Computerized Decision Support for Bladder Cancer Treatment Response Assessment in CT Urography: Effect on Diagnostic Accuracy in Multi-Institution Multi-Specialty Study.

This observer study investigates the effect of computerized artificial intelligence (AI)-based decision support system (CDSS-T) on physicians' diagnostic accuracy in assessing bladder cancer treatment response. The performance of 17 observers was evaluated when assessing bladder cancer treatment response without and with CDSS-T using pre- and post-chemotherapy CTU scans in 123 patients having 157 pre- and post-treatment cancer pairs. The impact of cancer case difficulty, observers' clinical experience, institution affiliation, specialty, and the assessment times on the observers' diagnostic performance with and without using CDSS-T were analyzed. It was found that the average performance of the 17 observers was significantly improved (p = 0.002) when aided by the CDSS-T. The cancer case difficulty, institution affiliation, specialty, and the assessment times influenced the observers' performance without CDSS-T. The AI-based decision support system has the potential to improve the diagnostic accuracy in assessing bladder cancer treatment response and result in more consistent performance among all physicians.

Intraobserver Variability in Bladder Cancer Treatment Response Assessment With and Without Computerized Decision Support.

We evaluated the intraobserver variability of physicians aided by a computerized decision-support system for treatment response assessment (CDSS-T) to identify patients who show complete response to neoadjuvant chemotherapy for bladder cancer, and the effects of the intraobserver variability on physicians' assessment accuracy. A CDSS-T tool was developed that uses a combination of deep learning neural network and radiomic features from computed tomography (CT) scans to detect bladder cancers that have fully responded to neoadjuvant treatment. Pre- and postchemotherapy CT scans of 157 bladder cancers from 123 patients were collected. In a multireader, multicase observer study, physician-observers estimated the likelihood of pathologic T0 disease by viewing paired pre/posttreatment CT scans placed side by side on an in-house-developed graphical user interface. Five abdominal radiologists, 4 diagnostic radiology residents, 2 oncologists, and 1 urologist participated as observers. They first provided an estimate without CDSS-T and then with CDSS-T. A subset of cases was evaluated twice to study the intraobserver variability and its effects on observer consistency. The mean areas under the curves for assessment of pathologic T0 disease were 0.85 for CDSS-T alone, 0.76 for physicians without CDSS-T and improved to 0.80 for physicians with CDSS-T (P = .001) in the original evaluation, and 0.78 for physicians without CDSS-T and improved to 0.81 for physicians with CDSS-T (P = .010) in the repeated evaluation. The intraobserver variability was significantly reduced with CDSS-T (P < .0001). The CDSS-T can significantly reduce physicians' variability and improve their accuracy for identifying complete response of muscle-invasive bladder cancer to neoadjuvant chemotherapy.

Deep Learning Approach for Assessment of Bladder Cancer Treatment Response.

We compared the performance of different Deep learning-convolutional neural network (DL-CNN) models for bladder cancer treatment response assessment based on transfer learning by freezing different DL-CNN layers and varying the DL-CNN structure. Pre- and posttreatment computed tomography scans of 123 patients (cancers, 129; pre- and posttreatment cancer pairs, 158) undergoing chemotherapy were collected. After chemotherapy 33% of patients had T0 stage cancer (complete response). Regions of interest in pre- and posttreatment scans were extracted from the segmented lesions and combined into hybrid pre -post image pairs (h-ROIs). Training (pairs, 94; h-ROIs, 6209), validation (10 pairs) and test sets (54 pairs) were obtained. The DL-CNN consisted of 2 convolution (C1-C2), 2 locally connected (L3-L4), and 1 fully connected layers. The DL-CNN was trained with h-ROIs to classify cancers as fully responding (stage T0) or not fully responding to chemotherapy. Two radiologists provided lesion likelihood of being stage T0 posttreatment. The test area under the ROC curve (AUC) was 0.73 for T0 prediction by the base DL-CNN structure with randomly initialized weights. The base DL-CNN structure with pretrained weights and transfer learning (no frozen layers) achieved test AUC of 0.79. The test AUCs for 3 modified DL-CNN structures (different C1-C2 max pooling filter sizes, strides, and padding, with transfer learning) were 0.72, 0.86, and 0.69. For the base DL-CNN with (C1) frozen, (C1-C2) frozen, and (C1-C2-L3) frozen, the test AUCs were 0.81, 0.78, and 0.71, respectively. The radiologists' AUCs were 0.76 and 0.77. DL-CNN performed better with pretrained than randomly initialized weights.

Needs Assessment Using a Structured Prioritization Schema: An Open Letter to PACS Vendors.

PURPOSE: The aim of this work was to prioritize in a quaternary academic environment necessary elements of a replacement PACS. METHODS: This quality improvement work was conducted at one academic medical center and was "not regulated" by the institutional review board. Three workgroups (10-15 members each) with unique resident, fellow, and attending radiologists; IT specialists; and departmental leaders convened in 2018 to prioritize elements for a PACS replacement project, including integrated IT tools. Each workgroup met two or three times and represented one of three missions (clinical, research, and education). Six elements assigned the highest priority were distilled from each workgroup. The resulting 18 elements were condensed into survey format and distributed to all department residents, fellows, and faculty members for 5-point Likert-type prioritization stratified by mission. Data were collected over 2 weeks. RESULTS: The survey response rate was 37% (71 of 192; 17 of 44 residents, 3 of 27 fellows, and 51 of 121 faculty members). Self-reported work effort was 63 ± 26% clinical, 14 ± 11% education, 15 ± 21% research, and 8 ± 14% administration. Aggregate priority ratings across all domains were highest for "stable system with predictable behavior" (mean, 4.51), "minimizes repetitive non-value-added work" (mean, 4.40), "interoperability" (mean, 4.12), and "near-instantaneous load times" (mean, 4.07). Clinical-specific ratings for these elements were even higher (means, 4.85-4.90). The lowest aggregate scores were mobile device compatibility (mean, 3.03), connectivity to nonaffiliated sites (mean, 3.01), and integrated instant messaging (mean, 2.87). CONCLUSIONS: The department prioritized a stable and interoperable system that minimized non-value-added work. In other words, participants wanted a functioning PACS. PACS vendors should prioritize a reliable experience over niche add-ons.

Bladder Cancer Treatment Response Assessment in CT using Radiomics with Deep-Learning.

Cross-sectional X-ray imaging has become the standard for staging most solid organ malignancies. However, for some malignancies such as urinary bladder cancer, the ability to accurately assess local extent of the disease and understand response to systemic chemotherapy is limited with current imaging approaches. In this study, we explored the feasibility that radiomics-based predictive models using pre- and post-treatment computed tomography (CT) images might be able to distinguish between bladder cancers with and without complete chemotherapy responses. We assessed three unique radiomics-based predictive models, each of which employed different fundamental design principles ranging from a pattern recognition method via deep-learning convolution neural network (DL-CNN), to a more deterministic radiomics feature-based approach and then a bridging method between the two, utilizing a system which extracts radiomics features from the image patterns. Our study indicates that the computerized assessment using radiomics information from the pre- and post-treatment CT of bladder cancer patients has the potential to assist in assessment of treatment response.

Bladder Cancer Segmentation in CT for Treatment Response Assessment: Application of Deep-Learning Convolution Neural Network-A Pilot Study.

Assessing the response of bladder cancer to neoadjuvant chemotherapy is crucial for reducing morbidity and increasing quality of life of patients. Changes in tumor volume during treatment is generally used to predict treatment outcome. We are developing a method for bladder cancer segmentation in CT using a pilot data set of 62 cases. 65 000 regions of interests were extracted from pre-treatment CT images to train a deep-learning convolution neural network (DL-CNN) for tumor boundary detection using leave-one-case-out cross-validation. The results were compared to our previous AI-CALS method. For all lesions in the data set, the longest diameter and its perpendicular were measured by two radiologists, and 3D manual segmentation was obtained from one radiologist. The World Health Organization (WHO) criteria and the Response Evaluation Criteria In Solid Tumors (RECIST) were calculated, and the prediction accuracy of complete response to chemotherapy was estimated by the area under the receiver operating characteristic curve (AUC). The AUCs were 0.73 ± 0.06, 0.70 ± 0.07, and 0.70 ± 0.06, respectively, for the volume change calculated using DL-CNN segmentation, the AI-CALS and the manual contours. The differences did not achieve statistical significance. The AUCs using the WHO criteria were 0.63 ± 0.07 and 0.61 ± 0.06, while the AUCs using RECIST were 0.65 ± 007 and 0.63 ± 0.06 for the two radiologists, respectively. Our results indicate that DL-CNN can produce accurate bladder cancer segmentation for calculation of tumor size change in response to treatment. The volume change performed better than the estimations from the WHO criteria and RECIST for the prediction of complete response.

Indirect Cost and Harm Attributable to Oral 13-Hour Inpatient Corticosteroid Prophylaxis before Contrast-enhanced CT.

PURPOSE: To estimate the effect of an oral 13-hour inpatient corticosteroid premedication regimen on length of stay, hospital cost, and hospital-acquired infections (HAIs) by using a combination of real and hypothetical study populations. MATERIALS AND METHODS: Institutional review board approval was obtained and informed consent waived for this HIPAA-compliant retrospective study. Inpatients who received an oral 13-hour corticosteroid premedication regimen before contrast material-enhanced CT (n = 1424) from 2008 to 2013 were matched by age, sex, and year when CT was performed to a control cohort (n = 1425) of patients who underwent contrast-enhanced CT without premedication and who had similar rates of 13 comorbid diseases. Length of stay in the hospital and time from admission to CT were compared by using the Mann-Whitney U test. Rates of prospectively reported HAIs were compared by using χ(2) tests. The indirect cost and risk of HAI with premedication were estimated by using published data. RESULTS: Premedicated inpatients had a significantly longer median length of stay (+25 hours; 158 vs 133 hours, P < .001), a significantly longer median time to CT (+25 hours, 42 vs 17 hours, respectively; P < .001), and a significantly greater risk of HAI (5.1% [72 of 1424] vs 3.1% [44 of 1424], respectively; P = .008) compared with nonpremedicated control subjects. On the basis of these data and existing references, the prolonged length of stay was estimated to result in 0.04 HAI-related deaths and a cost of $159 131 (in U.S. dollars) for each prevented reaction of any severity and 32 HAI-related deaths and a cost of $131 211 400 for each prevented reaction-related death. CONCLUSION: Oral 13-hour inpatient corticosteroid prophylaxis is associated with substantial cost relative to its modest benefit, and may cause more indirect harm than the direct harm that it prevents.

Treatment Response Assessment for Bladder Cancer on CT Based on Computerized Volume Analysis, World Health Organization Criteria, and RECIST.

OBJECTIVE: The purpose of this study was to evaluate the accuracy of our autoinitialized cascaded level set 3D segmentation system as compared with the World Health Organization (WHO) criteria and the Response Evaluation Criteria In Solid Tumors (RECIST) for estimation of treatment response of bladder cancer in CT urography. MATERIALS AND METHODS: CT urograms before and after neoadjuvant chemo-therapy treatment were collected from 18 patients with muscle-invasive localized or locally advanced bladder cancers. The disease stage as determined on pathologic samples at cystectomy after chemotherapy was considered as reference standard of treatment response. Two radiologists measured the longest diameter and its perpendicular on the pre- and posttreatment scans. Full 3D contours for all tumors were manually outlined by one radiologist. The autoinitialized cascaded level set method was used to automatically extract 3D tumor boundary. The prediction accuracy of pT0 disease (complete response) at cystectomy was estimated by the manual, autoinitialized cascaded level set, WHO, and RECIST methods on the basis of the AUC. RESULTS: The AUC for prediction of pT0 disease at cystectomy was 0.78 ± 0.11 for autoinitialized cascaded level set compared with 0.82 ± 0.10 for manual segmentation. The difference did not reach statistical significance (p = 0.67). The AUCs using RECIST criteria were 0.62 ± 0.16 and 0.71 ± 0.12 for the two radiologists, both lower than those of the two 3D methods. The AUCs using WHO criteria were 0.56 ± 0.15 and 0.60 ± 0.13 and thus were lower than all other methods. CONCLUSION: The pre- and posttreatment 3D volume change estimates obtained by the radiologist's manual outlines and the autoinitialized cascaded level set segmentation were more accurate for irregularly shaped tumors than were those based on RECIST and WHO criteria.

Assessment of 1 mSv urinary tract stone CT with model-based iterative reconstruction.

OBJECTIVE: The purpose of this study was to evaluate stone detection, assessment of secondary signs of stone disease, and diagnostic confidence utilizing submillisievert CT with model-based iterative reconstruction (MBIR) in a North American population with diverse body habitus. MATERIALS AND METHODS: Fifty-two adults underwent stone CT using a split-dose protocol; weight-based projected volume CT dose index (CTDIvol) and dose-length product (DLP) were divided into two separate acquisitions at 80% and 20% dose levels. Images were reconstructed with MBIR. Five blinded readers counted stones in three size categories and rated "overall diagnostic confidence" and "detectability of secondary signs of stone disease" on a 0-4 scale at both dose levels. Effective dose (ED) in mSv was calculated as DLP multiplied by conversion coefficient, k, equal to 0.017. RESULTS: Mean ED (80%, 3.90±1.44 mSv; vs 20%, 0.97±0.34 mSv [p<0.001]) and number of stones detected (80%, 193.6±25.0; vs 20%, 154.4±15.4 [p=0.03]) were higher in scans at 80% dose level. Intrareader correlation between scans at 80% and 20% dose levels was excellent (0.83-0.97). With 80% scans as reference standard, mean sensitivity and specificity at 20% varied with stone size (<3 mm, 74% and 77%; ≥3 mm, 92% and 82%). The 20% scans scored lower than 80% scans in diagnostic confidence (2.46±0.50; vs 3.21±0.36 [p<0.005]) and detectability of secondary signs (2.41±0.39; vs 3.19±0.29 [p<0.005]). CONCLUSION: Aggressively dose-reduced (~1 mSv) MBIR scans detected most urinary tract stones of 3 mm or larger but underperformed the low-dose reference standard (3-4 mSv) scans in small (<3 mm) stone detection and diagnostic confidence.

Residents' ability to interpret radiology images: development and improvement of an assessment tool.

RATIONALE AND OBJECTIVES: Despite increasing radiology coverage, nonradiology residents continue to preliminarily interpret basic radiologic studies independently, yet their ability to do so accurately is not routinely assessed. MATERIALS AND METHODS: An online test of basic radiologic image interpretation was developed through an iterative process. Educational objectives were established, then questions and images were gathered to create an assessment. The test was administered online to first-year interns (postgraduate year [PGY] 1) from 14 different specialties, as well as a sample of third- and fourth-year radiology residents (PGY3/R2 and PGY4/R3). RESULTS: Over a 2-year period, 368 residents were assessed, including PGY1 (n = 349), PGY3/R2 (n = 14), and PGY4/R3 (n = 5) residents. Overall, the test discriminated effectively between interns (average score = 66%) and advanced residents (R2 = 86%, R3 = 89%; P < .05). Item analysis indicated discrimination indices ranging from -0.72 to 48.3 (mean = 3.12, median 0.58) for individual questions, including four questions with negative discrimination indices. After removal of the negatively indexed questions, the overall predictive value of the instrument persisted and discrimination indices increased for all but one of the remaining questions (range 0.027-70.8, mean 5.76, median 0.94). CONCLUSIONS: Validation of an initial iteration of an assessment of basic image-interpretation skills led to revisions that improved the test. The results offer a specific test of radiologic reading skills with validation evidence for residents. More generally, results demonstrate a principled approach to test development.

Contrast medium-induced nephrotoxicity risk assessment in adult inpatients: a comparison of serum creatinine level- and estimated glomerular filtration rate-based screening methods.

PURPOSE: To compare serum creatinine (SCr) level- and estimated glomerular filtration rate (eGFR)-based screening methods for identifying adult inpatients at risk for contrast medium-induced nephrotoxicity (CIN). MATERIALS AND METHODS: Institutional review board approval was obtained; informed consent was waived for this HIPAA-compliant retrospective study. Computed tomographic examinations performed during 10 years in adult inpatients with stable renal function were identified (n = 28 390). The proportion of inpatients meeting various eGFR (≥60, <60, <45, <30, 30-44, 45-59 mL/min/1.73 m(2)) and SCr (<1.5, ≥1.5, ≥1.6, ≥1.7, ≥1.8, ≥1.9, ≥2.0 mg/dL) thresholds were contrasted with each other and with published guidelines (≥2.0 mg/dL [SCr] and <45 mL/min/1.73 m(2) [eGFR]) using McNemar and binomial tests. RESULTS: Most inpatients were considered low risk for CIN with commonly used thresholds: 92.6% (26 285 of 28 390) had SCr <1.5 mg/dL; 91.3% (25 922 of 28 390) had eGFR of ≥45 mL/min/1.73 m(2). Using SCr threshold of ≥1.5 mg/dL, identified inpatients had the following eGFRs: 19.6% (413 of 2105), 45-59 mL/min/1.73 m(2); 51.1% (1075 of 2105), 30-44 mL/min/1.73 m(2); 28.6% (603 of 2105), <30 mL/min/1.73 m(2); and 0.7% (14 of 2105), ≥60 mL/min/1.73 m(2) . Using SCr threshold of ≥2.0 mg/dL, identified inpatients had the following eGFRs: 100% (658 of 658), <45 mL/min/1.73 m(2); 74.6% (491 of 658), <30 mL/min/1.73 m(2). Threshold of SCr ≥2.0 mg/dL could not be used to identify eGFR <30 mL/min/1.73 m(2) in 0.4% (112 of 28 390) and <45 mL/min/1.73 m(2) in 6.4% (1810 of 28 390) of all inpatients. Using eGFR <45 mL/min/1.73 m(2) instead of SCr of ≥1.5 mg/dL would result in a significant but small increase in identified inpatients (8.7% [2468 of 28 390; 95% confidence interval: 8.4%, 9.0%] vs 7.4% [2105 of 28 390; 95% confidence interval: 7.1%, 7.7%]; P < .0001). CONCLUSION: Screening using eGFR <45 mL/min/1.73 m(2) instead of common SCr thresholds would significantly increase the number of inpatients identified to be at risk for CIN but would reduce misidentification of a large number of inpatients at low risk according to eGFR criteria.