Publications by authors named "Bradley J Erickson"

101 Publications

Semantic Instance Segmentation of Kidney Cysts in MR Images: A Fully Automated 3D Approach Developed Through Active Learning.

J Digit Imaging 2021 Apr 5. Epub 2021 Apr 5.

Division of Nephrology and Hypertension, Mayo Clinic, Rochester, MN 55905, USA.

Total kidney volume (TKV) is the main imaging biomarker used to monitor disease progression and to classify patients affected by autosomal dominant polycystic kidney disease (ADPKD) for clinical trials. However, patients with similar TKVs may have drastically different cystic presentations and phenotypes. In an effort to quantify these cystic differences, we developed the first 3D semantic instance cyst segmentation algorithm for kidneys in MR images. We have reformulated both the object detection/localization task and the instance-based segmentation task into a semantic segmentation task. This allowed us to solve this unique imaging problem efficiently, even for patients with thousands of cysts. To do this, a convolutional neural network (CNN) was trained to learn cyst edges and cyst cores. Images were converted from instance cyst segmentations to semantic edge-core segmentations by applying a 3D erosion morphology operator to up-sampled versions of the images. The reduced cysts were labeled as core; the eroded areas were dilated in 2D and labeled as edge. The network was trained on 30 MR images and validated on 10 MR images using a fourfold cross-validation procedure. The final ensemble model was tested on 20 MR images not seen during the initial training/validation. The results from the test set were compared to segmentations from two readers. The presented model achieved an averaged R value of 0.94 for cyst count, 1.00 for total cyst volume, 0.94 for cystic index, and an averaged Dice coefficient of 0.85. These results demonstrate the feasibility of performing cyst segmentations automatically in ADPKD patients.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1007/s10278-021-00452-3DOI Listing
April 2021

Generative Adversarial Networks to Synthesize Missing T1 and FLAIR MRI Sequences for Use in a Multisequence Brain Tumor Segmentation Model.

Radiology 2021 Mar 9:203786. Epub 2021 Mar 9.

From the Departments of Radiology (G.M.C., K.A.P., J.C.C., B.J.E.), Neurology (D.H.L., W.O.T.), Laboratory Medicine and Pathology (R.B.J.), and Health Sciences Research (J.E.E.P.), Mayo Clinic, 200 First St SW, Rochester, MN 55905; Department of Health Sciences Research, Mayo Clinic, Jacksonville, Fla (A.D.W.); Mayo Clinic Alix School of Medicine, Rochester, Minn (D.C.V.); Neuroradiology Unit, Scientific Institute for Research, Hospitalization, and Healthcare San Raffaele Scientific Institute, Milan, Italy (M.B.); and Department of Clinical, Surgical, Diagnostic, and Pediatric Sciences, University of Pavia, Pavia, Italy (F.S.).

Background Missing MRI sequences represent an obstacle in the development and use of deep learning (DL) models that require multiple inputs. Purpose To determine if synthesizing brain MRI scans using generative adversarial networks (GANs) allows for the use of a DL model for brain lesion segmentation that requires T1-weighted images, postcontrast T1-weighted images, fluid-attenuated inversion recovery (FLAIR) images, and T2-weighted images. Materials and Methods In this retrospective study, brain MRI scans obtained between 2011 and 2019 were collected, and scenarios were simulated in which the T1-weighted images and FLAIR images were missing. Two GANs were trained, validated, and tested using 210 glioblastomas (GBMs) (Multimodal Brain Tumor Image Segmentation Benchmark [BRATS] 2017) to generate T1-weighted images from postcontrast T1-weighted images and FLAIR images from T2-weighted images. The quality of the generated images was evaluated with mean squared error (MSE) and the structural similarity index (SSI). The segmentations obtained with the generated scans were compared with those obtained with the original MRI scans using the dice similarity coefficient (DSC). The GANs were validated on sets of GBMs and central nervous system lymphomas from the authors' institution to assess their generalizability. Statistical analysis was performed using the Mann-Whitney, Friedman, and Dunn tests. Results Two hundred ten GBMs from the BRATS data set and 46 GBMs (mean patient age, 58 years ± 11 [standard deviation]; 27 men [59%] and 19 women [41%]) and 21 central nervous system lymphomas (mean patient age, 67 years ± 13; 12 men [57%] and nine women [43%]) from the authors' institution were evaluated. The median MSE for the generated T1-weighted images ranged from 0.005 to 0.013, and the median MSE for the generated FLAIR images ranged from 0.004 to 0.103. The median SSI ranged from 0.82 to 0.92 for the generated T1-weighted images and from 0.76 to 0.92 for the generated FLAIR images. The median DSCs for the segmentation of the whole lesion, the FLAIR hyperintensities, and the contrast-enhanced areas using the generated scans were 0.82, 0.71, and 0.92, respectively, when replacing both T1-weighted and FLAIR images; 0.84, 0.74, and 0.97 when replacing only the FLAIR images; and 0.97, 0.95, and 0.92 when replacing only the T1-weighted images. Conclusion Brain MRI scans generated using generative adversarial networks can be used as deep learning model inputs in case MRI sequences are missing. © RSNA, 2021 See also the editorial by Zhong in this issue.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1148/radiol.2021203786DOI Listing
March 2021

A Deep Learning Tool for Automated Radiographic Measurement of Acetabular Component Inclination and Version After Total Hip Arthroplasty.

J Arthroplasty 2021 Feb 16. Epub 2021 Feb 16.

Department of Health Sciences Research, Mayo Clinic, Rochester, MN; Department of Orthopedic Surgery, Mayo Clinic, Rochester, MN.

Background: Inappropriate acetabular component angular position is believed to increase the risk of hip dislocation after total hip arthroplasty. However, manual measurement of these angles is time consuming and prone to interobserver variability. The purpose of this study was to develop a deep learning tool to automate the measurement of acetabular component angles on postoperative radiographs.

Methods: Two cohorts of 600 anteroposterior (AP) pelvis and 600 cross-table lateral hip postoperative radiographs were used to develop deep learning models to segment the acetabular component and the ischial tuberosities. Cohorts were manually annotated, augmented, and randomly split to train-validation-test data sets on an 8:1:1 basis. Two U-Net convolutional neural network models (one for AP and one for cross-table lateral radiographs) were trained for 50 epochs. Image processing was then deployed to measure the acetabular component angles on the predicted masks for anatomical landmarks. Performance of the tool was tested on 80 AP and 80 cross-table lateral radiographs.

Results: The convolutional neural network models achieved a mean Dice similarity coefficient of 0.878 and 0.903 on AP and cross-table lateral test data sets, respectively. The mean difference between human-level and machine-level measurements was 1.35° (σ = 1.07°) and 1.39° (σ = 1.27°) for the inclination and anteversion angles, respectively. Differences of 5⁰ or more between human-level and machine-level measurements were observed in less than 2.5% of cases.

Conclusion: We developed a highly accurate deep learning tool to automate the measurement of angular position of acetabular components for use in both clinical and research settings.

Level Of Evidence: III.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1016/j.arth.2021.02.026DOI Listing
February 2021

Deep Learning Artificial Intelligence Model for Assessment of Hip Dislocation Risk Following Primary Total Hip Arthroplasty From Postoperative Radiographs.

J Arthroplasty 2021 Feb 16. Epub 2021 Feb 16.

Department of Radiology, Radiology Informatics Laboratory, Mayo Clinic, Rochester, MN.

Background: Dislocation is a common complication following total hip arthroplasty (THA), and accounts for a high percentage of subsequent revisions. The purpose of this study is to illustrate the potential of a convolutional neural network model to assess the risk of hip dislocation based on postoperative anteroposterior pelvis radiographs.

Methods: We retrospectively evaluated radiographs for a cohort of 13,970 primary THAs with 374 dislocations over 5 years of follow-up. Overall, 1490 radiographs from dislocated and 91,094 from non-dislocated THAs were included in the analysis. A convolutional neural network object detection model (YOLO-V3) was trained to crop the images by centering on the femoral head. A ResNet18 classifier was trained to predict subsequent hip dislocation from the cropped imaging. The ResNet18 classifier was initialized with ImageNet weights and trained using FastAI (V1.0) running on PyTorch. The training was run for 15 epochs using 10-fold cross validation, data oversampling, and augmentation.

Results: The hip dislocation classifier achieved the following mean performance (standard deviation): accuracy = 49.5 (4.1%), sensitivity = 89.0 (2.2%), specificity = 48.8 (4.2%), positive predictive value = 3.3 (0.3%), negative predictive value = 99.5 (0.1%), and area under the receiver operating characteristic curve = 76.7 (3.6%). Saliency maps demonstrated that the model placed the greatest emphasis on the femoral head and acetabular component.

Conclusion: Existing prediction methods fail to identify patients at high risk of dislocation following THA. Our radiographic classifier model has high sensitivity and negative predictive value, and can be combined with clinical risk factor information for rapid assessment of risk for dislocation following THA. The model further suggests radiographic locations which may be important in understanding the etiology of prosthesis dislocation. Importantly, our model is an illustration of the potential of automated imaging artificial intelligence models in orthopedics.

Level Of Evidence: Level III.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1016/j.arth.2021.02.028DOI Listing
February 2021

Automated segmentation of endometrial cancer on MR images using deep learning.

Sci Rep 2021 Jan 8;11(1):179. Epub 2021 Jan 8.

Department of Radiology, MMIV Mohn Medical Imaging and Visualization Centre, Haukeland University Hospital, Bergen, Norway.

Preoperative MR imaging in endometrial cancer patients provides valuable information on local tumor extent, which routinely guides choice of surgical procedure and adjuvant therapy. Furthermore, whole-volume tumor analyses of MR images may provide radiomic tumor signatures potentially relevant for better individualization and optimization of treatment. We apply a convolutional neural network for automatic tumor segmentation in endometrial cancer patients, enabling automated extraction of tumor texture parameters and tumor volume. The network was trained, validated and tested on a cohort of 139 endometrial cancer patients based on preoperative pelvic imaging. The algorithm was able to retrieve tumor volumes comparable to human expert level (likelihood-ratio test, [Formula: see text]). The network was also able to provide a set of segmentation masks with human agreement not different from inter-rater agreement of human experts (Wilcoxon signed rank test, [Formula: see text], [Formula: see text], and [Formula: see text]). An automatic tool for tumor segmentation in endometrial cancer patients enables automated extraction of tumor volume and whole-volume tumor texture features. This approach represents a promising method for automatic radiomic tumor profiling with potential relevance for better prognostication and individualization of therapeutic strategy in endometrial cancer.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1038/s41598-020-80068-9DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC7794479PMC
January 2021

The RSNA International COVID-19 Open Radiology Database (RICORD).

Radiology 2021 04 5;299(1):E204-E213. Epub 2021 Jan 5.

From the Department of Radiology, Stanford University, Stanford, Calif (E.B.T., J.S., B.P.P.); Department of Radiology, University of Pennsylvania Hospital, Philadelphia, Pa (S.S., M. Hershman, L.R.); Department of Radiology, Stanford University School of Medicine, Stanford University Medical Center, 725 Welch Rd, Room 1675, Stanford, CA 94305-5913 (M.P.L.); Department of Medical Imaging, University of Toronto, Unity Health Toronto, Toronto, Canada (E.C.); Department of Radiology, Mayo Clinic, Rochester, Minn (B.J.E., P.R.); Department of Radiology, Weill Cornell Medicine, New York, NY (G.S.); MD.ai, New York, NY (A.S.); Department of Radiology, Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital, Harvard Medical School, Charlestown, Mass (J.K.C.); Department of Diagnostic and Interventional Radiology, Cairo University Kasr Alainy Faculty of Medicine, Cairo, Egypt (M. Hafez); Department of Radiology, The Ottawa Hospital, Ottawa, Canada (S.J.); Department of Radiology and Biomedical Imaging, Center for Intelligent Imaging, San Francisco, Calif (J.M.); Department of Radiology, Koç University School of Medicine, Koç University Hospital, Istanbul, Turkey (E.A.); Department of Radiology, ETZ Hospital, Tilburg, the Netherlands (E.R.R.); Department of Radiology, University of Ghent, Ghent, Belgium (E.R.R.); Department of Diagnostic Imaging, Universidade Federal de São Paulo, São Paulo, Brazil (F.C.K.); Department of Radiology, Netherlands Cancer Institute, Amsterdam, the Netherlands (L.T.); Department of Radiology, NYU Grossman School of Medicine, Center for Advanced Imaging Innovation and Research, Laura and Isaac Perlmutter Cancer Center, New York, NY (L.M.); Department of Radiology, University of Wisconsin School of Medicine and Public Health, Madison, Wis (J.P.K.); and Department of Thoracic Imaging, University of Texas MD Anderson Cancer Center, Houston, Tex (C.C.W.).

The coronavirus disease 2019 (COVID-19) pandemic is a global health care emergency. Although reverse-transcription polymerase chain reaction testing is the reference standard method to identify patients with COVID-19 infection, chest radiography and CT play a vital role in the detection and management of these patients. Prediction models for COVID-19 imaging are rapidly being developed to support medical decision making. However, inadequate availability of a diverse annotated data set has limited the performance and generalizability of existing models. To address this unmet need, the RSNA and Society of Thoracic Radiology collaborated to develop the RSNA International COVID-19 Open Radiology Database (RICORD). This database is the first multi-institutional, multinational, expert-annotated COVID-19 imaging data set. It is made freely available to the machine learning community as a research and educational resource for COVID-19 chest imaging. Pixel-level volumetric segmentation with clinical annotations was performed by thoracic radiology subspecialists for all COVID-19-positive thoracic CT scans. The labeling schema was coordinated with other international consensus panels and COVID-19 data annotation efforts, the European Society of Medical Imaging Informatics, the American College of Radiology, and the American Association of Physicists in Medicine. Study-level COVID-19 classification labels for chest radiographs were annotated by three radiologists, with majority vote adjudication by board-certified radiologists. RICORD consists of 240 thoracic CT scans and 1000 chest radiographs contributed from four international sites. It is anticipated that RICORD will ideally lead to prediction models that can demonstrate sustained performance across populations and health care systems.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1148/radiol.2021203957DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC7993245PMC
April 2021

Automated Aneurysm Detection: Emerging from the Shallow End of the Deep Learning Pool.

Radiology 2021 01 3;298(1):164-165. Epub 2020 Nov 3.

From the Department of Radiology, Mayo Clinic, 200 First St SW, Rochester, MN 55905.

View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1148/radiol.2020203853DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC7771990PMC
January 2021

Clinical, biological, radiological, and pathological comparison of sparsely and densely granulated somatotroph adenomas: a single center experience from a cohort of 131 patients with acromegaly.

Pituitary 2021 Apr 19;24(2):192-206. Epub 2020 Oct 19.

Division of Anatomic Pathology, Mayo Clinic, Rochester, MN, USA.

Purpose: Growth hormone-producing pituitary adenomas are divided into two clinically relevant histologic subtypes, densely (DG-A) and sparsely (SG-A) granulated. Histologic subtype was evaluated in a large cohort of patients with acromegaly, separating DG-A and SG-A, and correlated with clinicopathological characteristics.

Methods: Patients with acromegaly undergoing surgery as initial therapy between 1995 and 2015 were identified. Histologic subtype was determined by keratin expression pattern with CAM5.2 and correlated with clinical and imaging parameters, somatostatin receptor subtype 2 (SST2) expression, post-surgical remission rate, and application of a prognostic scoring system incorporating proliferation and invasiveness.

Results: One hundred thirty-one patients were included. Tumors were classified as DG-A (75, 57.3%), SG-A (29, 22.1%), intermediate (I-A) (9, 6.9%), and unclassified (18, 13.7%) when CAM5.2 was negative. DG-A and I-A were combined for analysis (DG/I-A) and compared to SG-A. Age, gender, proliferation, and post-surgical remission did not differ. SG-A were larger [2 vs. 1.5 cm (median), p = 0.03], more frequently invasive [65.5% vs. 32.9%, p = 0.004], associated with higher MRI T2-weighted signal ratio [1.01 vs. 0.82 (median), p = 0.01], showed lower SST2 expression (p < 0.0001), and scored higher in the prognostic classification (p = 0.004). Surgical remission occurred in 41.7% DG/I-A and 41.4% SG-A (p = 1.0). On multivariate analysis, absence of invasion (p = 0.009) and lower pre-operative IGF-1 index (p = 0.0002) were associated with post-surgical remission.

Conclusion: CAM5.2 allowed distinction between DG/I-A and SG-A in most but not all cases. Histologic subtype did not predict surgical outcome. Absence of invasion and lower pre-operative IGF-1 index were the only significant predictors of post-surgical remission in this cohort.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1007/s11102-020-01096-2DOI Listing
April 2021

Magician's Corner: 7. Using Convolutional Neural Networks to Reduce Noise in Medical Images.

Radiol Artif Intell 2020 Sep 30;2(5):e200036. Epub 2020 Sep 30.

Department of Radiology, Mayo Clinic, 200 First St SW, Rochester, MN 55905.

This article shows how to train a convolutional neural network to reduce noise in CT images, although the principles apply to medical and nonmedical images; authors also explore mathematical and visually weighted loss functions to adjust the appearance.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1148/ryai.2020200036DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC7529432PMC
September 2020

Automatic semantic segmentation of kidney cysts in MR images of patients affected by autosomal-dominant polycystic kidney disease.

Abdom Radiol (NY) 2021 03 17;46(3):1053-1061. Epub 2020 Sep 17.

Department of Radiology, Mayo Clinic, Rochester, MN, 55905, USA.

Purpose: For patients affected by autosomal-dominant polycystic kidney disease (ADPKD), successful differentiation of cysts is useful for automatic classification of patient phenotypes, clinical decision-making, and disease progression. The objective was to develop and evaluate a fully automated semantic segmentation method to differentiate and analyze renal cysts in patients with ADPKD.

Methods: An automated deep learning approach using a convolutional neural network was trained, validated, and tested on a set of 60 MR T2-weighted images. A three-fold cross-validation approach was used to train three models on distinct training and validation sets (n = 40). An ensemble model was then built and tested on the hold out cases (n = 20), with each of the cases compared to manual segmentations performed by two readers. Segmentation agreement between readers and the automated method was assessed.

Results: The automated approach was found to perform at the level of interobserver variability. The automated approach had a Dice coefficient (mean ± standard deviation) of 0.86 ± 0.10 vs Reader-1 and 0.84 ± 0.11 vs. Reader-2. Interobserver Dice was 0.86 ± 0.08. In terms of total cyst volume (TCV), the automated approach had a percent difference of 3.9 ± 19.1% vs Reader-1 and 8.0 ± 24.1% vs Reader-2, whereas interobserver variability was - 2.0 ± 16.4%.

Conclusion: This study developed and validated a fully automated approach for performing semantic segmentation of kidney cysts in MR images of patients affected by ADPKD. This approach will be useful for exploring additional imaging biomarkers of ADPKD and automatically classifying phenotypes.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1007/s00261-020-02748-4DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC7940295PMC
March 2021

Complete abdomen and pelvis segmentation using U-net variant architecture.

Med Phys 2020 Nov 7;47(11):5609-5618. Epub 2020 Oct 7.

Department of Radiology, Mayo Clinic, 200 First St SW, Rochester, MN, 55905, USA.

Purpose: Organ segmentation of computed tomography (CT) imaging is essential for radiotherapy treatment planning. Treatment planning requires segmentation not only of the affected tissue, but nearby healthy organs-at-risk, which is laborious and time-consuming. We present a fully automated segmentation method based on the three-dimensional (3D) U-Net convolutional neural network (CNN) capable of whole abdomen and pelvis segmentation into 33 unique organ and tissue structures, including tissues that may be overlooked by other automated segmentation approaches such as adipose tissue, skeletal muscle, and connective tissue and vessels. Whole abdomen segmentation is capable of quantifying exposure beyond a handful of organs-at-risk to all tissues within the abdomen.

Methods: Sixty-six (66) CT examinations of 64 individuals were included in the training and validation sets and 18 CT examinations from 16 individuals were included in the test set. All pixels in each examination were segmented by image analysts (with physician correction) and assigned one of 33 labels. Segmentation was performed with a 3D U-Net variant architecture which included residual blocks, and model performance was quantified on 18 test cases. Human interobserver variability (using semiautomated segmentation) was also reported on two scans, and manual interobserver variability of three individuals was reported on one scan. Model performance was also compared to several of the best models reported in the literature for multiple organ segmentation.

Results: The accuracy of the 3D U-Net model ranges from a Dice coefficient of 0.95 in the liver, 0.93 in the kidneys, 0.79 in the pancreas, 0.69 in the adrenals, and 0.51 in the renal arteries. Model accuracy is within 5% of human segmentation in eight of 19 organs and within 10% accuracy in 13 of 19 organs.

Conclusions: The CNN approaches the accuracy of human tracers and on certain complex organs displays more consistent prediction than human tracers. Fully automated deep learning-based segmentation of CT abdomen has the potential to improve both the speed and accuracy of radiotherapy dose prediction for organs-at-risk.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1002/mp.14422DOI Listing
November 2020

Evaluating the Use of rCBV as a Tumor Grade and Treatment Response Classifier Across NCI Quantitative Imaging Network Sites: Part II of the DSC-MRI Digital Reference Object (DRO) Challenge.

Tomography 2020 06;6(2):203-208

Division of Neuroimaging Research, Barrow Neurological Institute, Phoenix, AZ.

We have previously characterized the reproducibility of brain tumor relative cerebral blood volume (rCBV) using a dynamic susceptibility contrast magnetic resonance imaging digital reference object across 12 sites using a range of imaging protocols and software platforms. As expected, reproducibility was highest when imaging protocols and software were consistent, but decreased when they were variable. Our goal in this study was to determine the impact of rCBV reproducibility for tumor grade and treatment response classification. We found that varying imaging protocols and software platforms produced a range of optimal thresholds for both tumor grading and treatment response, but the performance of these thresholds was similar. These findings further underscore the importance of standardizing acquisition and analysis protocols across sites and software benchmarking.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.18383/j.tom.2020.00012DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC7289259PMC
June 2020

Consensus recommendations for a dynamic susceptibility contrast MRI protocol for use in high-grade gliomas.

Neuro Oncol 2020 09;22(9):1262-1275

Department of Biophysics, Medical College of Wisconsin, Milwaukee, Wisconsin, USA.

Despite the widespread clinical use of dynamic susceptibility contrast (DSC) MRI, DSC-MRI methodology has not been standardized, hindering its utilization for response assessment in multicenter trials. Recently, the DSC-MRI Standardization Subcommittee of the Jumpstarting Brain Tumor Drug Development Coalition issued an updated consensus DSC-MRI protocol compatible with the standardized brain tumor imaging protocol (BTIP) for high-grade gliomas that is increasingly used in the clinical setting and is the default MRI protocol for the National Clinical Trials Network. After reviewing the basis for controversy over DSC-MRI protocols, this paper provides evidence-based best practices for clinical DSC-MRI as determined by the Committee, including pulse sequence (gradient echo vs spin echo), BTIP-compliant contrast agent dosing (preload and bolus), flip angle (FA), echo time (TE), and post-processing leakage correction. In summary, full-dose preload, full-dose bolus dosing using intermediate (60°) FA and field strength-dependent TE (40-50 ms at 1.5 T, 20-35 ms at 3 T) provides overall best accuracy and precision for cerebral blood volume estimates. When single-dose contrast agent usage is desired, no-preload, full-dose bolus dosing using low FA (30°) and field strength-dependent TE provides excellent performance, with reduced contrast agent usage and elimination of potential systematic errors introduced by variations in preload dose and incubation time.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1093/neuonc/noaa141DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC7523451PMC
September 2020

Association of Maximal Extent of Resection of Contrast-Enhanced and Non-Contrast-Enhanced Tumor With Survival Within Molecular Subgroups of Patients With Newly Diagnosed Glioblastoma.

JAMA Oncol 2020 04;6(4):495-503

Department of Neurological Surgery, University of California, San Francisco.

Importance: Per the World Health Organization 2016 integrative classification, newly diagnosed glioblastomas are separated into isocitrate dehydrogenase gene 1 or 2 (IDH)-wild-type and IDH-mutant subtypes, with median patient survival of 1.2 and 3.6 years, respectively. Although maximal resection of contrast-enhanced (CE) tumor is associated with longer survival, the prognostic importance of maximal resection within molecular subgroups and the potential importance of resection of non-contrast-enhanced (NCE) disease is poorly understood.

Objective: To assess the association of resection of CE and NCE tumors in conjunction with molecular and clinical information to develop a new road map for cytoreductive surgery.

Design, Setting, And Participants: This retrospective, multicenter cohort study included a development cohort from the University of California, San Francisco (761 patients diagnosed from January 1, 1997, through December 31, 2017, with 9.6 years of follow-up) and validation cohorts from the Mayo Clinic (107 patients diagnosed from January 1, 2004, through December 31, 2014, with 5.7 years of follow-up) and the Ohio Brain Tumor Study (99 patients with data collected from January 1, 2008, through December 31, 2011, with a median follow-up of 10.9 months). Image accessors were blinded to patient groupings. Eligible patients underwent surgical resection for newly diagnosed glioblastoma and had available survival, molecular, and clinical data and preoperative and postoperative magnetic resonance images. Data were analyzed from November 15, 2018, to March 15, 2019.

Main Outcomes And Measures: Overall survival.

Results: Among the 761 patients included in the development cohort (468 [61.5%] men; median age, 60 [interquartile range, 51.6-67.7] years), younger patients with IDH-wild-type tumors and aggressive resection of CE and NCE tumors had survival similar to that of patients with IDH-mutant tumors (median overall survival [OS], 37.3 [95% CI, 31.6-70.7] months). Younger patients with IDH-wild-type tumors and reduction of CE tumor but residual NCE tumors fared worse (median OS, 16.5 [95% CI, 14.7-18.3] months). Older patients with IDH-wild-type tumors benefited from reduction of CE tumor (median OS, 12.4 [95% CI, 11.4-14.0] months). The results were validated in the 2 external cohorts. The association between aggressive CE and NCE in patients with IDH-wild-type tumors was not attenuated by the methylation status of the promoter region of the DNA repair enzyme O6-methylguanine-DNA methyltransferase.

Conclusions And Relevance: This study confirms an association between maximal resection of CE tumor and OS in patients with glioblastoma across all subgroups. In addition, maximal resection of NCE tumor was associated with longer OS in younger patients, regardless of IDH status, and among patients with IDH-wild-type glioblastoma regardless of the methylation status of the promoter region of the DNA repair enzyme O6-methylguanine-DNA methyltransferase. These conclusions may help reassess surgical strategies for individual patients with newly diagnosed glioblastoma.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1001/jamaoncol.2019.6143DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC7042822PMC
April 2020

Can my computer tell me if this tumor is IDH mutated?

Neuro Oncol 2020 03;22(3):311-312

Department of Radiology, Mayo Clinic, Rochester, Minnesota.

View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1093/neuonc/noaa002DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC7058444PMC
March 2020

A Survey of Deep-Learning Applications in Ultrasound: Artificial Intelligence-Powered Ultrasound for Improving Clinical Workflow.

J Am Coll Radiol 2019 Sep;16(9 Pt B):1318-1328

Radiology Informatics Lab, Department of Radiology, Mayo Clinic, Rochester, Minnesota.

Ultrasound is the most commonly used imaging modality in clinical practice because it is a nonionizing, low-cost, and portable point-of-care imaging tool that provides real-time images. Artificial intelligence (AI)-powered ultrasound is becoming more mature and getting closer to routine clinical applications in recent times because of an increased need for efficient and objective acquisition and evaluation of ultrasound images. Because ultrasound images involve operator-, patient-, and scanner-dependent variations, the adaptation of classical machine learning methods to clinical applications becomes challenging. With their self-learning ability, deep-learning (DL) methods are able to harness exponentially growing graphics processing unit computing power to identify abstract and complex imaging features. This has given rise to tremendous opportunities such as providing robust and generalizable AI models for improving image acquisition, real-time assessment of image quality, objective diagnosis and detection of diseases, and optimizing ultrasound clinical workflow. In this report, the authors review current DL approaches and research directions in rapidly advancing ultrasound technology and present their outlook on future directions and trends for DL techniques to further improve diagnosis, reduce health care cost, and optimize ultrasound clinical workflow.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1016/j.jacr.2019.06.004DOI Listing
September 2019

Cerebral blood volume and apparent diffusion coefficient - Valuable predictors of non-response to bevacizumab treatment in patients with recurrent glioblastoma.

J Neurol Sci 2019 Oct 23;405:116433. Epub 2019 Aug 23.

Department of Radiology, Mayo Clinic, 200 First St SW, Rochester, MN 55905, United States of America. Electronic address:

Background: Glioblastoma multiforme (GBM) is the most common primary brain tumor in adults. The core of standard of care for newly diagnosed GBM was established in 2005 and includes maximum feasible surgical resection followed by radiation and temozolomide, with subsequent temozolomide with or without tumor-treating fields. Unfortunately, nearly all patients experience a recurrence. Bevacizumab (BV) is a commonly used second-line agent for such recurrences, but it has not been shown to impact overall survival, and short-term response is variable.

Methods: We collected MRI perfusion and diffusion images from 54 subjects with recurrent GBM treated only with radiation and temozolomide. They were subsequently treated with BV. Using machine learning, we created a model to predict short term response (6 months) and overall survival. We set time thresholds to maximize the separation of responders/survivors versus non-responders/short survivors.

Results: We were able to segregate 21 (68%) of 31 subjects into unlikely to respond categories based on Progression Free Survival at 6 months (PFS6) criteria. Twenty-two (69%) of 32 subjects could similarly be identified as unlikely to survive long using the machine learning algorithm.

Conclusion: With the use of machine learning techniques to evaluate imaging features derived from pre- and post-treatment multimodal MRI, it is possible to identify an important fraction of patients who are either highly unlikely to respond, or highly likely to respond. This can be helpful is selecting patients that either should or should not be treated with BV.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1016/j.jns.2019.116433DOI Listing
October 2019

Automatic Measurement of Kidney and Liver Volumes from MR Images of Patients Affected by Autosomal Dominant Polycystic Kidney Disease.

J Am Soc Nephrol 2019 08 3;30(8):1514-1522. Epub 2019 Jul 3.

Division of Nephrology and Hypertension, Department of Internal Medicine;

Background: The formation and growth of cysts in kidneys, and often liver, in autosomal dominant polycystic kidney disease (ADPKD) cause progressive increases in total kidney volume (TKV) and liver volume (TLV). Laborious and time-consuming manual tracing of kidneys and liver is the current gold standard. We developed a fully automated segmentation method for TKV and TLV measurement that uses a deep learning network optimized to perform semantic segmentation of kidneys and liver.

Methods: We used 80% of a set of 440 abdominal magnetic resonance images (T2-weighted HASTE coronal sequences) from patients with ADPKD to train the network and the remaining 20% for validation. Both kidneys and liver were also segmented manually. To evaluate the method's performance, we used an additional test set of images from 100 patients, 45 of whom were also involved in longitudinal analyses.

Results: TKV and TLV measured by the automated approach correlated highly with manually traced TKV and TLV (intraclass correlation coefficients, 0.998 and 0.996, respectively), with low bias and high precision (<0.1%±2.7% for TKV and -1.6%±3.1% for TLV); this was comparable with inter-reader variability of manual tracing (<0.1%±3.5% for TKV and -1.5%±4.8% for TLV). For longitudinal analysis, bias and precision were <0.1%±3.2% for TKV and 1.4%±2.9% for TLV growth.

Conclusions: These findings demonstrate a fully automated segmentation method that measures TKV, TLV, and changes in these parameters as accurately as manual tracing. This technique may facilitate future studies in which automated and reproducible TKV and TLV measurements are needed to assess disease severity, disease progression, and treatment response.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1681/ASN.2018090902DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC6683702PMC
August 2019

RIL-Contour: a Medical Imaging Dataset Annotation Tool for and with Deep Learning.

J Digit Imaging 2019 08;32(4):571-581

Radiology Informatics Laboratory, Department of Radiology, Mayo Clinic, Rochester, MN, USA.

Deep-learning algorithms typically fall within the domain of supervised artificial intelligence and are designed to "learn" from annotated data. Deep-learning models require large, diverse training datasets for optimal model convergence. The effort to curate these datasets is widely regarded as a barrier to the development of deep-learning systems. We developed RIL-Contour to accelerate medical image annotation for and with deep-learning. A major goal driving the development of the software was to create an environment which enables clinically oriented users to utilize deep-learning models to rapidly annotate medical imaging. RIL-Contour supports using fully automated deep-learning methods, semi-automated methods, and manual methods to annotate medical imaging with voxel and/or text annotations. To reduce annotation error, RIL-Contour promotes the standardization of image annotations across a dataset. RIL-Contour accelerates medical imaging annotation through the process of annotation by iterative deep learning (AID). The underlying concept of AID is to iteratively annotate, train, and utilize deep-learning models during the process of dataset annotation and model development. To enable this, RIL-Contour supports workflows in which multiple-image analysts annotate medical images, radiologists approve the annotations, and data scientists utilize these annotations to train deep-learning models. To automate the feedback loop between data scientists and image analysts, RIL-Contour provides mechanisms to enable data scientists to push deep newly trained deep-learning models to other users of the software. RIL-Contour and the AID methodology accelerate dataset annotation and model development by facilitating rapid collaboration between analysts, radiologists, and engineers.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1007/s10278-019-00232-0DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC6646456PMC
August 2019

A Roadmap for Foundational Research on Artificial Intelligence in Medical Imaging: From the 2018 NIH/RSNA/ACR/The Academy Workshop.

Radiology 2019 06 16;291(3):781-791. Epub 2019 Apr 16.

From the Department of Radiology, Stanford University, Stanford, CA 94305 (C.P.L., M.P.L.); Department of Radiology, Grandview Medical Center, Birmingham, Ala (B.A.); Department of Radiology, Mayo Clinic, Rochester, Minn (B.J.E.); Department of Radiology, Massachusetts General Hospital, Harvard Medical School, Boston, Mass (J.K.C.); GE Healthcare, Chicago, Ill (K.B.); Department of Radiology, Hospital of the University of Pennsylvania, Philadelphia, Pa (T.S.C., J.D.R.); Department of Radiology, Thomas Jefferson University Hospital, Philadelphia, Pa (A.E.F.); Department of Radiology, Icahn School of Medicine at Mount Sinai, New York, NY (D.S.M.); Biomedical Imaging Center, Rensselaer Polytechnic Institute, Troy, NY (G.W.); and National Institute of Biomedical Imaging and Bioengineering, National Institutes of Health, Washington, DC (K.K.).

Imaging research laboratories are rapidly creating machine learning systems that achieve expert human performance using open-source methods and tools. These artificial intelligence systems are being developed to improve medical image reconstruction, noise reduction, quality assurance, triage, segmentation, computer-aided detection, computer-aided classification, and radiogenomics. In August 2018, a meeting was held in Bethesda, Maryland, at the National Institutes of Health to discuss the current state of the art and knowledge gaps and to develop a roadmap for future research initiatives. Key research priorities include: 1, new image reconstruction methods that efficiently produce images suitable for human interpretation from source data; 2, automated image labeling and annotation methods, including information extraction from the imaging report, electronic phenotyping, and prospective structured image reporting; 3, new machine learning methods for clinical imaging data, such as tailored, pretrained model architectures, and federated machine learning methods; 4, machine learning methods that can explain the advice they provide to human users (so-called explainable artificial intelligence); and 5, validated methods for image de-identification and data sharing to facilitate wide availability of clinical imaging data sets. This research roadmap is intended to identify and prioritize these needs for academic research laboratories, funding agencies, professional societies, and industry.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1148/radiol.2019190613DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC6542624PMC
June 2019

Evaluating Multisite rCBV Consistency from DSC-MRI Imaging Protocols and Postprocessing Software Across the NCI Quantitative Imaging Network Sites Using a Digital Reference Object (DRO).

Tomography 2019 03;5(1):110-117

Division of Neuroimaging Research, Barrow Neurological Institute, Phoenix, AZ.

Relative cerebral blood volume (rCBV) cannot be used as a response metric in clinical trials, in part, because of variations in biomarker consistency and associated interpretation across sites, stemming from differences in image acquisition and postprocessing methods (PMs). This study leveraged a dynamic susceptibility contrast magnetic resonance imaging digital reference object to characterize rCBV consistency across 12 sites participating in the Quantitative Imaging Network (QIN), specifically focusing on differences in site-specific imaging protocols (IPs; n = 17), and PMs (n = 19) and differences due to site-specific IPs and PMs (n = 25). Thus, high agreement across sites occurs when 1 managing center processes rCBV despite slight variations in the IP. This result is most likely supported by current initiatives to standardize IPs. However, marked intersite disagreement was observed when site-specific software was applied for rCBV measurements. This study's results have important implications for comparing rCBV values across sites and trials, where variability in PMs could confound the comparison of therapeutic effectiveness and/or any attempts to establish thresholds for categorical response to therapy. To overcome these challenges and ensure the successful use of rCBV as a clinical trial biomarker, we recommend the establishment of qualifying and validating site- and trial-specific criteria for scanners and acquisition methods (eg, using a validated phantom) and the software tools used for dynamic susceptibility contrast magnetic resonance imaging analysis (eg, using a digital reference object where the ground truth is known).
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.18383/j.tom.2018.00041DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC6403027PMC
March 2019

Standardizing total kidney volume measurements for clinical trials of autosomal dominant polycystic kidney disease.

Clin Kidney J 2019 Feb 29;12(1):71-77. Epub 2018 Aug 29.

Mayo Clinic, Rochester, MN, USA.

Background: The ability of unstandardized methods to track kidney growth in clinical trials for autosomal dominant polycystic kidney disease (ADPKD) has not been critically evaluated.

Methods: The Tolvaptan Efficacy and Safety Management of ADPKD and its Outcomes (TEMPO) 3:4 study involved baseline and annual magnetic resonance follow-up imaging yearly for 3 years. Total kidney volume (TKV) measurements were performed on these four time points in addition to the baseline imaging in TEMPO 4:4, initially by Perceptive Informatics (Waltham, MA, USA) using planimetry (original dataset) and for this study by the Mayo Translational PKD Center using semiautomated and complementary automated methods (sequential dataset). In the original dataset, the same reader was assigned to all scans of individual patients in TEMPO 3:4, but readers were reassigned in TEMPO 4:4. Two placebo-treated cohorts were included. In the first ( = 158), intervals between the end of TEMPO 3:4 and the start of TEMPO 4:4 scan visits ranged from 12 to 403 days; in the second ( = 95), the same scan (measured twice) visit was used for both.

Results: Growth rates in TEMPO 3:4 were similar in the original and sequential datasets (5.5 and 5.9%/year). Growth rates during the TEMPO 3:4 to TEMPO 4:4 interval were higher in the original (13.7%/year) but were not different in the sequential dataset (4.0%/year). Comparing volumes from the same images, TKVs showed a bias of 2.2% [95% confidence interval (CI) -5.2-9.7] in the original and -0.16% (95% CI -1.91-1.58) in the sequential dataset.

Conclusions: Despite using the same software, TKV and growth rate changes were present, likely due to reader differences in the transition from TEMPO 3:4 to TEMPO 4:4 in the original but not in the sequential dataset. Robust, standardized methods are essential in ADPKD trials to minimize errors in serial TKV measurements.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1093/ckj/sfy078DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC6366146PMC
February 2019

Using germline variants to estimate glioma and subtype risks.

Neuro Oncol 2019 03;21(4):451-461

Department of Laboratory Medicine and Pathology, Mayo Clinic, Rochester, Minnesota, USA.

Background: Twenty-five single nucleotide polymorphisms (SNPs) are associated with adult diffuse glioma risk. We hypothesized that the inclusion of these 25 SNPs with age at diagnosis and sex could estimate risk of glioma as well as identify glioma subtypes.

Methods: Case-control design and multinomial logistic regression were used to develop models to estimate the risk of glioma development while accounting for histologic and molecular subtypes. Case-case design and logistic regression were used to develop models to predict isocitrate dehydrogenase (IDH) mutation status. A total of 1273 glioma cases and 443 controls from Mayo Clinic were used in the discovery set, and 852 glioma cases and 231 controls from UCSF were used in the validation set. All samples were genotyped using a custom Illumina OncoArray.

Results: Patients in the highest 5% of the risk score had more than a 14-fold increase in relative risk of developing an IDH mutant glioma. Large differences in lifetime absolute risk were observed at the extremes of the risk score percentile. For both IDH mutant 1p/19q non-codeleted glioma and IDH mutant 1p/19q codeleted glioma, the lifetime risk increased from almost null to 2.3% and almost null to 1.7%, respectively. The SNP-based model that predicted IDH mutation status had a validation concordance index of 0.85.

Conclusions: These results suggest that germline genotyping can provide new tools for the initial management of newly discovered brain lesions. Given the low lifetime risk of glioma, risk scores will not be useful for population screening; however, they may be useful in certain clinically defined high-risk groups.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1093/neuonc/noz009DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC6422428PMC
March 2019

Automated Abdominal Segmentation of CT Scans for Body Composition Analysis Using Deep Learning.

Radiology 2019 03 11;290(3):669-679. Epub 2018 Dec 11.

From the Department of Biomedical Engineering and Physiology (A.D.W.) and Department of Radiology (P.K., T.L.K., K.A.P., P.K., T.S., M.S., N.T., B.J.E.), Mayo Clinic, 200 First St SW, Rochester, MN 55905.

Purpose To develop and evaluate a fully automated algorithm for segmenting the abdomen from CT to quantify body composition. Materials and Methods For this retrospective study, a convolutional neural network based on the U-Net architecture was trained to perform abdominal segmentation on a data set of 2430 two-dimensional CT examinations and was tested on 270 CT examinations. It was further tested on a separate data set of 2369 patients with hepatocellular carcinoma (HCC). CT examinations were performed between 1997 and 2015. The mean age of patients was 67 years; for male patients, it was 67 years (range, 29-94 years), and for female patients, it was 66 years (range, 31-97 years). Differences in segmentation performance were assessed by using two-way analysis of variance with Bonferroni correction. Results Compared with reference segmentation, the model for this study achieved Dice scores (mean ± standard deviation) of 0.98 ± 0.03, 0.96 ± 0.02, and 0.97 ± 0.01 in the test set, and 0.94 ± 0.05, 0.92 ± 0.04, and 0.98 ± 0.02 in the HCC data set, for the subcutaneous, muscle, and visceral adipose tissue compartments, respectively. Performance met or exceeded that of expert manual segmentation. Conclusion Model performance met or exceeded the accuracy of expert manual segmentation of CT examinations for both the test data set and the hepatocellular carcinoma data set. The model generalized well to multiple levels of the abdomen and may be capable of fully automated quantification of body composition metrics in three-dimensional CT examinations. © RSNA, 2018 Online supplemental material is available for this article. See also the editorial by Chang in this issue.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1148/radiol.2018181432DOI Listing
March 2019

The RSNA Pediatric Bone Age Machine Learning Challenge.

Radiology 2019 02 27;290(2):498-503. Epub 2018 Nov 27.

From the Department of Radiology, Stanford University, 300 Pasteur Dr, MC 5105, Stanford, CA 94305 (S.S.H.); Department of Radiology, The Ohio State University Wexner Medical Center, Columbus, Ohio (L.M.P.); Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital/Harvard Medical School, Boston, Mass (J.K.C.); Massachusetts General Hospital & Brigham and Women's Hospital Center for Clinical Data Science, Boston, Mass (A.B.M., K.A.); Department of Radiology, University of Toronto, Toronto, Ontario, Canada (A.B.); Department of Radiology, St. Michael's Hospital, Toronto, Ontario, Canada (M.C.); Department of Diagnostic Imaging, Warren Alpert Medical School of Brown University, Rhode Island Hospital, Providence, RI (I.P.); Universidade Federal de Goiás, Goiânia, Brazil (L.A.P., R.T.S.); Universidade Federal de São Paulo, São Paulo, Brazil (N.A., F.C.K.); Visiana, Hørsholm, Denmark (H.H.T.); MD.ai, New York, NY (L.C.); Department of Radiology, Weill Cornell Medicine, New York, NY (G.S.) Department of Radiology, University of California-San Francisco, San Francisco, Calif (M.D.K.); Department of Radiology, Mayo Clinic, Rochester, Minn (B.J.E.); and Department of Radiology, Thomas Jefferson University, Philadelphia, Pa (A.E.F.).

Purpose The Radiological Society of North America (RSNA) Pediatric Bone Age Machine Learning Challenge was created to show an application of machine learning (ML) and artificial intelligence (AI) in medical imaging, promote collaboration to catalyze AI model creation, and identify innovators in medical imaging. Materials and Methods The goal of this challenge was to solicit individuals and teams to create an algorithm or model using ML techniques that would accurately determine skeletal age in a curated data set of pediatric hand radiographs. The primary evaluation measure was the mean absolute distance (MAD) in months, which was calculated as the mean of the absolute values of the difference between the model estimates and those of the reference standard, bone age. Results A data set consisting of 14 236 hand radiographs (12 611 training set, 1425 validation set, 200 test set) was made available to registered challenge participants. A total of 260 individuals or teams registered on the Challenge website. A total of 105 submissions were uploaded from 48 unique users during the training, validation, and test phases. Almost all methods used deep neural network techniques based on one or more convolutional neural networks (CNNs). The best five results based on MAD were 4.2, 4.4, 4.4, 4.5, and 4.5 months, respectively. Conclusion The RSNA Pediatric Bone Age Machine Learning Challenge showed how a coordinated approach to solving a medical imaging problem can be successfully conducted. Future ML challenges will catalyze collaboration and development of ML tools and methods that can potentially improve diagnostic accuracy and patient care. © RSNA, 2018 Online supplemental material is available for this article. See also the editorial by Siegel in this issue.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1148/radiol.2018180736DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC6358027PMC
February 2019

What Does Deep Learning See? Insights From a Classifier Trained to Predict Contrast Enhancement Phase From CT Images.

AJR Am J Roentgenol 2018 12 7;211(6):1184-1193. Epub 2018 Nov 7.

1 Department of Radiology, Radiology Informatics Laboratory, Mayo Clinic, 3507 17th Ave NW, Rochester, MN 55901.

Objective: Deep learning has shown great promise for improving medical image classification tasks. However, knowing what aspects of an image the deep learning system uses or, in a manner of speaking, sees to make its prediction is difficult.

Materials And Methods: Within a radiologic imaging context, we investigated the utility of methods designed to identify features within images on which deep learning activates. In this study, we developed a classifier to identify contrast enhancement phase from whole-slice CT data. We then used this classifier as an easily interpretable system to explore the utility of class activation map (CAMs), gradient-weighted class activation maps (Grad-CAMs), saliency maps, guided backpropagation maps, and the saliency activation map, a novel map reported here, to identify image features the model used when performing prediction.

Results: All techniques identified voxels within imaging that the classifier used. SAMs had greater specificity than did guided backpropagation maps, CAMs, and Grad-CAMs at identifying voxels within imaging that the model used to perform prediction. At shallow network layers, SAMs had greater specificity than Grad-CAMs at identifying input voxels that the layers within the model used to perform prediction.

Conclusion: As a whole, voxel-level visualizations and visualizations of the imaging features that activate shallow network layers are powerful techniques to identify features that deep learning models use when performing prediction.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.2214/AJR.18.20331DOI Listing
December 2018

Deep Learning in Radiology: Does One Size Fit All?

J Am Coll Radiol 2018 03 31;15(3 Pt B):521-526. Epub 2018 Jan 31.

Radiology Informatics Laboratory, Department of Radiology, Mayo Clinic, Rochester, Minnesota.

Deep learning (DL) is a popular method that is used to perform many important tasks in radiology and medical imaging. Some forms of DL are able to accurately segment organs (essentially, trace the boundaries, enabling volume measurements or calculation of other properties). Other DL networks are able to predict important properties from regions of an image-for instance, whether something is malignant, molecular markers for tissue in a region, even prognostic markers. DL is easier to train than traditional machine learning methods, but requires more data and much more care in analyzing results. It will automatically find the features of importance, but understanding what those features are can be a challenge. This article describes the basic concepts of DL systems and some of the traps that exist in building DL systems and how to identify those traps.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1016/j.jacr.2017.12.027DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC5877825PMC
March 2018

Brain Gliomas: Multicenter Standardized Assessment of Dynamic Contrast-enhanced and Dynamic Susceptibility Contrast MR Images.

Radiology 2018 06 22;287(3):933-943. Epub 2018 Jan 22.

From the Neuroradiology Unit and CERMAC, Vita-Salute San Raffaele University and IRCCS San Raffaele Scientific Institute, Via Olgettina 60, 20132, Milan, Italy (N.A., A. Castellano, M. Cadioli, G.M.C.); Neuroradiology Unit, Fondazione IRCCS Istituto Neurologico Carlo Besta, Milan, Italy (V.C., A.B., D.A.); Department of Radiology, Humanitas Clinical and Research Hospital, Rozzano, Milan, Italy (M.G.); Department of Neuroradiology, Fondazione IRCCS Cà Granda-Ospedale Maggiore Policlinico, Milan, Italy (A. Costa); IRCCS Neuromed, Pozzilli (Isernia), Italy (G.G.); Neuroradiology Unit, C. Mondino National Neurologic Institute, Pavia, Italy (P.V.); Pathology Unit, IRCCS San Raffaele Scientific Institute, Milan, Italy (M.R.T.); Methodology for Clinical Research Laboratory, Department of Oncology, IRCCS Mario Negri, Milan, Italy (V.T.); Department of Radiology, Mayo Clinic, Rochester, Minn (B.J.E.); and Department of Neuroscience and Imaging and ITAB-Institute of Advanced Biomedical Technologies, University G. d'Annunzio, Chieti, Italy (M. Caulo).

Purpose To evaluate the feasibility of a standardized protocol for acquisition and analysis of dynamic contrast material-enhanced (DCE) and dynamic susceptibility contrast (DSC) magnetic resonance (MR) imaging in a multicenter clinical setting and to verify its accuracy in predicting glioma grade according to the new World Health Organization 2016 classification. Materials and Methods The local research ethics committees of all centers approved the study, and informed consent was obtained from patients. One hundred patients with glioma were prospectively examined at 3.0 T in seven centers that performed the same preoperative MR imaging protocol, including DCE and DSC sequences. Two independent readers identified the perfusion hotspots on maps of volume transfer constant (K), plasma (v) and extravascular-extracellular space (v) volumes, initial area under the concentration curve, and relative cerebral blood volume (rCBV). Differences in parameters between grades and molecular subtypes were assessed by using Kruskal-Wallis and Mann-Whitney U tests. Diagnostic accuracy was evaluated by using receiver operating characteristic curve analysis. Results The whole protocol was tolerated in all patients. Perfusion maps were successfully obtained in 94 patients. An excellent interreader reproducibility of DSC- and DCE-derived measures was found. Among DCE-derived parameters, v and v had the highest accuracy (are under the receiver operating characteristic curve [A] = 0.847 and 0.853) for glioma grading. DSC-derived rCBV had the highest accuracy (A = 0.894), but the difference was not statistically significant (P > .05). Among lower-grade gliomas, a moderate increase in both v and rCBV was evident in isocitrate dehydrogenase wild-type tumors, although this was not significant (P > .05). Conclusion A standardized multicenter acquisition and analysis protocol of DCE and DSC MR imaging is feasible and highly reproducible. Both techniques showed a comparable, high diagnostic accuracy for grading gliomas. RSNA, 2018 Online supplemental material is available for this article.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1148/radiol.2017170362DOI Listing
June 2018

Randomised controlled trial to determine the efficacy and safety of prescribed water intake to prevent kidney failure due to autosomal dominant polycystic kidney disease (PREVENT-ADPKD).

BMJ Open 2018 01 21;8(1):e018794. Epub 2018 Jan 21.

Centre for Transplant and Renal Research, The Westmead Institute for Medical Research, The University of Sydney, Sydney, Australia.

Introduction: Maintaining fluid intake sufficient to reduce arginine vasopressin (AVP) secretion has been hypothesised to slow kidney cyst growth in autosomal dominant polycystic kidney disease (ADPKD). However, evidence to support this as a clinical practice recommendation is of poor quality. The aim of the present study is to determine the long-term efficacy and safety of prescribed water intake to prevent the progression of height-adjusted total kidney volume (ht-TKV) in patients with chronic kidney disease (stages 1-3) due to ADPKD.

Methods And Analysis: A multicentre, prospective, parallel-group, open-label, randomised controlled trial will be conducted. Patients with ADPKD (n=180; age ≤65 years, estimated glomerular filtration rate (eGFR) ≥30 mL/min/1.73 m) will be randomised (1:1) to either the control (standard treatment+usual fluid intake) or intervention (standard treatment+prescribed fluid intake) group. Participants in the intervention arm will be prescribed an individualised daily fluid intake to reduce urine osmolality to ≤270 mOsmol/kg, and supported with structured clinic and telephonic dietetic review, self-monitoring of urine-specific gravity, short message service text reminders and internet-based tools. All participants will have 6-monthly follow-up visits, and ht-TKV will be measured by MRI at 0, 18 and 36 months. The primary end point is the annual rate of change in ht-TKV as determined by serial renal MRI in control vs intervention groups, from baseline to 3 years. The secondary end points are differences between the two groups in systemic AVP activity, renal disease (eGFR, blood pressure, renal pain), patient adherence, acceptability and safety.

Ethics And Dissemination: The trial was approved by the Human Research Ethics Committee, Western Sydney Local Health District. The results will inform clinicians, patients and policy-makers regarding the long-term safety, efficacy and feasibility of prescribed fluid intake as an approach to reduce kidney cyst growth in patients with ADPKD.

Trial Registration Number: ANZCTR12614001216606.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1136/bmjopen-2017-018794DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC5780847PMC
January 2018

Phase 1/2 trial of temsirolimus and sorafenib in the treatment of patients with recurrent glioblastoma: North Central Cancer Treatment Group Study/Alliance N0572.

Cancer 2018 04 3;124(7):1455-1463. Epub 2018 Jan 3.

Department of Radiation Oncology, Mayo Clinic, Rochester, Minnesota.

Background: Mitogen-activated protein kinase (MAPK) activation and mammalian target of rapamycin (mTOR)-dependent signaling are hallmarks of glioblastoma. In the current study, the authors conducted a phase 1/2 study of sorafenib (an inhibitor of Raf kinase and vascular endothelial growth factor receptor 2 [VEGFR-2]) and the mTOR inhibitor temsirolimus in patients with recurrent glioblastoma.

Methods: Patients with recurrent glioblastoma who developed disease progression after surgery or radiotherapy plus temozolomide and with ≤2 prior chemotherapy regimens were eligible. The phase 1 endpoint was the maximum tolerated dose (MTD), using a cohorts-of-3 design. The 2-stage phase 2 study included separate arms for VEGF inhibitor (VEGFi)-naive patients and patients who progressed after prior VEGFi.

Results: The MTD was sorafenib at a dose of 200 mg twice daily and temsirolimus at a dose of 20 mg weekly. In the first 41 evaluable patients who were treated at the phase 2 dose, there were 7 who were free of disease progression at 6 months (progression-free survival at 6 months [PFS6]) in the VEGFi-naive group (17.1%); this finding met the prestudy threshold of success. In the prior VEGFi group, only 4 of the first 41 evaluable patients treated at the phase 2 dose achieved PFS6 (9.8%), and this did not meet the prestudy threshold for success. The median PFS for the 2 groups was 2.6 months and 1.9 months, respectively. The median overall survival for the 2 groups was 6.3 months and 3.9 months, respectively. At least 1 adverse event of grade ≥3 was observed in 75.5% of the VEGFi-naive patients and in 73.9% of the prior VEGFi patients.

Conclusions: The limited activity of sorafenib and temsirolimus at the dose and schedule used in the current study was observed with considerable toxicity of grade ≥3. Significant dose reductions that were required in this treatment combination compared with tolerated single-agent doses may have contributed to the lack of efficacy. Cancer 2018;124:1455-63. © 2018 American Cancer Society.
View Article and Find Full Text PDF

Download full-text PDF

Source
http://dx.doi.org/10.1002/cncr.31219DOI Listing
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC5867230PMC
April 2018