Dataset Information

Automated Generation of Radiologic Descriptions on Brain Volume Changes From T1-Weighted MR Images: Initial Assessment of Feasibility.

ABSTRACT: Purpose: To examine the feasibility and potential difficulties of automatically generating radiologic reports (RRs) to articulate the clinically important features of brain magnetic resonance (MR) images. Materials and Methods: We focused on examining brain atrophy by using magnetization-prepared rapid gradient-echo (MPRAGE) images. The technology was based on multi-atlas whole-brain segmentation that identified 283 structures, from which larger superstructures were created to represent the anatomic units most frequently used in RRs. Through two layers of data-reduction filters, based on anatomic and clinical knowledge, raw images (~10 MB) were converted to a few kilobytes of human-readable sentences. The tool was applied to images from 92 patients with memory problems, and the results were compared to RRs independently produced by three experienced radiologists. The mechanisms of disagreement were investigated to understand where machine-human interface succeeded or failed. Results: The automatically generated sentences had low sensitivity (mean: 24.5%) and precision (mean: 24.9%) values; these were significantly lower than the inter-rater sensitivity (mean: 32.7%) and precision (mean: 32.2%) of the radiologists. The causes of disagreement were divided into six error categories: mismatch of anatomic definitions (7.2 ± 9.3%), data-reduction errors (11.4 ± 3.9%), translator errors (3.1 ± 3.1%), difference in the spatial extent of used anatomic terms (8.3 ± 6.7%), segmentation quality (9.8 ± 2.0%), and threshold for sentence-triggering (60.2 ± 16.3%). Conclusion: These error mechanisms raise interesting questions about the potential of automated report generation and the quality of image reading by humans. The most significant discrepancy between the human and automatically generated RRs was caused by the sentence-triggering threshold (the degree of abnormality), which was fixed to z-score >2.0 for the automated generation, while the thresholds by radiologists varied among different anatomical structures.

SUBMITTER: Akazawa K

PROVIDER: S-EPMC6354548 | biostudies-literature | 2019

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Automated Generation of Radiologic Descriptions on Brain Volume Changes From T1-Weighted MR Images: Initial Assessment of Feasibility.

Akazawa Kentaro K Sakamoto Ryo R Nakajima Satoshi S Wu Dan D Li Yue Y Oishi Kenichi K Faria Andreia V AV Yamada Kei K Togashi Kaori K Lyketsos Constantine G CG Miller Michael I MI Mori Susumu S

Frontiers in neurology 20190124

<b>Purpose:</b> To examine the feasibility and potential difficulties of automatically generating radiologic reports (RRs) to articulate the clinically important features of brain magnetic resonance (MR) images. <b>Materials and Methods:</b> We focused on examining brain atrophy by using magnetization-prepared rapid gradient-echo (MPRAGE) images. The technology was based on multi-atlas whole-brain segmentation that identified 283 structures, from which larger superstructures were created to repr ...[more]

PMID: 30733701

Similar Datasets

Project description:To retrospectively determine the prevalence and positive predictive value (PPV) of the hemorrhage exclusion sign on T1-weighted magnetic resonance (MR) images in conjunction with findings on T2-weighted images in the detection of prostate cancer, with use of whole-mount step-section pathologic specimens from prostatectomy as the reference standard.The institutional review board approved this retrospective study, which was compliant with HIPAA, and the requirement to obtain informed consent was waived. Two hundred ninety-two patients with biopsy-proved prostate cancer underwent endorectal MR imaging followed by prostatectomy. The hemorrhage exclusion sign was defined as the presence of a well-defined area of low signal intensity surrounded by areas of high signal intensity on T1-weighted images. Two readers independently assessed the presence and extent of postbiopsy changes and the hemorrhage exclusion sign. The presence of a corresponding area of homogeneous low signal intensity on T2-weighted images was also recorded. The prevalence and PPV of the hemorrhage exclusion sign were calculated.Readers 1 and 2 found postbiopsy changes in the peripheral zone in 184 (63%) and 189 (64.7%) of the 292 patients, respectively. In these patients, the hemorrhage exclusion sign was observed in 39 of 184 patients (21.2%) by reader 1 and 36 of 189 patients (19.0%) by reader 2. A corresponding area of homogeneous low signal intensity was seen on T2-weighted images in the same location as the hemorrhage exclusion sign in 23 of 39 patients (59%) by reader 1 and 19 of 36 patients (53%) by reader 2. The PPV of the hemorrhage exclusion sign alone was 56% (22 of 39 patients) for reader 1 and 50% (18 of 36 patients) for reader 2 but increased to 96% (22 of 23 patients) and 95% (18 of 19 patients) when the sign was identified in an area of homogeneous low signal intensity on T2-weighted images.Postbiopsy change is a known pitfall in the interpretation of T2-weighted images. The authors have shown that a potential benefit of postbiopsy change is the presence of excluded hemorrhage, which, in conjunction with a corresponding area of homogeneous low signal intensity at T2-weighted imaging, is highly accurate for cancer identification.

Project description:PurposeTo develop and evaluate a complete deep learning pipeline that allows fully automated end-diastolic left ventricle (LV) cardiac MRI segmentation, including trabeculations and automatic quality control of the predicted segmentation.Materials and methodsThis multicenter retrospective study includes training, validation, and testing datasets of 272, 27, and 150 cardiac MR images, respectively, collected between 2012 and 2018. The reference standard was the manual segmentation of four LV anatomic structures performed on end-diastolic short-axis cine cardiac MRI: LV trabeculations, LV myocardium, LV papillary muscles, and the LV blood cavity. The automatic pipeline was composed of five steps with a DenseNet architecture. Intraobserver agreement, interobserver agreement, and interaction time were recorded. The analysis includes the correlation between the manual and automated segmentation, a reproducibility comparison, and Bland-Altman plots.ResultsThe automated method achieved mean Dice coefficients of 0.96 ± 0.01 (standard deviation) for LV blood cavity, 0.89 ± 0.03 for LV myocardium, and 0.62 ± 0.08 for LV trabeculation (mean absolute error, 3.63 g ± 3.4). Automatic quantification of LV end-diastolic volume, LV myocardium mass, LV trabeculation, and trabeculation mass-to-total myocardial mass (TMM) ratio showed a significant correlation with the manual measures (r = 0.99, 0.99, 0.90, and 0.83, respectively; all P < .01). On a subset of 48 patients, the mean Dice value for LV trabeculation was 0.63 ± 0.10 or higher compared with the human interobserver (0.44 ± 0.09; P < .01) and intraobserver measures (0.58 ± 0.09; P < .01). Automatic quantification of the trabeculation mass-to-TMM ratio had a higher correlation (0.92) compared with the intra- and interobserver measures (0.74 and 0.39, respectively; both P < .01).ConclusionAutomated deep learning framework can achieve reproducible and quality-controlled segmentation of cardiac trabeculations, outperforming inter- and intraobserver analyses.Supplemental material is available for this article.© RSNA, 2020.

Project description:PurposeTo retrospectively determine the accuracy of diffusion-weighted (DW) magnetic resonance (MR) imaging for identifying cancer in the prostate peripheral zone (PZ) and to assess the accuracy of tumor volume measurements made with T2-weighted imaging and combined T2-weighted and DW MR imaging by using surgical pathologic examination as the reference standard.Materials and methodsThe institutional review board issued a waiver of informed consent for this HIPAA-compliant study. Forty-two patients underwent endorectal MR at 1.5 T before undergoing radical prostatectomy for prostate cancer and had at least one PZ tumor larger than 0.1 cm(3) at surgical pathologic examination. On T2-weighted images, an experienced radiologist outlined suspected PZ tumors. Two apparent diffusion coefficient (ADC) cutoff values were identified by using the Youden index and published literature. Image cluster analysis was performed on voxels within the suspected tumor regions. Associations between volume measurements from imaging and from pathologic examination were assessed by using concordance correlation coefficients (CCCs). The sensitivity and specificity of ADCs for identifying malignant PZ voxels were calculated.ResultsIn identifying malignant voxels, respective ADC cutoff values of 0.0014 and 0.0016 mm(2)/sec yielded sensitivity of 82% and 95% and specificity of 85% and 65%, respectively. Sixty PZ cancer lesions larger than 0.1 cm(3) were found at pathologic examination; 43 were detected by the radiologist. CCCs between imaging and pathologic tumor volume measurements were 0.36 for T2-weighted imaging, and 0.46 and 0.60 for combined T2-weighted and DW MR imaging with ADC cutoffs of 0.0014 and 0.0016 mm(2)/sec, respectively; the CCC of combined T2-weighted and DW MR imaging (ADC cutoff, 0.0016 mm(2)/sec) was significantly higher (P = .006) than that of T2-weighted imaging alone.ConclusionAdding DW MR to T2-weighted imaging can significantly improve the accuracy of prostate PZ tumor volume measurement.Supplemental materialhttp://radiology.rsnajnls.org/cgi/content/full/252/2/449/DC1.

Dataset Information

Automated Generation of Radiologic Descriptions on Brain Volume Changes From T1-Weighted MR Images: Initial Assessment of Feasibility.

Publications

Automated Generation of Radiologic Descriptions on Brain Volume Changes From T1-Weighted MR Images: Initial Assessment of Feasibility.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets