Deep Learning-based Prescription of Cardiac MRI Planes

PURPOSE

To develop and evaluate a system to prescribe imaging planes for cardiac MRI based on deep learning (DL)-based localization of key anatomic landmarks.

MATERIALS AND METHODS

Annotated landmarks on 892 long-axis (LAX) and 493 short-axis (SAX) cine steady-state free precession series from cardiac MR images were retrospectively collected between February 2012 and June 2017. U-Net-based heatmap regression was used for localization of cardiac landmarks, which were used to compute cardiac MRI planes. Performance was evaluated by comparing localization distances and plane angle differences between DL predictions and ground truth. The plane angulations from DL were compared with those prescribed by the technologist at the original time of acquisition. Data were split into 80% for training and 20% for testing, and results confirmed with fivefold cross-validation.

RESULTS

On LAX images, DL localized the apex within mean 12.56 mm ± 19.11 (standard deviation) and the mitral valve (MV) within 7.68 mm ± 6.91. On SAX images, DL localized the aortic valve within 5.78 mm ± 5.68, MV within 5.90 mm ± 5.24, pulmonary valve within 6.55 mm ± 6.39, and tricuspid valve within 6.39 mm ± 5.89. On the basis of these localizations, average angle bias and mean error of DL-predicted imaging planes relative to ground truth annotations were as follows: SAX, -1.27° ± 6.81 and 4.93° ± 4.86; four chambers, 0.38° ± 6.45 and 5.16° ± 3.80; three chambers, 0.13° ± 12.70 and 9.02° ± 8.83; and two chamber, 0.25° ± 9.08 and 6.53° ± 6.28, respectively.

CONCLUSION

DL-based anatomic localization is a feasible strategy for planning cardiac MRI planes. This approach can produce imaging planes comparable to those defined by ground truth landmarks.© RSNA, 2019 Supplemental material is available for this article.

Automated selection of myocardial inversion time with a convolutional neural network: Spatial temporal ensemble myocardium inversion network (STEMI-NET)

PURPOSE

Delayed enhancement imaging is an essential component of cardiac MRI, which is used widely for the evaluation of myocardial scar and viability. The selection of an optimal inversion time (TI) or null point (TINP ) to suppress the background myocardial signal is required. The purpose of this study was to assess the feasibility of automated selection of TINP using a convolutional neural network (CNN). We hypothesized that a CNN may use spatial and temporal imaging characteristics from an inversion-recovery scout to select TINP , without the aid of a human observer.

METHODS

We retrospectively collected 425 clinically acquired cardiac MRI exams performed at 1.5 T that included inversion-recovery scout acquisitions. We developed a VGG19 classifier ensembled with long short-term memory to identify the TINP . We compared the performance of the ensemble CNN in predicting TINP against ground truth, using linear regression analysis. Ground truth was defined as the expert physician annotation of the optimal TI. In a backtrack approach, saliency maps were generated to interpret the classification outcome and to increase the model's transparency.

RESULTS

Prediction of TINP from our ensemble VGG19 long short-term memory closely matched with expert annotation (ρ = 0.88). Ninety-four percent of the predicted TINP were within ±36 ms, and 83% were at or after expert TI selection.

CONCLUSION

In this study, we show that a CNN is capable of automated prediction of myocardial TI from an inversion-recovery experiment. Merging the spatial and temporal characteristics of the VGG-19 and long short-term-memory CNN structures appears to be sufficient to predict myocardial TI from TI scout.

Automated CT and MRI Liver Segmentation and Biometry Using a Generalized Convolutional Neural Network

Purpose

To assess feasibility of training a convolutional neural network (CNN) to automate liver segmentation across different imaging modalities and techniques used in clinical practice and apply this to enable automation of liver biometry.

Methods

We trained a 2D U-Net CNN for liver segmentation in two stages using 330 abdominal MRI and CT exams acquired at our institution. First, we trained the neural network with non-contrast multi-echo spoiled-gradient-echo (SGPR)images with 300 MRI exams to provide multiple signal-weightings. Then, we used transfer learning to generalize the CNN with additional images from 30 contrast-enhanced MRI and CT exams.We assessed the performance of the CNN using a distinct multi-institutional data set curated from multiple sources (n = 498 subjects). Segmentation accuracy was evaluated by computing Dice scores. Utilizing these segmentations, we computed liver volume from CT and T1-weighted (T1w) MRI exams, and estimated hepatic proton- density-fat-fraction (PDFF) from multi-echo T2*w MRI exams. We compared quantitative volumetry and PDFF estimates between automated and manual segmentation using Pearson correlation and Bland-Altman statistics.

Results

Dice scores were 0.94 ± 0.06 for CT (n = 230), 0.95 ± 0.03 (n = 100) for T1w MR, and 0.92 ± 0.05 for T2*w MR (n = 169). Liver volume measured by manual and automated segmentation agreed closely for CT (95% limit-of-agreement (LoA) = [-298 mL, 180 mL]) and T1w MR (LoA = [-358 mL, 180 mL]). Hepatic PDFF measured by the two segmentations also agreed closely (LoA = [-0.62%, 0.80%]).

Conclusions

Utilizing a transfer-learning strategy, we have demonstrated the feasibility of a CNN to be generalized to perform liver segmentations across different imaging techniques and modalities. With further refinement and validation, CNNs may have broad applicability for multimodal liver volumetry and hepatic tissue characterization.

Voice Enabled Framework to Support Post-Surgical Discharge Monitoring

Unplanned surgical readmissions pose a challenging problem for the American healthcare system. We propose to combine consumer electronic voice recognition technology with the FHIR standard to create a post-surgical discharge monitoring app to identify and alert physicians to a patient's deteriorating status.

The economic impact of surgical care for morbidly obese endometrial cancer patients: a nationwide study

BACKGROUND

Obesity significantly impacts the cost of cancer treatment, yet the impact of morbid obesity on inpatient hospital charges related to endometrial cancer treatment is not well-defined.

OBJECTIVES

The purpose of this study was to determine the charges that are associated with inpatient surgery, hospitalization, and postoperative care of morbidly obese patients with endometrial cancer.

STUDY DESIGN

Data were obtained from the National Inpatient Sample from 2010. Chi-square test, t-test, and linear regression were used for statistical analyses.

RESULTS

Six thousand five hundred sixty patients who underwent hysterectomy for endometrial cancer were identified. Mean age was 62 years (range, 22-99 years). The majority were white (78%), and the remainder were black (10%), Hispanic, (8%), Asian (3%), and Native American (1%). Insurance types were private (45%), Medicare (45%), Medicaid (5%), and uninsured (7%). One thousand eighty-eight of these patients (17%) were coded as morbidly obese. The mean postoperative stay for the morbidly obese was 4.0 days (range, 0-46 days) compared with 3.5 days (range, 0-81 days) for the non-morbidly obese patients (P < .01). Morbidly obese patients required more intensive care with mechanical ventilation (5.5% vs 1.6%; P < .01). The median hospital charges were higher for morbidly obese patients compared with their counterparts ($46,654 vs $41,164; P < .01). After adjustment for charges that were associated with insurance type, hospital type, and the surgery that was performed, the incremental increase in hospital charges that were associated with treating the morbidly obese patient was $5096 per patient (95% confidence interval, $2593-$7598; P < .01).

CONCLUSION

In this economic analysis, the health care charges that were associated with inpatient endometrial cancer treatment in the morbidly obese patient was significantly higher compared the non-morbidly obese patient. Resources are needed to support the needs of this population, and programs to encourage weight loss and optimize general health should be encouraged.

Robotic versus laparoscopic versus open surgery in morbidly obese endometrial cancer patients - a comparative analysis of total charges and complication rates

OBJECTIVE

To compare the complications and charges of robotic vs. laparoscopic vs. open surgeries in morbidly obese patients treated for endometrial cancer.

METHODS

Data were obtained from the Nationwide Inpatient Sample from 2011. Chi-squared, Wilcoxon rank sum two-sample tests, and multivariate analyses were used for statistical analyses.

RESULTS

Of 1087 morbidly obese (BMI ≥40kg/m(2)) endometrial cancer patients (median age: 59years, range: 22 to 89), 567 (52%) had open surgery (OS), 98 (9%) laparoscopic (LS), and 422 (39%) robotic surgery (RS). 23% of OS, 13% of LS, and 8% of RS patients experienced an intraoperative or postoperative complication including: blood transfusions, mechanical ventilation, urinary tract injury, gastrointestinal injury, wound debridement, infection, venous thromboembolism, and lymphedema (p<0.0001). RS and LS patients were less likely to receive blood transfusions compared to OS (5% and 6% vs. 14%, respectively; p<0.0001). The median lengths of hospitalization for OS, LS, and RS patients were 4, 1, and 1days, respectively (p<0.0001). Median total charges associated with OS, LS, and RS were $39,281, $40,997, and $45,030 (p=0.037), respectively.

CONCLUSIONS

In morbidly obese endometrial cancer patients, minimally invasive robotic or laparoscopic surgeries were associated with fewer complications and less days of hospitalization relative to open surgery. Compared to laparoscopic approach, robotic surgeries had comparable rates of complications but higher charges.

Comparison of bevacizumab alone or with chemotherapy in recurrent ovarian cancer patients

BACKGROUND

To compare the efficacy of chemotherapy (C) combined with bevacizumab (Bev) versus Bev alone in recurrent, heavily pretreated epithelial ovarian cancer (EOC).

METHODS

A multicenter analysis of patients treated from 2004 to 2011 was performed. Demographic, treatment, response, and adverse event information were obtained. Progression-free (PFS) and overall survival (OS) were analyzed.

RESULTS

Of 277 patients (median age: 58years), the majority had Stage III and IV (86%) disease, and 72% had serous histology. 244 (88%) were treated with C+Bev and 33 (12%) with Bev. Corresponding median progression-free survival (PFS) was 8.7 and 6.7months, and median overall survival (OS) was 14.3 and 10.5months, respectively. The chemotherapeutic agents combined with Bev and the median OS include: pegylated liposomal doxorubicin (n=19, OS of 20.4months), taxanes (n=55, OS of 20.2months), gemcitabine (n=106, OS of 14.1months), topotecan (n=43, OS of 13months), and cyclophosphamide (n=21, OS of 13months). There was no significant difference in toxicities between the C+Bev vs. Bev alone group.

CONCLUSION

This retrospective analysis supports that combination chemotherapy and bevacizumab prolongs PFS and OS compared with bevacizumab alone.

The centralization of robotic surgery in high-volume centers for endometrial cancer patients--a study of 6560 cases in the U.S

OBJECTIVE

To evaluate the hospital and patient factors associated with robotic surgery for endometrial cancer in the United States.

METHODS

Data was obtained from the Nationwide Inpatient Sample from the year 2010. Chi-squared and multivariate analyses were used for statistical analysis.

RESULTS

Of the 6560 endometrial cancer patients who underwent surgery, the median age was 62 (range: 22 to 99). 1647 (25%) underwent robotic surgery, 820 (13%) laparoscopic, and 4093 (62%) had open surgery. The majority was White (65%). Hospitals with 76 or more hysterectomy cases for endometrial cancer patients per year (4% of hospitals in the study) performed 31% of all hysterectomies and 40% of all robotic hysterectomies (p<0.01). 29% of Whites had robotic surgery compared to 15% of Hispanics, 12% of Blacks, and 11% of Asians (p<0.01). Patients with upper-middle and high incomes underwent robotic surgery more than patients with low or middle incomes (p<0.01). 27% of Medicare patients and 26% of patients with private insurance had robotic surgery compared to only 14% of Medicaid patients and 12% of uninsured patients (p<0.01).

CONCLUSIONS

The majority of robotic surgeries for endometrial cancer were performed at a small number of high-volume hospitals in the United States. Socioeconomic status, insurance type, and race were also important predictors for the use of RS. Further studies are warranted to better understand the barriers to receiving minimally invasive surgery.

Survival differences in women with serous tubal, ovarian, peritoneal, and uterine carcinomas

OBJECTIVE

The fallopian tube has been implicated as the primary origin of pelvic serous cancers. We proposed to determine the survival outcomes of serous tubal, ovarian, peritoneal, and uterine cancer patients.

STUDY DESIGN

Data were obtained from the National Cancer Institute between 2004 and 2009. Kaplan-Meier and Cox proportional hazards models were used for analysis.

RESULTS

Of 12,336 high-grade serous cancer patients, 563 were tubal (TC), 8560 ovarian (OC), 1037 primary peritoneal (PPC), and 2176 uterine cancer (USC). The median ages of these patients were 63 vs 62 vs 67 vs 68 years, respectively. The majority were white (89% vs 88% vs 91% vs 74%). The overall 5 year, disease-specific survival was 37%. The survivals of those with TC, OC, PPC, and USC were 50%, 37%, 26%, and 40% (P < .01). There was no detailed staging on PPC cancers. Adjusted for stage, the survival of those with stage I, II, III, and IV TC were 73%, 62%, 44%, and 22% (P < .01), OC were 83%, 64%, 34%, and 15% (P < .01), and USC were 88%, 72%, 55%, and 17% (P < .01). On multivariate analysis, younger age, white race, earlier stage, and tubal origin were independent predictors for improved survival.

CONCLUSION

In advanced-staged serous cancer patients, tubal cancer patients have better survivals compared with ovarian, peritoneal, and uterine cancer.

Foreign- vs US-born Asians and the association of type I uterine cancer

OBJECTIVE

The purpose of this study was to determine the association of type I endometrioid uterine cancer in US-born vs immigrant Asian women.

STUDY DESIGN

Data were obtained from the Surveillance, Epidemiology, and End Results Program from 2001-2009. Chi-squared, Kaplan-Meier, and binomial logistic regression analyses were used for statistics.

RESULTS

Of 4834 Asian women with uterine cancer, 62% were US-born and 38% were immigrants. Of these women, 2972 (61%) had type I (grade 1 or 2, endometrioid histologic type) uterine cancer. Compared with patients with type II disease (grade 3, clear cell and serous histologic type), patients with type I disease were younger (age 55 vs 59 years; P < .01) and had lower stage disease (90% vs 71%; P < .01). US-born Asian women had a significantly higher proportion of type I uterine cancers in contrast to their immigrant counterparts (65% vs 56%; P < .01). Of all immigrants, the proportion of type I cancers was lowest in Japanese women followed by Chinese and Filipino women, respectively (48% vs 52% vs 58%; P < .01). The 5-year disease-specific survivals of US-born vs immigrant Asian women with type I cancer was 92% for both groups. Over 3 time periods (2001-2003, 2004-2006, and 2007-2009), there was an increase in type I cancers among US-born Asian women (61% to 65% to 68%; P < .01).

CONCLUSION

US-born Asian women are more likely to be diagnosed with type I uterine cancer compared with immigrants. Over the study period, there was a trend towards an increase in type I cancers among US-born Asian women.

Bevacizumab in treatment of high-risk ovarian cancer--a cost-effectiveness analysis

The objective of this study was to evaluate a cost-effectiveness strategy of bevacizumab in a subset of high-risk advanced ovarian cancer patients with survival benefit. Methods. A subset analysis of the International Collaboration on Ovarian Neoplasms 7 trial showed that additions of bevacizumab (B) and maintenance bevacizumab (mB) to paclitaxel (P) and carboplatin (C) improved the overall survival (OS) of high-risk advanced cancer patients. Actual and estimated costs of treatment were determined from Medicare payment. Incremental cost-effectiveness ratio per life-year saved was established. Results. The estimated cost of PC is $535 per cycle; PCB + mB (7.5 mg/kg) is $3,760 per cycle for the first 6 cycles and then $3,225 per cycle for 12 mB cycles. Of 465 high-risk stage IIIC (>1 cm residual) or stage IV patients, the previously reported OS after PC was 28.8 months versus 36.6 months in those who underwent PCB + mB. With an estimated 8-month improvement in OS, the incremental cost-effectiveness ratio of B was $167,771 per life-year saved. Conclusion. In this clinically relevant subset of women with high-risk advanced ovarian cancer with overall survival benefit after bevacizumab, our economic model suggests that the incremental cost of bevacizumab was approximately $170,000.

MiR-378 as a biomarker for response to anti-angiogenic treatment in ovarian cancer

OBJECTIVE

To determine the role of miR-378 as a biomarker for anti-angiogenic therapy response in ovarian cancer.

METHODS

Expression of miR-378 was analyzed in ovarian cancer cell lines and human tumors vs. normal ovarian epithelial cells by qRT-PCR. After miR-378 transfection in SKOV3 cells, dysregulated genes were identified using microarray. Data from The Cancer Genome Atlas (TCGA) was utilized to correlate miR-378 expression with progression-free survival (PFS) among patients treated with anti-angiogenic therapy by using Kaplan-Meier and Cox proportional hazards.

RESULTS

MiR-378 was overexpressed in ovarian cancer cells and tumors vs. normal ovarian epithelial cells. Overexpressing miR-378 in ovarian cancer cells altered expression of genes associated with angiogenesis (ALCAM, EHD1, ELK3, TLN1), apoptosis (RPN2, HIPK3), and cell cycle regulation (SWAP-70, LSM14A, RDX). In the TCGA dataset, low vs. high miR-378 expression was associated with longer PFS in a subset of patients with recurrent ovarian cancer treated with bevacizumab (9.2 vs. 4.2months; p=0.04). On multivariate analysis, miR-378 expression was an independent predictor for PFS after anti-angiogenic treatment (HR=2.04, 95% CI: 1.12-3.72; p=0.02). Furthermore, expression levels of two miR-378 targets (ALCAM and EHD1) were associated with PFS in this subgroup of patients who received anti-angiogenic therapy (9.4 vs. 4.2months, p=0.04 for high vs. low ALCAM; 7.9 vs. 2.3months, p<0.01 for low vs. high EHD1).

CONCLUSIONS

Our data suggest that miR-378 is overexpressed in ovarian cancer cells and tumors vs. normal ovarian epithelial cells. MiR-378 and its downstream targets may serve as markers for response to anti-angiogenic therapy.

Applications for oncologic drugs: a descriptive analysis of the oncologic drugs advisory committee reviews

BACKGROUND

Despite advances in cancer research, the majority of drug applications submitted to the U.S. Food and Drug Administration (FDA) are not approved. It is important to identify the concerns of the Oncologic Drugs Advisory Committee (ODAC) from rejected applications.

METHODS

All applications referred to the ODAC from 2001 to 2012 were reviewed.

RESULTS

Of 46 applications, 31 (67%) were for full and 15 (33%) were for supplemental approval, 34 (74%) were for solid and 12 (26%) were for hematologic tumors. In all, 22 (48%) were not approved. ODAC comments addressed missing or inadequate data (65%), excessive toxicity (55%), inappropriate study endpoints (45%), poor study design (40%), and insufficient sample size (30%). To define efficacy, 19 applications used response rates (RR) (median = 38%), and 19 applications used hazard ratios (HR) (median = 0.67). For all organ systems combined, the median cumulative grade 3 or 4 toxicity was 64%. Drugs with higher RR, lower HR, and lower toxicity were more likely to be approved versus other drugs (89% vs. 45%; p = .02). Over time (2001-2004, 2005-2008, 2009-2012), there was an increase in the following: number of applications submitted for review (from 11 to 12 to 23, respectively), number of approvals (from 6 to 6 to 12, respectively), and proportion of trials using progression-free survival as a primary endpoint (from 0% to 50% to 70%, respectively; p = .01).

CONCLUSION

Of all applications, common ODAC concerns included inadequate data, excessive toxicity, and inappropriate study endpoints. Over time, there was an approximate doubling of FDA application submissions and approved oncology drugs.

The inhibition of miR-21 promotes apoptosis and chemosensitivity in ovarian cancer

BACKGROUND

MicroRNAs have been implicated in tumorigenesis, drug resistance, and prognosis in cancer. We investigated the role of microRNA-21 (miR-21) in regulating ovarian cancer drug resistance.

METHODS

We used parental and cisplatin resistant ovarian cell lines to demonstrate the role of miR-21 in drug resistance and investigated the gene targets of miR-21. Fresh tumor specimens were used to validate our in vitro findings.

RESULTS

Cisplatin resistant ovarian cells were four-fold more resistant compared to the parental cell line. MiR-21 was overexpressed in the resistant cell line on microRNA microarray, which was subsequently validated with qRT-PCR. Using anti-microRNA inhibitors, we demonstrated that miR-21 attenuation reversed the drug resistant phenotype in both the resistant and parental cell lines. The inhibition of miR-21 induced apoptosis based on annexin V-FITC immunostaining. Using Western blot analysis, miR-21 knockdown enhanced the expression of tumor suppressor PDCD4, and attenuated apoptosis inhibitor c-IAP2. Using 101 specimens from advanced ovarian cancer patients enrolled in The Cancer Genome Atlas, we found that women with tumors that overexpressed miR-21 were associated with a shorter progression-free survival.

CONCLUSION

Our data suggest that miR-21 regulates drug resistance via apoptosis and cellular survival pathways. Targeting miR-21 may have clinical utility in the treatment of resistant ovarian cancer.

Roles for substance P and gastrin-releasing peptide as neurotransmitters released by primary afferent pruriceptors

Recent studies support roles for neurokinin-1 (NK-1) and gastrin-releasing peptide (GRP) receptor-expressing spinal neurons in itch. We presently investigated expression of substance P (SP) and GRP in pruritogen-responsive primary sensory neurons and roles for these neuropeptides in itch signaling. Responses of dorsal root ganglion (DRG) cells to various pruritogens were observed by calcium imaging. DRG cells were then processed for SP, GRP, and isolectin B-4 (IB4; a marker for nonpeptidergic neurons) immunofluorescence. Of pruritogen-responsive DRG cells, 11.8-26.8%, 21.8-40.0%, and 21.4-26.8% were immunopositive for SP, GRP, and IB4, respectively. In behavioral studies, both systemic and intrathecal administration of a NK-1 receptor antagonist significantly attenuated scratching evoked by chloroquine and a protease-activated receptor 2 agonist, SLIGRL, but not histamine, bovine adrenal medulla peptide 8-22 (BAM8-22), or serotonin. Systemic or intrathecal administration of a GRP receptor antagonist attenuated scratching evoked by chloroquine and SLIGRL but not BAM8-22 or histamine. The GRP receptor antagonist enhanced scratching evoked by serotonin. These results indicate that SP and GRP expressed in primary sensory neurons are partially involved as neurotransmitters in histamine-independent itch signaling from the skin to the spinal cord.

Cross-sensitization of histamine-independent itch in mouse primary sensory neurons

Overexpression of pruritogens and their precursors may contribute to the sensitization of histamine-dependent and -independent itch-signaling pathways in chronic itch. We presently investigated self- and cross-sensitization of scratching behavior elicited by various pruritogens, and their effects on primary sensory neurons. The MrgprC11 agonist BAM8-22 exhibited self- and reciprocal cross-sensitization of scratching evoked by the protease-activated receptor-2 (PAR-2) agonist SLIGRL. The MrgprA3 agonist chloroquine unidirectionally cross-sensitized BAM8-22-evoked scratching. Histamine unidirectionally cross-sensitized scratching evoked by chloroquine and BAM8-22. SLIGRL unidirectionally cross-sensitized scratching evoked by chloroquine. Dorsal root ganglion (DRG) cells responded to various combinations of pruritogens and algogens. Neither chloroquine, BAM8-22 nor histamine had any effect on responses of DRG cell responses to subsequently applied pruritogens, implying that their behavioral self- and cross-sensitization effects are mediated indirectly. SLIGRL unilaterally cross-sensitized responses of DRG cells to chloroquine and BAM8-22, consistent with the behavioral data. These results indicate that unidirectional cross-sensitization of histamine-independent itch-signaling pathways might occur at a peripheral site through PAR-2. PAR-2 expressed in pruriceptive nerve endings is a potential target to reduce sensitization associated with chronic itch.