Multi-institutional Dataset Research Articles

MP42-09 NEIGHBORHOOD DISADVANTAGE CORRELATES WITH OUTCOMES IN MUSCLE INVASIVE BLADDER CANCER

You have accessJournal of UrologyCME1 May 2022MP42-09 NEIGHBORHOOD DISADVANTAGE CORRELATES WITH OUTCOMES IN MUSCLE INVASIVE BLADDER CANCER Devon Langston, Sarah Beebe, Tasha Posid, Victor Heh, and Kamal Pohar Devon LangstonDevon Langston More articles by this author , Sarah BeebeSarah Beebe More articles by this author , Tasha PosidTasha Posid More articles by this author , Victor HehVictor Heh More articles by this author , and Kamal PoharKamal Pohar More articles by this author View All Author Informationhttps://doi.org/10.1097/JU.0000000000002608.09AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Area deprivation index (ADI), a measure of neighborhood socioeconomic disadvantage (NSD), has never been studied in bladder cancer. ADI is an integrated metric of neighborhood-level material disadvantage, which incorporates factors including income, employment, education, and housing quality of a local area as derived from US census data. We evaluated the association between NSD, demographics, and outcomes for Ohioans diagnosed with muscle invasive bladder cancer (MIBC) treated by radical cystectomy (RC) at a tertiary medical center. METHODS: 643 consecutive MIBC patients were identified and entered in a RC database (2012-2018). Patient clinical, pathological, and surgical characteristics were compared across ADI quartiles (least, low, high, or most NSD) using Chi2 tests (categorical) and ANOVA (continuous). RESULTS: Relative to all other groups, patients in the high and most NSD quartiles tend to be female (23.2%, P=0.011 in four-quartile comparison). The most NSD quartile tends to be non-white (6.7%, P=0.273), and current smokers (29.7%, P=0.001), whereas patients in the least NSD quartile are more likely to be ex-smokers (57.6%, P=0.046). Patients in the most NSD quartile are less likely to complete neo-adjuvant chemotherapy prior to RC (58.2%, P=0.034). Patients in the most NSD quartile have a higher rate of nodal metastatic disease (N+) (2.9%, P=0.622), and lympovascular invasion (LVI) (30.1%, P=0.434), pathological T-stage 15.9% pT4, P=0.89). Patients in the most NSD quartile have a soft-tissue positive margin rate of 5.44% (P=0.407). Additionally, patients in the most NSD quartile tend to have a shorter time from cystectomy to death (median 17 months [IQR 9-42], P=0.489). CONCLUSIONS: Although the size of our dataset fails to achieve statistical significance for many comparisons, a series of consistent trends are noted. MIBC patients in the most NSD quartile who are treated by cystectomy receive NAC less often, have higher rates of N+ and LVI, which may be associated with worse overall survival. Commissioning a larger multi-institutional dataset would allow greater statistical power. Source of Funding: None © 2022 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 207Issue Supplement 5May 2022Page: e734 Advertisement Copyright & Permissions© 2022 by American Urological Association Education and Research, Inc.MetricsAuthor Information Devon Langston More articles by this author Sarah Beebe More articles by this author Tasha Posid More articles by this author Victor Heh More articles by this author Kamal Pohar More articles by this author Expand All Advertisement PDF DownloadLoading ...

MP24-11 IS PATHOLOGICAL UPSTAGING TO T3 RENAL CELL CARCINOMA ASSOCIATED WITH A SIMILAR PROGNOSIS TO NON-UPSTAGED PATHOLOGIC T3 DISEASE? A MULTICENTER ANALYSIS

You have accessJournal of UrologyCME1 May 2022MP24-11 IS PATHOLOGICAL UPSTAGING TO T3 RENAL CELL CARCINOMA ASSOCIATED WITH A SIMILAR PROGNOSIS TO NON-UPSTAGED PATHOLOGIC T3 DISEASE? A MULTICENTER ANALYSIS Ryan Nasseri, Kevin Hakimi, Ava Saidian, Riccardo Autorino, Brian Lane, Yasuha Fujii, Michele Marchioni, Hajime Tanaka, Dattatraya Patil, Francesco Porpiglia, Sabrina Noyes, Chandru Sundaram, James Porter, Andrea Minervini, Umberto Capitianio, Francesco Montorsi, Viraj Master, and Ithaar Derweesh Ryan NasseriRyan Nasseri More articles by this author , Kevin HakimiKevin Hakimi More articles by this author , Ava SaidianAva Saidian More articles by this author , Riccardo AutorinoRiccardo Autorino More articles by this author , Brian LaneBrian Lane More articles by this author , Yasuha FujiiYasuha Fujii More articles by this author , Michele MarchioniMichele Marchioni More articles by this author , Hajime TanakaHajime Tanaka More articles by this author , Dattatraya PatilDattatraya Patil More articles by this author , Francesco PorpigliaFrancesco Porpiglia More articles by this author , Sabrina NoyesSabrina Noyes More articles by this author , Chandru SundaramChandru Sundaram More articles by this author , James PorterJames Porter More articles by this author , Andrea MinerviniAndrea Minervini More articles by this author , Umberto CapitianioUmberto Capitianio More articles by this author , Francesco MontorsiFrancesco Montorsi More articles by this author , Viraj MasterViraj Master More articles by this author , and Ithaar DerweeshIthaar Derweesh More articles by this author View All Author Informationhttps://doi.org/10.1097/JU.0000000000002563.11AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: Pathological upstaging to T3a disease may occur following radical (RN) or partial nephrectomy (PN) for patients with T1/T2 renal cell carcinoma (RCC). Follow up and perceived prognosis for these patients is similar to those who initially presented with clinical T3 RCC. We sought to compare survival outcomes in patients who underwent surgical therapy for upstaged versus non-upstaged pT3 RCC. METHODS: We conducted a retrospective analysis of a multi-institutional dataset of patients who underwent radical (RN) or partial nephrectomy (PN) with final pathologic stage of pT3. Patients were classified as being upstaged (US) from cT1 or cT2 or non-upstaged (NUS) with cT3 disease. Primary outcome was cancer specific survival (CSS)/cancer specific mortality (CSM), with secondary outcomes being Overall Survival (OS)/all-cause mortality (ACM) and Progression/progression-free survival (PFS). Multivariable Cox regression analysis (MVA) was conducted for predictors of mortality outcomes and Kaplan Meier Analyses (KMA) were conducted for survival outcomes comparing US and NUS groups. RESULTS: 1232 patients were analyzed (853 US/379 NUS; median follow up 46.2 months). MVA for CSM found that clear cell RCC (HR 2.1, p <0.001), positive surgical margin (HR 1.7, p=0.047), and NUS (HR 1.5, p=0.003) were independently associated with worsened CSM. MVA for ACM revealed age (HR 1.01, p=0.002), RN (HR 2.5, p <0.001), preoperative GFR <60 (HR 1.4, p=0.007), clear-cell RCC (HR 1.4, p=0.007), positive surgical margin (HR 2.2, p <0.001), and NUS (1.27, p=0.023) were independently associated with worsened ACM. MVA for Progression revealed Age (HR 1.01, p=0.022), positive margin (HR 1.9, p=0.016), and NUS (HR 1.67, p <0.001) to be predictive. Comparing NUS and US groups, KMA revealed worsened 5 year CSS (56% vs. 75%, p <0.001; Figure), 5 year OS (42% vs. 58%, p <0.001) and 5 year PFS (33% vs. 58%, p <0.001). CONCLUSIONS: Pathologic upstaging of clinical T1/T2 to T3 RCC was associated with improved survival outcomes when compared to non-upstaged pathologic T3 RCC. Such differences need to be taken into account when designing follow up protocols and for refinement of risk stratification in the T3 RCC stage group. Further investigation is requisite. Source of Funding: Stephen Weissman Kidney Cancer Research Fund © 2022 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 207Issue Supplement 5May 2022Page: e396 Advertisement Copyright & Permissions© 2022 by American Urological Association Education and Research, Inc.MetricsAuthor Information Ryan Nasseri More articles by this author Kevin Hakimi More articles by this author Ava Saidian More articles by this author Riccardo Autorino More articles by this author Brian Lane More articles by this author Yasuha Fujii More articles by this author Michele Marchioni More articles by this author Hajime Tanaka More articles by this author Dattatraya Patil More articles by this author Francesco Porpiglia More articles by this author Sabrina Noyes More articles by this author Chandru Sundaram More articles by this author James Porter More articles by this author Andrea Minervini More articles by this author Umberto Capitianio More articles by this author Francesco Montorsi More articles by this author Viraj Master More articles by this author Ithaar Derweesh More articles by this author Expand All Advertisement PDF DownloadLoading ...

PD54-04 CHALLENGING THE DOGMA: SHOULD UNILATERAL PELVIC LYMPH NODE DISSECTION BE ENVISIONED IN THE MULTIPARAMETRIC PROSTATE MRI ERA?

PD54-04 CHALLENGING THE DOGMA: SHOULD UNILATERAL PELVIC LYMPH NODE DISSECTION BE ENVISIONED IN THE MULTIPARAMETRIC PROSTATE MRI ERA?

Deep learning of chest X-rays can predict mechanical ventilation outcome in ICU-admitted COVID-19 patients

The COVID-19 pandemic repeatedly overwhelms healthcare systems capacity and forced the development and implementation of triage guidelines in ICU for scarce resources (e.g. mechanical ventilation). These guidelines were often based on known risk factors for COVID-19. It is proposed that image data, specifically bedside computed X-ray (CXR), provide additional predictive information on mortality following mechanical ventilation that can be incorporated in the guidelines. Deep transfer learning was used to extract convolutional features from a systematically collected, multi-institutional dataset of COVID-19 ICU patients. A model predicting outcome of mechanical ventilation (remission or mortality) was trained on the extracted features and compared to a model based on known, aggregated risk factors. The model reached a 0.702 area under the curve (95% CI 0.707-0.694) at predicting mechanical ventilation outcome from pre-intubation CXRs, higher than the risk factor model. Combining imaging data and risk factors increased model performance to 0.743 AUC (95% CI 0.746-0.732). Additionally, a post-hoc analysis showed an increase performance on high-quality than low-quality CXRs, suggesting that using only high-quality images would result in an even stronger model.

Fecal Transplant for Recurrent Clostridium difficile Colitis in CF-LVAD Supported Patients

<h3>Purpose</h3> Clostridium difficile (CD) colitis may result in diarrhea with fluid shifts and hypovolemia as well as sepsis in patients supported on LVAD therapy. Treatment of CD with fecal transplant has utilized for patients with recurrent CD. Use of fecal transplant (FT) for treatment of CD in the LVAD population has not been described. <h3>Methods</h3> A retrospective study was conducted on LVAD patients implanted at a single center from 2010 to 2020 to identify those with CD that proceeded with FT. Patient characteristics, infection history, and subsequent medical course were collected and analyzed. The primary outcome was efficacy in a small cohort of LVAD-supported patients with CD. <h3>Results</h3> Patients identified (n=2) included an LVAD patient implanted as destination therapy (DT) and an LVAD patient who later underwent heart transplant. Prior to FT, both patients were trialed on courses of oral vancomycin for treatment of CD. Both patients developed recurrent CD despite oral vancomycin and were deemed to be failures of response. Patients underwent FT 379 and 1533 days after LVAD implant. FT occurred 301 days and 1749 days after initial CD diagnosis. All patients tolerated fecal transplant without complications. Following FT, the patient supported with DT LVAD did not have CD recurrence after last follow up of 481 days, and the patient initially supported with LVAD had one CD recurrence after 663 days (after heart transplant). <h3>Conclusion</h3> These data provide initial evidence of treatment safety and response to FT in patients supported on CF-LVAD therapy who have previously failed antibiotic therapy. FT may represent an effective therapy for treatment of CD in LVAD patients similar to the general population, but multi-institutional datasets should confirm efficacy.

Autosegmentation based on different-sized training datasets of consistently-curated volumes and impact on rectal contours in prostate cancer radiation therapy.

Background and purposeAutosegmentation techniques are emerging as time-saving means for radiation therapy (RT) contouring, but the understanding of their performance on different datasets is limited. The aim of this study was to determine agreement between rectal volumes by an existing autosegmentation algorithm and manually-delineated rectal volumes in prostate cancer RT. We also investigated contour quality by different-sized training datasets and consistently-curated volumes for retrained versions of this same algorithm.Materials and methodsSingle-institutional data from 624 prostate cancer patients treated to 50–70 Gy were used. Manually-delineated clinical rectal volumes (clinical) and consistently-curated volumes recontoured to one anatomical guideline (reference) were compared to autocontoured volumes by a commercial autosegmentation tool based on deep-learning (v1; n = 891, multiple-institutional data) and retrained versions using subsets of the curated volumes (v32/64/128/256; n = 32/64/128/256). Evaluations included dose-volume histogram metrics, Dice similarity coefficients, and Hausdorff distances; differences between groups were quantified using parametric or non-parametric hypothesis testing.ResultsVolumes by v1-256 (76–78 cm3) were larger than reference (75 cm3) and clinical (76 cm3). Mean doses by v1-256 (24.2–25.2 Gy) were closer to reference (24.2 Gy) than to clinical (23.8 Gy). Maximum doses were similar for all volumes (65.7–66.0 Gy). Dice for v1-256 and reference (0.87–0.89) were higher than for v1-256 and clinical (0.86–0.87) with corresponding Hausdorff comparisons including reference smaller than comparisons including clinical (5–6 mm vs. 7–8 mm).ConclusionUsing small single-institutional RT datasets with consistently-defined rectal volumes when training autosegmentation algorithms created contours of similar quality as the same algorithm trained on large multi-institutional datasets.

COVID-19 prognostic modeling using CT radiomic features and machine learning algorithms: Analysis of a multi-institutional dataset of 14,339 patients

BackgroundWe aimed to analyze the prognostic power of CT-based radiomics models using data of 14,339 COVID-19 patients. MethodsWhole lung segmentations were performed automatically using a deep learning-based model to extract 107 intensity and texture radiomics features. We used four feature selection algorithms and seven classifiers. We evaluated the models using ten different splitting and cross-validation strategies, including non-harmonized and ComBat-harmonized datasets. The sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) were reported. ResultsIn the test dataset (4,301) consisting of CT and/or RT-PCR positive cases, AUC, sensitivity, and specificity of 0.83 ± 0.01 (CI95%: 0.81–0.85), 0.81, and 0.72, respectively, were obtained by ANOVA feature selector + Random Forest (RF) classifier. Similar results were achieved in RT-PCR-only positive test sets (3,644). In ComBat harmonized dataset, Relief feature selector + RF classifier resulted in the highest performance of AUC, reaching 0.83 ± 0.01 (CI95%: 0.81–0.85), with a sensitivity and specificity of 0.77 and 0.74, respectively. ComBat harmonization did not depict statistically significant improvement compared to a non-harmonized dataset. In leave-one-center-out, the combination of ANOVA feature selector and RF classifier resulted in the highest performance. ConclusionLung CT radiomics features can be used for robust prognostic modeling of COVID-19. The predictive power of the proposed CT radiomics model is more reliable when using a large multicentric heterogeneous dataset, and may be used prospectively in clinical setting to manage COVID-19 patients.

Development and validation of a deep learning model for detection of breast cancers in mammography from multi-institutional datasets.

ObjectivesThe objective of this study was to develop and validate a state-of-the-art, deep learning (DL)-based model for detecting breast cancers on mammography.MethodsMammograms in a hospital development dataset, a hospital test dataset, and a clinic test dataset were retrospectively collected from January 2006 through December 2017 in Osaka City University Hospital and Medcity21 Clinic. The hospital development dataset and a publicly available digital database for screening mammography (DDSM) dataset were used to train and to validate the RetinaNet, one type of DL-based model, with five-fold cross-validation. The model’s sensitivity and mean false positive indications per image (mFPI) and partial area under the curve (AUC) with 1.0 mFPI for both test datasets were externally assessed with the test datasets.ResultsThe hospital development dataset, hospital test dataset, clinic test dataset, and DDSM development dataset included a total of 3179 images (1448 malignant images), 491 images (225 malignant images), 2821 images (37 malignant images), and 1457 malignant images, respectively. The proposed model detected all cancers with a 0.45–0.47 mFPI and had partial AUCs of 0.93 in both test datasets.ConclusionsThe DL-based model developed for this study was able to detect all breast cancers with a very low mFPI. Our DL-based model achieved the highest performance to date, which might lead to improved diagnosis for breast cancer.

Multiscale Local Enhancement Deep Convolutional Networks for the Automated 3D Segmentation of Gross Tumor Volumes in Nasopharyngeal Carcinoma: A Multi-Institutional Dataset Study.

PurposeAccurate segmentation of gross target volume (GTV) from computed tomography (CT) images is a prerequisite in radiotherapy for nasopharyngeal carcinoma (NPC). However, this task is very challenging due to the low contrast at the boundary of the tumor and the great variety of sizes and morphologies of tumors between different stages. Meanwhile, the data source also seriously affect the results of segmentation. In this paper, we propose a novel three-dimensional (3D) automatic segmentation algorithm that adopts cascaded multiscale local enhancement of convolutional neural networks (CNNs) and conduct experiments on multi-institutional datasets to address the above problems.Materials and MethodsIn this study, we retrospectively collected CT images of 257 NPC patients to test the performance of the proposed automatic segmentation model, and conducted experiments on two additional multi-institutional datasets. Our novel segmentation framework consists of three parts. First, the segmentation framework is based on a 3D Res-UNet backbone model that has excellent segmentation performance. Then, we adopt a multiscale dilated convolution block to enhance the receptive field and focus on the target area and boundary for segmentation improvement. Finally, a central localization cascade model for local enhancement is designed to concentrate on the GTV region for fine segmentation to improve the robustness. The Dice similarity coefficient (DSC), positive predictive value (PPV), sensitivity (SEN), average symmetric surface distance (ASSD) and 95% Hausdorff distance (HD95) are utilized as qualitative evaluation criteria to estimate the performance of our automated segmentation algorithm.ResultsThe experimental results show that compared with other state-of-the-art methods, our modified version 3D Res-UNet backbone has excellent performance and achieves the best results in terms of the quantitative metrics DSC, PPR, ASSD and HD95, which reached 74.49 ± 7.81%, 79.97 ± 13.90%, 1.49 ± 0.65 mm and 5.06 ± 3.30 mm, respectively. It should be noted that the receptive field enhancement mechanism and cascade architecture can have a great impact on the stable output of automatic segmentation results with high accuracy, which is critical for an algorithm. The final DSC, SEN, ASSD and HD95 values can be increased to 76.23 ± 6.45%, 79.14 ± 12.48%, 1.39 ± 5.44mm, 4.72 ± 3.04mm. In addition, the outcomes of multi-institution experiments demonstrate that our model is robust and generalizable and can achieve good performance through transfer learning.ConclusionsThe proposed algorithm could accurately segment NPC in CT images from multi-institutional datasets and thereby may improve and facilitate clinical applications.

Federated Learning for Multicenter Collaboration in Ophthalmology: Improving Classification Performance in Retinopathy of Prematurity

To compare the performance of deep learning classifiers for the diagnosis of plus disease in retinopathy of prematurity (ROP) trained using 2 methods for developing models on multi-institutional data sets: centralizing data versus federated learning (FL) in which no data leave each institution. Evaluation of a diagnostic test or technology. Deep learning models were trained, validated, and tested on 5255 wide-angle retinal images in the neonatal intensive care units of 7 institutions as part of the Imaging and Informatics in ROP study. All images were labeled for the presence of plus, preplus, or no plus disease with a clinical label and a reference standard diagnosis (RSD) determined by 3 image-based ROP graders and the clinical diagnosis. We compared the area under the receiver operating characteristic curve (AUROC) for models developed on multi-institutional data, using a central approach initially, followed by FL, and compared locally trained models with both approaches. We compared the model performance (κ) with the label agreement (between clinical and RSD), data set size, and number of plus disease cases in each training cohort using the Spearman correlation coefficient (CC). Model performance using AUROC and linearly weighted κ. Four settings of experiment were used: FL trained on RSD against central trained on RSD, FL trained on clinical labels against central trained on clinical labels, FL trained on RSD against central trained on clinical labels, and FL trained on clinical labels against central trained on RSD (P= 0.046, P= 0.126, P= 0.224, and P= 0.0173, respectively). Four of the 7 (57%) models trained on local institutional data performed inferiorly to the FL models. The model performance for local models was positively correlated with the label agreement (between clinical and RSD labels, CC= 0.389, P= 0.387), total number of plus cases (CC= 0.759, P= 0.047), and overall training set size (CC= 0.924, P= 0.002). We found that a trained FL model performs comparably to a centralized model, confirming that FL may provide an effective, more feasible solution for interinstitutional learning. Smaller institutions benefit more from collaboration than larger institutions, showing the potential of FL for addressing disparities in resource access.

The Role of Standards in Accelerating the Uptake of Artificial Intelligence in Dermatology

Background The use of artificial intelligence (AI) for dermatology is showing great promise in research contexts. However, the clinical use of AI in dermatology is still limited. The uptake of medical imaging standards for dermatology imaging is also limited. Standards adoption is more widespread in other imaging specialties (eg, radiology) as is the clinical use of AI. Digital Image Communication in Medicine (DICOM) is the standard for medical imaging. DICOM standardizes image formats and associated metadata. Further, DICOM facilitates interoperability between actors in the digital health ecosystem. Objective This study aimed to identify how medical imaging standards, in particular DICOM, can support the clinical use of AI. Methods Design Science Research Methodology was used to determine the role of DICOM in AI-based medical imaging workflows. Scenarios were identified and synthesized using expert consensus. Results The key benefits of using DICOM to improve the clinical use of AI were the potential to encode artefacts derived from the AI process as DICOM objects and store them alongside the original images. Such objects include downsized or down-sampled images, segmentation objects, or visual explainability maps (eg, class activation maps). DICOM can facilitate interoperability between actors in the medical imaging workflow pipeline and permits the inclusion of AI evidence creators in this pipeline. Owing to standardized image formats and metadata, DICOM can be beneficial for the curation of multi-institutional data sets. The key challenge of using DICOM is limited uptake in some specialties including dermatology. Conclusions DICOM offers potential to accelerate the clinical adoption of AI in dermatology by addressing several technological issues. More widespread uptake of DICOM in dermatology imaging is required to achieve this potential. Conflicts of Interest None declared.

Systemic Chemotherapies Retain Antitumor Activity in Desmoid Tumors Independent of Specific Mutations in CTNNB1 or APC: A Multi-institutional Retrospective Study.

Determine whether specific CTNNB1 or APC mutations in patients with desmoid tumor were associated with differences in clinical responses to systemic treatments. We established a multi-institutional dataset of previously treated patients with desmoid tumor across four U.S. sarcoma centers, including demographic and clinicopathologic characteristics, treatment regimens, and clinical and radiographic responses. CTNNB1 or APC mutation status was determined from prior pathology records, or archival tissue was requested and analyzed by Sanger sequencing and/or next-generation sequencing. Evaluable patients with mutation results were analyzed to determine clinical progression-free survival (cPFS), RECIST 1.1 PFS (rPFS), time to next treatment (TTNT), and overall survival (OS). Kaplan-Meier analysis and Cox proportional hazards regression were performed to identify differences in cPFS, rPFS, TTNT, and OS by mutation subtype, desmoid tumor location, and treatment regimen. A total of 259 evaluable patients were analyzed for at least one of the survival outcomes, with 177 patients having mutation data. First- and second-line cPFS, rPFS, and TTNT were not significantly affected by mutation subtype; however, APC-mutant desmoid tumors demonstrated nonstatistically significant inferior outcomes. Extremity/trunk desmoid tumor location and treatment with doxorubicin-based, methotrexate/vinca alkaloids and sorafenib regimens were associated with better clinical outcomes compared with surgery or "other" therapies, including estrogen-receptor blockade and imatinib. OS was significantly worse with APC or CTNNB1 negative/other mutations. Mutation subtype did not affect responses to specific systemic therapies. APC mutations and nonextremity desmoid tumor locations remain prognostic for worse outcomes, and earlier initiation of systemic therapy for these higher-risk desmoid tumors should be prospectively evaluated. See related commentary by Greene and Van Tine, p. 3911.

Abstract PD14-04: Circulating tumor DNA characterization of invasive lobular carcinoma in patients with metastatic breast cancer

Abstract Background:Invasive lobular carcinoma (ILC) is the second most common histology of breast cancer, accounting for approximately 10-15% of cases. Prior studies have demonstrated that loss of E-cadherin, as well as alterations in tissue including CDH1, FOXA1, TBX3 and PTEN loss, that were more commonly observed in Luminal A ILC, while GATA3 was more commonly observed in invasive ductal carcinoma (IDC) (Ciriello et al., Cell 2015). However, data regarding the characterization of circulating tumor DNA (ctDNA) in patients (pts) with metastatic ILC are limited. We hypothesized that there would be distinct mutational profiles between pts with metastatic ILC and IDC that could be characterized using ctDNA. Methods:This retrospective cohort study included de-identified clinical, pathological, and ctDNA data from pts with metastatic breast cancer (MBC) combined under a data use agreement and approved by the institutional review boards of three sites: Washington University in St. Louis (MO), Northwestern University (Chicago, IL), and Massachusetts General Hospital (Boston, MA). All pts included in the study had ctDNA testing per standard of care with plasma-based genotyping performed by Guardant360 (Redwood City, CA) between 2015-2020. Histological classification (ILC vs. IDC) was defined based on review of pathology reports from the primary tumor or from breast biopsies of de novo MBC, and additional clinical and pathological variables were obtained via electronic medical record review. Single nucleotide variants (SNVs) were annotated using OncoKB and ClinVar and only pathogenic variants were included. Mutational profiles were compared across histologic subtypes using Fisher’s exact test to assess differences in alteration frequency across subtypes. Multivariable analysis was performed. Results:A total of 994 pts with MBC underwent ctDNA testing and were included in the analysis. 10.7% of pts had ILC (N=106) and 89.3% had IDC (N=888). 89.4% of ILC cases were categorized as hormone-receptor positive (HR+) compared with 67.1% of IDC cases. Pts with ILC had a lower frequency of triple-negative (6.7% vs. 17.7%) and HER2 positive (3.9% vs. 15.2%) breast cancer compared with IDC. Pts with ILC had a significantly higher number of pathogenic SNVs compared with IDC (mean 4.45 vs. 2.77; P=0.0037). In contrast, pts with ILC had a significantly lower number of copy number alterations as compared to pts with IDC (mean 0.40 vs. 1.03; P=0.0017). No differences were observed in mutant allele frequency between pts with ILC and IDC. The 5 most common alterations observed in pts with ILC were the following: PIK3CA, TP53, ESR1, ERBB2, and ARID1A. Alterations in AR, BRAF, CDH1, ERBB2, FGFR2, IDH2, KRAS, NF1, PIK3CA, SMAD4, and TERT were significantly higher in ILC than IDC (all P&amp;lt;0.05). In contrast, mutations in GATA3, and amplifications in ERBB2 and MYC were significantly more common in pts with IDC (all P&amp;lt;0.05). In multivariable analysis, mutations in BRAF, CDH1, ERBB2, IDH2, TERT remained significantly higher in ILC, while amplification of MYC was significantly higher in IDC (all P&amp;lt;0.05). After restricting the analysis to pts with HR+ HER2 negative MBC, the following genes were significant in multivariate analysis: CDH1 and ERBB2 for pts with ILC and MYC amplification for pts with IDC (all P&amp;lt;0.05). Discussion:In this large, multi-institutional dataset, pts with metastatic ILC were characterized by a significantly higher number of SNVs in ctDNA compared to pts with IDC, suggesting higher mutational burden. We report several alterations that were significantly different in ILC vs. IDC. These results demonstrate the ctDNA profile of pts with ILC, and future studies should explore serial plasma-based genotyping to track ILC evolution to develop targeted precision medicine based therapeutic approaches for this unique subset of pts with MBC. Citation Format: Andrew A Davis, Lorenzo Gerratana, Katherine Clifton, Marko Velimirovic, Whitney L Hensing, Ami N Shah, Paolo D’Amico, Carolina Reduzzi, Qiang Zhang, Charles S Dai, Elyssa N Denault, Nusayba A Bagegni, Mateusz Opyrchal, Foluso O Ademuyiwa, Ron Bose, William J Gradishar, Amir Behdad, Cynthia X Ma, Aditya Bardia, Massimo Cristofanilli. Circulating tumor DNA characterization of invasive lobular carcinoma in patients with metastatic breast cancer [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr PD14-04.

General and custom deep learning autosegmentation models for organs in head and neck, abdomen, and male pelvis.

To reduce workload and inconsistencies in organ segmentation for radiation treatment planning, we developed and evaluated general and custom autosegmentation models on computed tomography (CT) for three major tumor sites using a well-established deep convolutional neural network (DCNN). Five CT-based autosegmentation models for 42 organs at risk (OARs) in head and neck (HN), abdomen (ABD), and male pelvis (MP) were developed using a full three-dimensional (3D) DCNN architecture. Two types of deep learning (DL) models were separately trained using either general diversified multi-institutional datasets or custom well-controlled single-institution datasets. To improve segmentation accuracy, an adaptive spatial resolution approach for small and/or narrow OARs and a pseudo scan extension approach, when CT scan length is too short to cover entire organs, were implemented. The performance of the obtained models was evaluated based on accuracy and clinical applicability of the autosegmented contours using qualitative visual inspection and quantitative calculation of dice similarity coefficient (DSC), mean distance to agreement (MDA), and time efficiency. The five DL autosegmentation models developed for the three anatomical sites were found to have high accuracy (DSC ranging from 0.8 to 0.98) for 74% OARs and marginally acceptable for 26% OARs. The custom models performed slightly better than the general models, even with smaller custom datasets used for the custom model training. The organ-based approaches improved autosegmentation accuracy for small or complex organs (e.g., eye lens, optic nerves, inner ears, and bowels). Compared with traditional manual contouring times, the autosegmentation times, including subsequent manual editing, if necessary, were substantially reduced by 88% for MP, 80% for HN, and 65% for ABD models. The obtained autosegmentation models, incorporating organ-based approaches, were found to be effective and accurate for most OARs in the male pelvis, head and neck, and abdomen. We have demonstrated that our multianatomical DL autosegmentation models are clinically useful for radiation treatment planning.

Locoregional Recurrence Prediction Using a Deep Neural Network of Radiological and Radiotherapy Images.

Radiation therapy (RT) is an important and potentially curative modality for head and neck squamous cell carcinoma (HNSCC). Locoregional recurrence (LR) of HNSCC after RT is ranging from 15% to 50% depending on the primary site and stage. In addition, the 5-year survival rate of patients with LR is low. To classify high-risk patients who might develop LR, a deep learning model for predicting LR needs to be established. In this work, 157 patients with HNSCC who underwent RT were analyzed. Based on the National Cancer Institute’s multi-institutional TCIA data set containing FDG-PET/CT/dose, a 3D deep learning model was proposed to predict LR without time-consuming segmentation or feature extraction. Our model achieved an averaged area under the curve (AUC) of 0.856. Adding clinical factors into the model improved the AUC to an average of 0.892 with the highest AUC of up to 0.974. The 3D deep learning model could perform individualized risk quantification of LR in patients with HNSCC without time-consuming tumor segmentation.

Clinical cure vs a novel trifecta system for evaluating long-term outcomes of minimally-invasive partial or total adrenalectomy for unilateral primary aldosteronism: results of a multicentric series.

Several predictive scores to evaluate outcomes of adrenal surgery for unilateral primary aldosteronism (UPA), have been conceived. We compared a novel trifecta that summarizes outcomes of adrenal surgery for UPA with the clinical cure proposed by Vorselaars. Between March 2011 and January 2022, a multi-institutional dataset was queried for UPA. Baseline, perioperative and functional data were collected. Clinical and biochemical complete and partial success rates according to Primary Aldosteronism Surgical Outcome (PASO) criteria were assessed for the overall cohort. Clinical cure was defined either as normotension without antihypertensive medications or normotension with lower or equal use of antihypertensive medications. Trifecta was defined as the coexistence of ≥50% antihypertensive therapeutic intensity score (TIS) reduction (ΔTIS), no electrolyte impairment at 3-months and no Clavien-Dindo (2-5) complications. Cox regression analyses were used to identify predictors of long-term clinical and biochemical success. For all analyses, a two-sided p <0.05 was considered significant. Baseline, perioperative and functional outcomes were analyzed. Out of 90 patients, at a median follow-up of 42 months (IQR 27-54) a complete and partial clinical success was observed in 60% and 17.7% of cases while a complete and partial biochemical success was achieved in 83.3% and 12.3% of cases, respectively. Overall trifecta and clinical cure rates were 21.1% and 58.9%, respectively. On multivariable Cox regression analysis, trifecta achievement (HR 2.87; 95% CI 1.45-5.58; p = 0.02) was the only independent predictor of complete clinical success at long-term follow-up. Despite its complex estimation and more restrictive criteria, trifecta but not clinical cure allows to independently predict composite PASO endpoints on the long run.

Cigarette smoking and complications in elective thoracolumbar fusions surgery: An analysis of 58,304 procedures.

ABSTRACTStudy Design:This was retrospective cohort study.Purpose:The current investigation uses a large, multi-institutional dataset to compare short-term morbidity and mortality rates between current smokers and nonsmokers undergoing thoracolumbar fusion surgery.Overview of Literature:The few studies that have addressed perioperative complications following thoracolumbar fusion surgeries are each derived from small cohorts from single institutions.Materials and Methods:A retrospective study was conducted on thoracolumbar fusion patients in the American College of Surgeons National Surgical Quality Improvement Program database (2006–2016). The primary outcome compared the rates of overall morbidity, severe postoperative morbidity, infections, pneumonia, deep venous thrombosis (DVT), pulmonary embolism (PE), transfusions, and mortality in smokers and nonsmokers.Results:A total of 57,677 patients were identified. 45,952 (78.8%) were nonsmokers and 12,352 (21.2%) smoked within 1 year of surgery. Smokers had fewer severe complications (1.6% vs. 2.0%, P = 0.014) and decreased discharge to skilled nursing facilities (6.3% vs. 11.5%, P < 0.001) compared to nonsmokers. They had lower incidences of transfusions (odds ratio [OR] = 0.9, confidence interval [CI] = 0.8–1.0, P = 0.009) and DVT (OR = 0.7, CI = 0.5–0.9, P = 0.039) as well as shorter length of stay (LOS) (OR = 0.9, CI = 0.9–0.99, P < 0.001). They had a higher incidence of postoperative pneumonia (OR = 1.4, CI = 1.1–1.8, P = 0.002). There was no difference in the remaining primary outcomes between smoking and nonsmoking cohorts.Conclusions:There is a positive correlation between smoking and postoperative pneumonia after thoracolumbar fusion. The incidence of blood transfusions, DVT, and LOS was decreased in smokers. Early postoperative mortality, severe complications, discharge to subacute rehabilitation facilities, extubation failure, PE, SSI, and return to OR were not associated with smoking.

FedSPL: federated self-paced learning for privacy-preserving disease diagnosis.

The growing expansion of data availability in medical fields could help improve the performance of machine learning methods. However, with healthcare data, using multi-institutional datasets is challenging due to privacy and security concerns. Therefore, privacy-preserving machine learning methods are required. Thus, we use a federated learning model to train a shared global model, which is a central server that does not contain private data, and all clients maintain the sensitive data in their own institutions. The scattered training data are connected to improve model performance, while preserving data privacy. However, in the federated training procedure, data errors or noise can reduce learning performance. Therefore, we introduce the self-paced learning, which can effectively select high-confidence samples and drop high noisy samples to improve the performances of the training model and reduce the risk of data privacy leakage. We propose the federated self-paced learning (FedSPL), which combines the advantage of federated learning and self-paced learning. The proposed FedSPL model was evaluated on gene expression data distributed across different institutions where the privacy concerns must be considered. The results demonstrate that the proposed FedSPL model is secure, i.e. it does not expose the original record to other parties, and the computational overhead during training is acceptable. Compared with learning methods based on the local data of all parties, the proposed model can significantly improve the predicted F1-score by approximately 4.3%. We believe that the proposed method has the potential to benefit clinicians in gene selections and disease prognosis.

Quantum deep reinforcement learning for clinical decision support in oncology: application to adaptive radiotherapy

Subtle differences in a patient’s genetics and physiology may alter radiotherapy (RT) treatment responses, motivating the need for a more personalized treatment plan. Accordingly, we have developed a novel quantum deep reinforcement learning (qDRL) framework for clinical decision support that can estimate an individual patient’s dose response mid-treatment and recommend an optimal dose adjustment. Our framework considers patients’ specific information including biological, physical, genetic, clinical, and dosimetric factors. Recognizing that physicians must make decisions amidst uncertainty in RT treatment outcomes, we employed indeterministic quantum states to represent human decision making in a real-life scenario. We paired quantum decision states with a model-based deep q-learning algorithm to optimize the clinical decision-making process in RT. We trained our proposed qDRL framework on an institutional dataset of 67 stage III non-small cell lung cancer (NSCLC) patients treated on prospective adaptive protocols and independently validated our framework in an external multi-institutional dataset of 174 NSCLC patients. For a comprehensive evaluation, we compared three frameworks: DRL, qDRL trained in a Qiskit quantum computing simulator, and qDRL trained in an IBM quantum computer. Two metrics were considered to evaluate our framework: (1) similarity score, defined as the root mean square error between retrospective clinical decisions and the AI recommendations, and (2) self-evaluation scheme that compares retrospective clinical decisions and AI recommendations based on the improvement in the observed clinical outcomes. Our analysis shows that our framework, which takes into consideration individual patient dose response in its decision-making, can potentially improve clinical RT decision-making by at least about 10% compared to unaided clinical practice. Further validation of our novel quantitative approach in a prospective study will provide a necessary framework for improving the standard of care in personalized RT.

Federated learning for computational pathology on gigapixel whole slide images.

Deep Learning-based computational pathology algorithms have demonstrated profound ability to excel in a wide array of tasks that range from characterization of well known morphological phenotypes to predicting non human-identifiable features from histology such as molecular alterations. However, the development of robust, adaptable and accurate deep learning-based models often rely on the collection and time-costly curation large high-quality annotated training data that should ideally come from diverse sources and patient populations to cater for the heterogeneity that exists in such datasets. Multi-centric and collaborative integration of medical data across multiple institutions can naturally help overcome this challenge and boost the model performance but is limited by privacy concerns among other difficulties that may arise in the complex data sharing process as models scale towards using hundreds of thousands of gigapixel whole slide images. In this paper, we introduce privacy-preserving federated learning for gigapixel whole slide images in computational pathology using weakly-supervised attention multiple instance learning and differential privacy. We evaluated our approach on two different diagnostic problems using thousands of histology whole slide images with only slide-level labels. Additionally, we present a weakly-supervised learning framework for survival prediction and patient stratification from whole slide images and demonstrate its effectiveness in a federated setting. Our results show that using federated learning, we can effectively develop accurate weakly-supervised deep learning models from distributed data silos without direct data sharing and its associated complexities, while also preserving differential privacy using randomized noise generation. We also make available an easy-to-use federated learning for computational pathology software package: http://github.com/mahmoodlab/HistoFL.