Unsupervised segmentation and quantification of COVID-19 lesions on computed Tomography scans using CycleGAN.

BACKGROUND: Lesion segmentation is a critical step in medical image analysis, and methods to identify pathology without time-intensive manual labeling of data are of utmost importance during a pandemic and in resource-constrained healthcare settings. Here, we describe a method for fully automated segmentation and quantification of pathological COVID-19 lung tissue on chest Computed Tomography (CT) scans without the need for manually segmented training data. METHODS: We trained a cycle-consistent generative adversarial network (CycleGAN) to convert images of COVID-19 scans into their generated healthy equivalents. Subtraction of the generated healthy images from their corresponding original CT scans yielded maps of pathological tissue, without background lung parenchyma, fissures, airways, or vessels. We then used these maps to construct three-dimensional lesion segmentations. Using a validation dataset, Dice scores were computed for our lesion segmentations and other published segmentation networks using ground truth segmentations reviewed by radiologists. RESULTS: The COVID-to-Healthy generator eliminated high Hounsfield unit (HU) voxels within pulmonary lesions and replaced them with lower HU voxels. The generator did not distort normal anatomy such as vessels, airways, or fissures. The generated healthy images had higher gas content (2.45 ± 0.93 vs 3.01 ± 0.84 L, P < 0.001) and lower tissue density (1.27 ± 0.40 vs 0.73 ± 0.29 Kg, P < 0.001) than their corresponding original COVID-19 images, and they were not significantly different from those of the healthy images (P < 0.001). Using the validation dataset, lesion segmentations scored an average Dice score of 55.9, comparable to other weakly supervised networks that do require manual segmentations. CONCLUSION: Our CycleGAN model successfully segmented pulmonary lesions in mild and severe COVID-19 cases. Our model's performance was comparable to other published models; however, our model is unique in its ability to segment lesions without the need for manual segmentations.

A Simulated Prospective Evaluation of a Deep Learning Model for Real-Time Prediction of Clinical Deterioration Among Ward Patients.

OBJECTIVES: The National Early Warning Score, Modified Early Warning Score, and quick Sepsis-related Organ Failure Assessment can predict clinical deterioration. These scores exhibit only moderate performance and are often evaluated using aggregated measures over time. A simulated prospective validation strategy that assesses multiple predictions per patient-day would provide the best pragmatic evaluation. We developed a deep recurrent neural network deterioration model and conducted a simulated prospective evaluation. DESIGN: Retrospective cohort study. SETTING: Four hospitals in Pennsylvania. PATIENTS: Inpatient adults discharged between July 1, 2017, and June 30, 2019. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: We trained a deep recurrent neural network and logistic regression model using data from electronic health records to predict hourly the 24-hour composite outcome of transfer to ICU or death. We analyzed 146,446 hospitalizations with 16.75 million patient-hours. The hourly event rate was 1.6% (12,842 transfers or deaths, corresponding to 260,295 patient-hours within the predictive horizon). On a hold-out dataset, the deep recurrent neural network achieved an area under the precision-recall curve of 0.042 (95% CI, 0.04-0.043), comparable with logistic regression model (0.043; 95% CI 0.041 to 0.045), and outperformed National Early Warning Score (0.034; 95% CI, 0.032-0.035), Modified Early Warning Score (0.028; 95% CI, 0.027- 0.03), and quick Sepsis-related Organ Failure Assessment (0.021; 95% CI, 0.021-0.022). For a fixed sensitivity of 50%, the deep recurrent neural network achieved a positive predictive value of 3.4% (95% CI, 3.4-3.5) and outperformed logistic regression model (3.1%; 95% CI 3.1-3.2), National Early Warning Score (2.0%; 95% CI, 2.0-2.0), Modified Early Warning Score (1.5%; 95% CI, 1.5-1.5), and quick Sepsis-related Organ Failure Assessment (1.5%; 95% CI, 1.5-1.5). CONCLUSIONS: Commonly used early warning scores for clinical decompensation, along with a logistic regression model and a deep recurrent neural network model, show very poor performance characteristics when assessed using a simulated prospective validation. None of these models may be suitable for real-time deployment.

Role of adrenal vein sampling in primary aldosteronism: Impact of imaging, localization, and age.

BACKGROUND: The role of adrenal vein sampling (AVS) has been debated, with some authorities advocating selective use in younger patients (≤40 years), and those localized by preoperative imaging. We examined our experience to determine the impact of AVS in patients who routinely underwent AVS with a high success rate. METHODS: A retrospective cohort study was performed using a prospectively maintained database of patients referred for evaluation of PA (1997-2013). Patients were classified as localized (L) if a unilateral mass was identified on imaging, and non-localized (NL) otherwise. RESULTS: Of 367 patients, 94% (n = 345) underwent successful AVS. Seventy-two percent (n = 265) were L; AVS was lateralizing 58% (n = 214). AVS changed management in 43% of patients (n = 158). In patients ≤40 years, AVS changed management in 30% (n = 15). In patients ≤40 years with a ≥1 cm adrenal mass, 12% (n = 3) would have undergone unnecessary surgery based on imaging results alone; in patients >40 years with a ≥1 cm adrenal mass, 3% (n = 5) would have undergone wrong-side surgery, and 30% (n = 50) would have undergone unnecessary surgery based on imaging. CONCLUSION: AVS changed management in a significant minority of patients regardless of age and imaging findings. AVS should be routinely recommended in all patients with PA, to direct operative therapy. J. Surg. Oncol. 2016;113:532-537. © 2016 Wiley Periodicals, Inc.

Surgery for metastatic neuroendocrine tumors with occult primaries.

INTRODUCTION: Neuroendocrine tumors (NETs) frequently metastasize prior to diagnosis. Although metastases are often identifiable on conventional imaging studies, primary tumors, particularly those in the midgut, are frequently difficult to localize preoperatively. MATERIALS AND METHODS: Patients with metastatic NETs with intact primaries were identified. Clinical and pathologic data were extracted from medical records. Primary tumors were classified as localized or occult based on preoperative imaging. The sensitivities and specificities of preoperative imaging modalities for identifying the primary tumors were calculated. Patient characteristics, tumor features, and survival in localized and occult cases were compared. RESULTS: Sixty-one patients with an intact primary tumor and metastatic disease were identified. In 28 of these patients (46%), the primary tumor could not be localized preoperatively. A median of three different preoperative imaging studies were utilized. Patients with occult primaries were more likely to have a delay (>6 mo) in surgical referral from time of onset of symptoms (57% versus 27%, P = 0.02). Among the 28 patients with occult primary tumors, 18 (64%) were found to have radiographic evidence of mesenteric lymphadenopathy corresponding, in all but one case, to a small bowel primary. In all but three patients (89%), the primary tumor could be identified intraoperatively. CONCLUSION: The primary tumor can be identified intraoperatively in a majority of patients with metastatic NETs, irrespective of preoperative localization status. Referral for surgical management should not, therefore, be influenced by the inability to localize the primary tumor.

Regional recurrence after lymphadenectomy for clinically evident lymph node metastases from papillary thyroid cancer: a cohort study.

BACKGROUND: The presence of lymph node (LN) metastases in papillary thyroid cancer (PTC) has limited prognostic utility for predicting disease-specific survival. Pathologic features of the LNs beyond their presence and location do not factor into the AJCC staging system. Most LN metastases are microscopic. The natural history of patients with PTC and clinically evident LN metastases (CELNM) has not been well characterized. METHODS: Patients with CELNM from PTC undergoing lymph node dissection (LND) by a single surgeon were identified (1999-2009). Patients with and without recurrence were compared by clinical and pathological factors using Student's t test, Fisher's exact test, and Chi-squared test. Logistic regression was used to determine the association between recurrence and CELNM after adjustment for confounders. RESULTS: Ninety-two patients were identified who underwent surgery for CELNM. With a median follow-up of 27.5 months, 27 patients (29%) developed regional nodal recurrence with no disease-related deaths. Patients with and without recurrence were similar with respect to all clinical and pathologic variables analyzed except for the number of metastatic LNs. With a similar number of LNs sampled in both groups, the mean number of metastatic LNs was significantly higher in the group of patients with recurrence versus those without recurrence (12 vs. 6, p = 0.03) and remained significantly associated with the likelihood of recurrence (p = 0.009) in the logistic regression model. CONCLUSIONS: Patients with CELNM have an appreciable incidence of early regional cervical recurrence but not distant metastases. Among clinical and pathological factors evaluated, the number of metastatic LNs is associated with recurrence.

THE IMPACT OF VIDEO VISITS ON MEASURES OF CLINICAL EFFICIENCY AND REIMBURSEMENT.

INTRODUCTION: Telehealth is gaining more attention in multiple specialties, including urology. Video visits in particular have shown high satisfaction and cost-saving for patients. However, there has been little investigation into how video visits compare to traditional clinic visits on measures of clinical efficiency and reimbursement. METHODS: Our dataset included 250 video visits of established patients at Michigan Medicine Department of Urology and 250 in-person clinic visits with the same providers completed between July 2016 and July 2017. Information on visit completion and cancellation rates; cycle time (time from check in to check out); reimbursement; and patient out-of-pocket expenses was collected using the electronic medical record and billing data. RESULTS: Completion rates were similar between video and clinic visits (58% versus 61%, respectively; p=0.24). Average cycle time for video visits was significantly shorter compared to clinic visits (24 min vs 80 min, respectively; p<0.01). Neither average commercial payer reimbursement (p=0.21) nor average out-of-pocket expense (p=0.22) was statistically different between video and traditional visits. More video visits than clinic visits were billed as level 3 encounters (85% and 63%, respectively, p=0.002). CONCLUSIONS: Our study demonstrates that video visits have the potential to reduce the time patients spend on follow-up care without negatively impacting reimbursement rates. However, these visits could reduce average billing levels. These findings suggest that the incorporation of video visits into practice may be more efficient for patients but may also reduce billing levels.

Integrating 31-Gene Expression Profiling With Clinicopathologic Features to Optimize Cutaneous Melanoma Sentinel Lymph Node Metastasis Prediction.

National guidelines recommend sentinel lymph node biopsy (SLNB) be offered to patients with > 10% likelihood of sentinel lymph node (SLN) positivity. On the other hand, guidelines do not recommend SLNB for patients with T1a tumors without high-risk features who have < 5% likelihood of a positive SLN. However, the decision to perform SLNB is less certain for patients with higher-risk T1 melanomas in which a positive node is expected 5%-10% of the time. We hypothesized that integrating clinicopathologic features with the 31-gene expression profile (31-GEP) score using advanced artificial intelligence techniques would provide more precise SLN risk prediction. METHODS: An integrated 31-GEP (i31-GEP) neural network algorithm incorporating clinicopathologic features with the continuous 31-GEP score was developed using a previously reported patient cohort (n = 1,398) and validated using an independent cohort (n = 1,674). RESULTS: Compared with other covariates in the i31-GEP, the continuous 31-GEP score had the largest likelihood ratio (G2 = 91.3, P < .001) for predicting SLN positivity. The i31-GEP demonstrated high concordance between predicted and observed SLN positivity rates (linear regression slope = 0.999). The i31-GEP increased the percentage of patients with T1-T4 tumors predicted to have < 5% SLN-positive likelihood from 8.5% to 27.7% with a negative predictive value of 98%. Importantly, for patients with T1 tumors originally classified with a likelihood of SLN positivity of 5%-10%, the i31-GEP reclassified 63% of cases as having < 5% or > 10% likelihood of positive SLN, for a more precise, personalized, and clinically actionable SLN-positive likelihood estimate. CONCLUSION: These data suggest the i31-GEP could reduce the number of SLNBs performed by identifying patients with likelihood under the 5% threshold for performance of SLNB and improve the yield of positive SLNBs by identifying patients more likely to have a positive SLNB.

Video Visits as a Substitute for Urological Clinic Visits.

OBJECTIVE: To evaluate whether video visits were being used as substitutes to clinic visits prior to COVID-19 at our institution's outpatient urology centers. METHODS: We reviewed 600 established patient video visits completed by 13 urology providers at a tertiary academic center in southeast Michigan. We compared these visits to a random, stratified sample of established patient clinic visits. We assessed baseline demographics and visit characteristics for both groups. We defined our primary outcome ("revisit rate") as the proportion of additional healthcare evaluation (ie, office, emergency room, hospitalization) by a urology provider within 30 days of the initial encounter. RESULTS: Patients seen by video visit tended to be younger (51 vs 61 years, P <.001), would have to travel further for a clinic appointment (82 vs 68 miles, P <.001), and were more likely to be female (36 vs 28%, P = .001). The most common diagnostic groups evaluated through video visits were nephrolithiasis (40%), oncology (18%) and andrology (14.3%). While the 30-day revisit rates were higher for clinic visits (4.3% vs 7.5%, P = .01) primarily due to previously scheduled appointments, revisits due to medical concerns were similar across both groups (0.5% vs 0.67%; P = .60). CONCLUSIONS: Video visits can be used to deliver care across a broad range of urologic diagnoses and can serve as a substitute for clinic visits.

Perioperative Hypoxia Secondary to Phenazopyridine-induced Methemoglobinemia in an Adolescent Patient Without Renal Insufficiency or Overdose: An Unusual Case.

Phenazopyridine is a common, well-tolerated medication with minimal side effects. Severe side effects are rare and include methemoglobinemia in setting of overdose, elderly patients, renal insufficiency, and chronic use. Here, we report a case of methemoglobinemia-induced perioperative hypoxia in an adolescent patient without renal insufficiency or overdose which has not been reported previously. This case underscores the importance of judicious use of this medication in all patients but notably in pediatric patients and those with chronic lung disease.

Strategies for Managing Smart Pump Alarm and Alert Fatigue: A Narrative Review.

Although smart infusion pumps are intended to prevent medication errors by alerting users about doses that exceed set thresholds, a large number of clinically insignificant alarms and alerts create the potential for alert and alarm fatigue. We searched the PubMed, Scopus, and Cumulative Index to Nursing and Allied Health Literature (CINAHL) databases for peer-reviewed literature (January 1, 2004-August 31, 2017) on managing smart pump alerts, alarms, and related fatigue. Twenty-nine articles that met the inclusion criteria were reviewed and organized into themes. Smart pumps give users two types of signals: alarms that indicate mechanical issues such as occlusion, air in the line, or low battery; and clinical alerts that indicate that a programmed dose exceeds a predefined safety limit. Mechanical alarms occur with greater frequency than clinical alerts, but alarms and alerts vary widely by pump model, patient population, time of day, month, and type of drug. Several causes of clinically insignificant alerts and alarms may be actionable, and strategies proposed in the literature include development of a multidisciplinary team to oversee the quality improvement effort with involvement of end users, standardization of medication administration practices, widening of drug limit library thresholds when clinically appropriate, maintaining up-to-date drug limit libraries, and interoperability. Whereas many strategies have been proposed, and case studies have been reported, none have been rigorously evaluated. In addition, more research is needed related to managing occlusion and air-in-line alarms, especially for complicated infusions. Future work should focus on the evaluation of specific and replicable alert and alarm reduction strategies with a greater emphasis on quantitative metrics.

Emergency Department Switching and Duplicate Computed Tomography Scans in Patients With Kidney Stones.

OBJECTIVE: To test whether duplicate imaging relates to a lack of information sharing among providers, we measured the association between emergency department (ED) switching during a kidney stone episode and receipt of a repeat computed tomography (CT) scan. METHODS: Using the MarketScan Commercial Claims and Encounters Database, we identified adults between the ages 18 and 64 with an ED visit for a diagnosis of kidney stones. Among patients who had an abdominal or pelvic CT scan at their initial encounter, we then determined the subset that made an ED revisit within 30 days of their first, distinguishing between those to the same vs a different ED. Finally, we fit multivariable logistic regression models to estimate the risk of receiving a repeat CT scan associated with ED switching. RESULTS: Twelve percent of patients who received a CT scan at their initial ED encounter had a revisit within 30 days of discharge. One-third of their revisits were made to a different ED than the index one. Duplicate CT scans were obtained at nearly 40% of all revisits. On multivariable analysis, the risk of receiving a repeat CT was 12% higher if this revisit was made to a different ED (risk ratio, 1.12; 95% confidence interval, 1.03-1.21; P = .010). CONCLUSION: Our study reveals that ED switching during an acute kidney stone episode is associated with higher levels of repeat CT imaging. These findings support the role of better health information exchange among providers to help reduce waste in the health-care system.

Delayed Nephron Sparing Surgery for Grade IV Renal Injury.

Renal injuries are a common occurrence in many trauma cases. The management of these cases varies, but, currently, a conservative, nonoperative approach is the norm. In cases where an operative intervention may be necessary, emergency total nephrectomies are the most commonly performed procedure reported in the literature. There is a dearth in the reporting of other surgical approaches, especially the delayed nephron sparing approach. Here, we present a unique case of a 29-year-old male that underwent a delayed nephron sparing nephrectomy for a persistent urinoma despite appropriate noninvasive therapy.

Epithelial cell adhesion molecule-positive circulating tumor cells as predictive biomarker in patients with prostate cancer.

OBJECTIVE: To assess the use of circulating tumor cells (CTCs) as a longitudinal endpoint factor for clinical monitoring of patients with prostate cancer and to evaluate the association among the baseline CTC number, various clinical characteristics, and survival. MATERIALS AND METHODS: The CTCs were enumerated using the CellSearch Food and Drug Administration-cleared CTC kit in 202 patients with prostate cancer. Variables, including metastatic site, prostate-specific antigen level, Gleason score, testosterone level, and use of androgen treatment, were tested for association with the CTC number. The probability of patient survival over time was estimated using the Kaplan-Meier method. RESULTS: The baseline CTC numbers were strongly associated with survival (P <.0001), with overall survival significantly poorer in patients with ≥5 CTCs. Significantly greater CTC numbers were observed in patients with bone metastasis (mean 41.12 CTCs) than in those with lymph node metastasis (mean 2.53 CTC, P = .026). Analysis of the association between the CTC count and prostate-specific antigen level revealed a weak positive correlation (correlation coefficient r = 0.2695, P = .0007). The CTC number also correlated with the Gleason score (P = .0138) and lower testosterone level (P <.0001). Patients without androgen depletion had significantly lower CTC numbers (mean 2.70) than those with androgen depletion (mean 26.39, P <.0001). CONCLUSION: The baseline CTC counts were predictive of patient survival and correlated significantly with the clinical characteristics of patients with prostate cancer. Our study results have confirmed previous findings that support the use of CTC enumeration as a prognostic biomarker for patients with prostate cancer.