerable impact on time and cost with worst-case differences being as much as 39.30% added print time and 34.56% added costs. Averaged across all investigated settings, horizontal model orientation on the print bed minimizes estimated print time for all 3D printers, while vertical model orientation minimizes cost with the exception of Stratasys J750 Digital Anatomy, in which horizontal orientation also minimized cost. Decreasing layer height for all investigated printers increased estimated print time and decreased estimated cost with the exception of Stratasys F370, in which cost increased. The difference in material cost was two orders of magnitude between the least and most-expensive printers. The difference in build rate (cm3/min) was one order of magnitude between the fastest and slowest printers.All investigated 3D printers in this study have the potential for clinical utility. Print time and print cost are dependent on orientation of anatomy and the printers and settings selected. Cost-effective clinical 3D printing of anatomic models should consider an appropriate printer for the complexity of the anatomy and the experience of the printer technicians.

  View details for DOI 10.1186/s41205-020-00091-4

  View details for PubMedID 33404847

  View details for PubMedCentralID PMC7786189

• Identifying a commercially-available 3D printing process that minimizes model distortion after annealing and autoclaving and the effect of steam sterilization on mechanical strength. 3D printing in medicine Chen, J. V., Tanaka, K. S., Dang, A. B., Dang, A. 2020; 6 (1): 9

  Abstract

  Fused deposition modeling 3D printing is used in medicine for diverse purposes such as creating patient-specific anatomical models and surgical instruments. For use in the sterile surgical field, it is necessary to understand the mechanical behavior of these prints across 3D printing materials and after autoclaving. It has been previously understood that steam sterilization weakens polylactic acid, however, annealing heat treatment of polylactic acid increases its crystallinity and mechanical strength. We aim to identify an optimal and commercially available 3D printing process that minimizes distortion after annealing and autoclaving and to quantify mechanical strength after these interventions.Thirty millimeters cubes with four different infill geometries were 3D printed and subjected to hot water-bath annealing then immediate autoclaving. Seven commercially available 3D printing materials were tested to understand their mechanical behavior after intervention. The dimensions in the X, Y, and Z axes were measured before and after annealing, and again after subsequent autoclaving. Standard and strength-optimized Army-Navy retractor designs were printed using the 3D printing material and infill geometry that deformed the least. These retractors were subjected to annealing and autoclaving interventions and tested for differences in mechanical strength.For both the annealing and subsequent autoclaving intervention, the material and infill geometry that deformed the least, respectively, was Essentium PLA Gray and "grid". Standard retractors without intervention failed at 95 N +/- 2.4 N. Annealed retractors failed at 127.3 N +/- 10 N. Autoclave only retractors failed at 15.7 N +/- 1.4 N. Annealed then autoclaved retractors failed at 19.8 N +/- 3.1 N. Strength-optimized retractors, after the annealing then autoclaving intervention, failed at 164.8 N +/- 12.5 N.For 30 mm cubes, the 3D printing material and infill geometry that deformed the least, respectively, was Essentium PLA and "grid". Hot water-bath annealing results in increased 3D printed model strength, however autoclaving 3D prints markedly diminishes strength. Strength-optimized 3D printed PLA Army-Navy retractors overcome the strength limitation due to autoclaving.

  View details for DOI 10.1186/s41205-020-00062-9

  View details for PubMedID 32297041

  View details for PubMedCentralID PMC7161250

• 3D printed PLA Army-Navy retractors when used as linear retractors yield clinically acceptable tolerances. 3D printing in medicine Chen, J. V., Dang, A. B., Lee, C. S., Dang, A. B. 2019; 5 (1): 16

  Abstract

  Modern low-cost 3D printing technologies offer the promise of access to surgical tools in resource scarce areas, however optimal designs for manufacturing have not yet been established. We explore how the optimization of 3D printing parameters when manufacturing polylactic acid filament based Army-Navy retractors vastly increases the strength of retractors, and investigate sources of variability in retractor strength, material cost, printing time, and parameter limitations.Standard retractors were printed from various polylactic acid filament spools intra-manufacturer and inter-manufacturer to measure variability in retractor strength. Printing parameters were systematically varied to determine optimum printing parameters. These parameters include retractor width, thickness, infill percentage, infill geometry, perimeter number, and a reinforced joint design. Estimated retractor mass from computer models allows us to estimate material cost.We found statistically significant differences in retractor strength between spools of the same manufacturer and between manufacturers. We determined the true strength optimized retractor to have 30% infill, 3 perimeters, 0.25 in. thickness, 0.75 in. width, and has "Triangle" infill geometry and reinforced joints, failing at more than 15X the threshold for clinically excessive retraction and costs $1.25 USD.The optimization of 3D printed Army-Navy retractors greatly improve the efficacy of this instrument and expedite the adoption of 3D printing technology in many diverse fields in medicine not necessarily limited to resource poor settings.

  View details for DOI 10.1186/s41205-019-0053-z

  View details for PubMedID 31754879

  View details for PubMedCentralID PMC6873412