scholarly journals Deep learning with multimodal representation for pancancer prognosis prediction

2019 ◽  
Vol 35 (14) ◽  
pp. i446-i454 ◽  
Author(s):  
Anika Cheerla ◽  
Olivier Gevaert
Keyword(s):  
Cancer Patients ◽  
Expression Data ◽  
Patient Handling ◽  
Multimodal Data ◽  
Patients With Cancer ◽  
Multiple Data ◽  
Single Feature ◽  
Clinical Methods ◽  
Cancer Types ◽  
Single Cancer

Abstract Motivation Estimating the future course of patients with cancer lesions is invaluable to physicians; however, current clinical methods fail to effectively use the vast amount of multimodal data that is available for cancer patients. To tackle this problem, we constructed a multimodal neural network-based model to predict the survival of patients for 20 different cancer types using clinical data, mRNA expression data, microRNA expression data and histopathology whole slide images (WSIs). We developed an unsupervised encoder to compress these four data modalities into a single feature vector for each patient, handling missing data through a resilient, multimodal dropout method. Encoding methods were tailored to each data type—using deep highway networks to extract features from clinical and genomic data, and convolutional neural networks to extract features from WSIs. Results We used pancancer data to train these feature encodings and predict single cancer and pancancer overall survival, achieving a C-index of 0.78 overall. This work shows that it is possible to build a pancancer model for prognosis that also predicts prognosis in single cancer sites. Furthermore, our model handles multiple data modalities, efficiently analyzes WSIs and represents patient multimodal data flexibly into an unsupervised, informative representation. We thus present a powerful automated tool to accurately determine prognosis, a key step towards personalized treatment for cancer patients. Availability and implementation https://github.com/gevaertlab/MultimodalPrognosis

scholarly journals Deep Learning with Multimodal Representation for Pancancer Prognosis Prediction

2019 ◽  
Author(s):  
Anika Cheerla ◽  
Olivier Gevaert
Keyword(s):  
Cancer Patients ◽  
Clinical Data ◽  
Survival Data ◽  
Cancer Survival ◽  
Patient Handling ◽  
Gene Expressions ◽  
Multiple Data ◽  
Single Feature ◽  
Single Cancer ◽  
Cancer Survival Data

AbstractEstimating the future course of cancer is invaluable to physicians; however, current clinical methods fail to effectively use the vast amount of multimodal data that is available for cancer patients.To tackle this problem, we constructed a deep neural network based model to predict the survival of patients for 20 different cancer types using gene expressions, microRNA data, clinical data and histopathology whole slide images (WSIs). We developed an unsupervised encoder to compress these four data modalities into a single feature vector for each patient, handling missing data through a resilient, multimodal dropout method. Encoding methods were tailored to each data type - using deep highway networks to extract features from genomic and clinical data, and convolutional neural networks extract features from pathology images. We then used these feature encodings trained on pancancer data to predict pancancer and single cancer survival data, achieving a C-index of 0.784 overall.This work shows that it is possible to build a pancancer model for prognosis that also predicts prognosis in single cancer sites. Furthermore, our model handles multiple data modalities, efficiently analyzes WSIs, and summarizes patient details flexibly into an unsupervised, informative profile. We thus present a powerful automated tool to accurately determine prognosis, a key step towards personalized treatment for cancer patients.

scholarly journals Use of Anti-angiogenic Drugs Potentially Associated With an Increase on Serum AST, LDH, CK, and CK-MB Activities in Patients With Cancer: A Retrospective Study

2021 ◽  
Vol 8 ◽  
Author(s):  
Qi Zheng ◽  
Hanzhou Wang ◽  
Wei Hou ◽  
Ying Zhang
Keyword(s):  
Creatine Kinase ◽  
Cancer Patients ◽  
Muscle Injury ◽  
Lactic Dehydrogenase ◽  
Cardiovascular Risks ◽  
Patients With Cancer ◽  
Specific Effects ◽  
Before And After ◽  
Cancer Types ◽  
Myocardial Muscle

Background: There is a large amount of evidence that anti-angiogenic drugs are effective safe. However, few studies have evaluated the specific effects of anti-angiogenic drugs on myocardial enzyme injury biomarkers: aspartate aminotransferase (AST), lactic dehydrogenase (LDH), creatine kinase (CK) and creatine kinase isoenzyme (CK-MB). The purpose of our study was to determine whether anti-angiogenic drugs serum AST, LDH, CK, and CK-MB activities of cancer patients treated with anti-angiogenic drugs.Methods: This study retrospectively analyzed 81 cancer patients. Patients who had used anti-angiogenic drugs were selected. Serum AST, LDH, CK, and CK-MB activities were measured before and after treatment with anti-angiogenic drugs for 3 weeks.Results: A total of 16 cancer types were analyzed. The distribution of the cancer types in the patients was mainly concentrated in lung, gastric, and colorectal cancers. The anti-angiogenic treatment markedly increased AST, LDH, CK, and CK-MB activities by 32.51, 7.29, 31.25, and 55.56%, respectively in serum.Conclusions: Our findings suggest that patients, who had used anti-angiogenic drugs were likely to have elevated AST, LDH, and CK, indicators of myocardial muscle injury. Use of anti-angiogenic drugs should not be assumed to be completely safe and without any cardiovascular risks.

scholarly journals BBCancer: an expression atlas of blood-based biomarkers in the early diagnosis of cancers

2019 ◽  
Author(s):  
Zhixiang Zuo ◽  
Huanjing Hu ◽  
Qingxian Xu ◽  
Xiaotong Luo ◽  
Di Peng ◽  
...  
Keyword(s):  
Gene Expression ◽  
Early Detection ◽  
Cancer Patients ◽  
High Throughput ◽  
Early Detection Of Cancer ◽  
Circular Rnas ◽  
Expression Data ◽  
Blood Samples ◽  
Cancer Types ◽  
High Throughput Gene Expression

Abstract The early detection of cancer holds the key to combat and control the increasing global burden of cancer morbidity and mortality. Blood-based screenings using circulating DNAs (ctDNAs), circulating RNA (ctRNAs), circulating tumor cells (CTCs) and extracellular vesicles (EVs) have shown promising prospects in the early detection of cancer. Recent high-throughput gene expression profiling of blood samples from cancer patients has provided a valuable resource for developing new biomarkers for the early detection of cancer. However, a well-organized online repository for these blood-based high-throughput gene expression data is still not available. Here, we present BBCancer (http://bbcancer.renlab.org/), a web-accessible and comprehensive open resource for providing the expression landscape of six types of RNAs, including messenger RNAs (mRNAs), long noncoding RNAs (lncRNAs), microRNAs (miRNAs), circular RNAs (circRNAs), tRNA-derived fragments (tRFRNAs) and Piwi-interacting RNAs (piRNAs) in blood samples, including plasma, CTCs and EVs, from cancer patients with various cancer types. Currently, BBCancer contains expression data of the six RNA types from 5040 normal and tumor blood samples across 15 cancer types. We believe this database will serve as a powerful platform for developing blood biomarkers.

Abstract 13145: Differences of Cancer Types in Hospital Mortality in Patients With Venous Thromboembolism

Circulation ◽  
2020 ◽  
Vol 142 (Suppl_3) ◽  
Author(s):  
Okushi Yuichiro ◽  
Kenya Kusunose ◽  
Takayuki Ise ◽  
Takeshi Tobiume ◽  
Koji Yamaguchi ◽  
...  
Keyword(s):  
Big Data ◽  
Venous Thromboembolism ◽  
Hospital Mortality ◽  
Cancer Patients ◽  
Biliary Tract ◽  
Gynecological Cancer ◽  
Cancer Type ◽  
Patients With Cancer ◽  
Liver Cancers ◽  
Cancer Types

Introduction: We sought to evaluate the clinical characteristics and outcomes of patients with cancer-associated VTE, compared with the matched cohort without cancer using real-world big data of VTE. Background: Cancer is associated with a high incidence of Venous Thromboembolism (VTE) and there are many guidelines/recommendations about VTE. However, the prognosis of cancer-VTE patients is not well known because of a lack of big data. Moreover, there is also no knowledge on how cancer type is related to prognosis. Methods: This study was based on the Diagnosis Procedure Combination database in the Japanese Registry of All Cardiac and Vascular Datasets (JROAD-DPC). We identified 28,247 patients who were first hospitalized with VTE from April 2012 to March 2017. 26.0% were cancer patients. Compared with national statistics of cancer incidence in 2015 from National Cancer Center of Japan, the proportion of gynecological cancer patients was higher, but other cancer types had similar prevalence rates. Propensity score (PS) was estimated with logistic regression model, with cancer as the dependent variable and 18 clinically relevant covariates. Results: We included 24,576 patients after exclusion. The median age was 71years (range: 59-80 years), and 42.0% were male. On PS-matched analysis with 12,418 patients, patients with cancer had higher total in-hospital mortality (9.5% vs. 3.8%, P<0.001; OR, 2.72, 95% CI: 2.33-3.19) and in-hospital mortality within 30days (6.8% vs. 3.2%, P<0.001; OR, 2.20, 95% CI: 1.85-2.62). On analysis for each type of cancer, in-hospital mortality in 10 types of cancer was significantly high, especially pancreas (OR: 9.65, 95%CI: 4.31-21.64), biliary tract (OR: 8.36, 95%CI: 2.42-28.89) and liver (OR: 7.33, 95%CI: 1.92-28.02). Conclusions: Patients with cancer had a higher in-hospital mortality for VTE than those without cancer, especially in pancreatic, biliary tract and liver cancers.

Suicide Risk in Cancer Patients From 1960 to 1999

2004 ◽  
Vol 22 (20) ◽  
pp. 4209-4216 ◽  
Author(s):  
Erlend Hem ◽  
Jon H. Loge ◽  
Tor Haldorsen ◽  
Øivind Ekeberg
Keyword(s):  
Relative Risk ◽  
Cancer Patients ◽  
Cancer Registry ◽  
Suicide Risk ◽  
International Classification Of Diseases ◽  
Patients With Cancer ◽  
Classification Of Diseases ◽  
Cancer Types ◽  
Male Patients ◽  
Standardized Mortality Ratios

Purpose Suicide risk is reportedly higher for cancer patients than for the general population, but estimates vary and analyses of trends are few. The aim of the present study was to determine whether cancer patients had a higher suicide risk between 1960 and 1999. Patients and Methods A cohort comprising patients from the Cancer Registry of Norway 1960 to 1997 was linked to suicide diagnosis in the Register of Deaths at Statistics Norway and observed during 1960 to 1999. The cohort consisted of all cancer patients registered in the Cancer Registry of Norway 1960 to 1997 (N = 490,245 patients with 520,823 cancer diagnoses). Suicide was defined according to death certificates based on the International Classification of Diseases (versions 7, 8, 9, and 10). Results During the period, 589 cancer patients (407 males and 182 females) committed suicide. The relative risk was elevated for males and females, with standardized mortality ratios (SMRs) of 1.55 (95% CI, 1.41 to 1.71) and 1.35 (95% CI, 1.17 to 1.56), respectively. Risk was highest in the first months after diagnosis. For both sexes, there was a significant decrease in the relative suicide risk over decades. The risk was markedly increased among male patients with cancer of respiratory organs (SMR, 4.08; 95% CI, 2.96 to 5.47). Otherwise, the SMRs varied from 0.76 to 3.67 across cancer types. Conclusion Cancer may be a risk factor for suicide, particularly shortly after diagnosis. However, the relative risk gradually decreased during the period 1960 to 1999.

scholarly journals Comparison and impact of COVID-19 for patients with cancer: a survival analysis of fatality rate controlling for age, sex and cancer type

2021 ◽  
Vol 28 (1) ◽  
pp. e100341
Author(s):  
Haiquan Li ◽  
Edwin Baldwin ◽  
Xiang Zhang ◽  
Colleen Kenost ◽  
Wenting Luo ◽  
...  
Keyword(s):  
Cancer Patients ◽  
Uterine Cancer ◽  
Fold Increase ◽  
Distant Metastases ◽  
Cancer Type ◽  
Fatality Rate ◽  
Patients With Cancer ◽  
Increased Risk ◽  
Cancer Types ◽  
Stratified Analysis

ObjectivesPrior research has reported an increased risk of fatality for patients with cancer, but most studies investigated the risk by comparing cancer to non-cancer patients among COVID-19 infections, where cancer might have contributed to the increased risk. This study is to understand COVID-19’s imposed HR of fatality while controlling for covariates, such as age, sex, metastasis status and cancer type.MethodsWe conducted survival analyses of 4606 cancer patients with COVID-19 test results from 16 March to 11 October 2020 in UK Biobank and estimated the overall HR of fatality with and without COVID-19 infection. We also examined the HRs of 13 specific cancer types with at least 100 patients using a stratified analysis.ResultsCOVID-19 resulted in an overall HR of 7.76 (95% CI 5.78 to 10.40, p<10−10) by following 4606 patients with cancer for 21 days after the tests. The HR varied among cancer type, with over a 10-fold increase in fatality rate (false discovery rate ≤0.02) for melanoma, haematological malignancies, uterine cancer and kidney cancer. Although COVID-19 imposed a higher risk for localised versus distant metastasis cancers, those of distant metastases yielded higher overall fatality rates due to their multiplicative effects.DiscussionThe results confirmed prior reports for the increased risk of fatality for patients with COVID-19 plus hematological malignancies and demonstrated similar findings of COVID-19 on melanoma, uterine, and kidney cancers.ConclusionThe results highlight the heightened risk that COVID-19 imposes on localised and haematological cancer patients and the necessity to vaccinate uninfected patients with cancer promptly, particularly for the cancer types most influenced by COVID-19. Results also suggest the importance of timely care for patients with localised cancer, whether they are infected by COVID-19 or not.

Frequency, demographics and risk (according to tumour type or site) of cancer-associated thrombosis among patients seen at outpatient DVT clinics

2010 ◽  
Vol 103 (02) ◽  
pp. 338-343 ◽  
Author(s):  
Shankaranarayana Paneesha ◽  
Aidan McManus ◽  
Roopen Arya ◽  
Nicholas Scriven ◽  
Timothy Farren ◽  
...  
Keyword(s):  
Cancer Patients ◽  
Similar Rate ◽  
Cancer Type ◽  
Tumour Site ◽  
Vein Thrombosis ◽  
Patients With Cancer ◽  
Skin Cancers ◽  
Cancer Types ◽  
Cancer Associated Thrombosis ◽  
Deep Vein

SummaryVenous thromboembolism (VTE) is a clinically important complication for both hospitalised and ambulatory cancer patients. In the current study, the frequency, demographics and risk (according to tumour site) of VTE were examined among patients seen at outpatient DVT (deep-vein thrombosis) clinics. Of 10,015 VTE cases, 1,361 were diagnosed with cancer, for an overall rate of cancer-associated VTE of 13.6% in this outpatient population. Patients with cancer-associated VTE were significantly older than cancer-free VTE cases (66.4 ± 12.7 vs. 58.8 ± 18.5 years; p<0.0001). The frequency of cancer-associated VTE peaked earlier among females than males, occurring in the sixth (137/639, 21.4% vs. 98/851, 11.3%; p<0.001) and seventh decades (213/980, 21.7% vs. 197/1096, 18%; p=0.036). VTE was described most frequently in common cancers – breast, prostate, colorectal and lung (56.1% of cases). The risk of VTE varied widely across 17 cancer types. Calculating odds ratios (OR) to assess the effect size of cancer type on VTE risk, the highest odds were observed for patients with pancreatic cancer (OR 9.65, 95% confidence interval [CI] (5.51–16.91). Tumours of the head and neck had higher odds than previously reported (OR 8.24, 95% CI 5.06–13.42). Reduced risk estimates were observed for skin cancers (melanoma and non-melanoma: OR 0.89, 95% CI 0.42–1.87; OR 0.74, 95% CI, 0.32–1.69, respectively). We conclude that outpatients have a similar rate of cancer-associated VTE as VTE patient populations previously reported, that cancer-associated VTE occurs in an older age group and earlier in females and that outpatients exhibit distinct tumour site-specific risk from that described among hospitalised cancer patients.

scholarly journals Adaptive immunity and neutralizing antibodies against SARS-CoV-2 variants of concern following vaccination in patients with cancer: The CAPTURE study

2021 ◽  
Author(s):  
Annika Fendler ◽  
Scott Shepherd ◽  
Lewis Au ◽  
Katalin Wilkinson ◽  
Mary Wu ◽  
...  
Keyword(s):  
Cancer Patients ◽  
Neutralizing Antibodies ◽  
Hematological Malignancies ◽  
Neutralizing Antibody ◽  
Patients With Cancer ◽  
Cell Responses ◽  
Solid Malignancies ◽  
Antibody Titers ◽  
Cancer Types ◽  
Anti Cd20

Abstract CAPTURE (NCT03226886) is a prospective cohort study of COVID-19 immunity in patients with cancer. Here we evaluated 585 patients following administration of two doses of BNT162b2 or AZD1222 vaccines, administered 12 weeks apart. Seroconversion rates after two doses were 85% and 59% in patients with solid and hematological malignancies, respectively. A lower proportion of patients had detectable neutralizing antibody titers (NAbT) against SARS-CoV-2 variants of concern (VOCs) vs wild-type (WT). Patients with hematological malignancies were more likely to have undetectable NAbT and had lower median NAbT vs solid cancers against both WT and VOCs. In comparison with individuals without cancer, patients with haematological, but not solid, malignancies had reduced NAb responses. Seroconversion showed poor concordance with NAbT against VOCs. Prior SARS-CoV-2 infection boosted NAb response including against VOCs, and anti-CD20 treatment was associated with undetectable NAbT. Vaccine-induced T-cell responses were detected in 80% of patients, and were comparable between vaccines or cancer types. Our results have implications for the management of cancer patients during the ongoing COVID-19 pandemic.

Predictors Of Vena Cava Filter Use For Venous Thromboembolism In Cancer Patients

Blood ◽  
2013 ◽  
Vol 122 (21) ◽  
pp. 935-935
Author(s):  
Gwendolyn Ho ◽  
Ann Brunson ◽  
Richard H. White ◽  
Ted Wun
Keyword(s):  
Risk Factors ◽  
Venous Thromboembolism ◽  
Cancer Patients ◽  
Brain Cancer ◽  
Vena Cava ◽  
Severity Of Illness ◽  
Bleeding Risk ◽  
Factors Associated ◽  
Patients With Cancer ◽  
Cancer Types

Abstract Background The use of vena cava filters (VCF) in the treatment of venous thromboembolism (VTE) is controversial. Few studies have evaluated the use of VCFs in cancer patients with acute thrombosis. Aims To determine frequency of VCF placement and factors associated with VCF use in patients with cancer hospitalized for acute VTE, and to compare these findings to patients without cancer hospitalized for acute VTE. Methods Using a retrospective observational study design, we analyzed hospital discharge records in California from 2005-2009 of cases presenting with acute VTE. Patients with cancer were identified by specific ICD-9-CM codes for the index VTE admission or a cancer diagnosis within 6 months prior to the index VTE. Bivariate and multivariable logistic regression analyses were used to determine predictive factors for placement of a VCF in cancer patients. Candidate risk factors included basic demographic parameters, cancer type, severity-of-illness (SOI) on admission, undergoing surgery, bleeding, and hospital characteristics. Results A VCF was placed in 19.6% of 14,000 cancer cases admitted with a principal diagnosis of acute VTE, versus 10.8% of 70,472 non-cancer cases admitted during the same time period. Among cancer cases, there was little variation in percentage that received a VCF based on age, and no significant variation across race or insurance type, except that self pay cancer patients had a lower rate of VCF placement. Variation across hospitals in the percentage of cancer cases that received a VCF was striking, ranging from 0% to 52% among hospitals that admitted a minimum of 15 acute VTE cases. There was a strong correlation (r=0.72, R2=0.52) in the frequency of VCF placement in cancer and non-cancer cases within individual hospitals. Among cancer types, the frequency of VCF placement was highest in cases with brain cancer (43%), with the observed frequency of VCF use among other cancer types ranging from 8%-23%. Patients with brain cancers, which has a high perceived bleeding risk were over 4 fold more likely to have a VCF placed compared to those cancers with low bleeding risk. Having acute leukemia did not predict for VCF placement. Only 8.2% of cancer patients had a strict contraindication to anticoagulation (acute bleeding or recent/imminent surgery), which are the only guideline-based indications for VCF placement. Active bleeding and undergoing surgery were each strongly associated with VCF use: 47% of cases that bled and 58% of cases who underwent surgery had a VCF placed. Results of the multivariable logistic model are shown in the table. In addition to bleeding and undergoing surgery, factors associated with VCF insertion included: larger hospital, urban location, private hospital and greater SOI at the time of admission. Conclusions The frequency of VCF use in cancer patients admitted for acute VTE is much higher than in non-cancer patients. Major risk factors for VCF use include bleeding, undergoing recent surgery, having brain cancer, urban location, and greater severity of illness. The frequency of VCF placement among cancer patients varied widely across hospitals. Given the extraordinary variation in the frequency of use of VCFs between hospitals, more research is needed to better define outcomes of VCF placement in cancer patients. Disclosures: Ho: American Society of Hematology: ASH HONORS trainee research award Other.

scholarly journals Clinical and Imaging Characteristics of Cancer Patients with COVID-19: A Pilot Study

2021 ◽  
Vol 14 (10) ◽  
Author(s):  
Anya Jafari ◽  
Zahra Mahboubi-Fooladi ◽  
Zahra Siavashpour ◽  
Afshin Rakhsha ◽  
Sahar Mirbaha ◽  
...  
Keyword(s):  
Cancer Patients ◽  
Cancer Disease ◽  
Gastrointestinal Malignancies ◽  
Patients With Cancer ◽  
Patient Interviews ◽  
Common Cancer ◽  
Imaging Characteristics ◽  
Stage 4 ◽  
Cancer Types ◽  
Tomography Scan

Background: Malignancy is a known risk factor of coronavirus disease 2019 (COVID-19) severe involvement. Information about this infection in patients with cancer is limited. Objectives: This study aimed at reporting the clinical and imaging characteristics of COVID-19 infection in patients with cancer. Methods: All the patients were known cases of a solid tumor with COVID-19 infection in one center, between February and May 2020. Clinical presentation and imaging involvement of COVID-19 infection in addition to cancer features were documented from medical records/patient interviews. Results: Thirty-one patients with solid tumors and COVID-19 involvement were included. The most prevalet presentation was fever, cough, and myalgia. Breast and gastrointestinal malignancies were the most common cancer types. The mortality rate was 22.5% and all deceased patients suffered from stage 4 of their underlying cancer disease. Lung computed tomography scan (CT scan) features in these patients were not different from the non-cancer patients with COVID-19. Conclusions: COVID-19 involvement in patients with cancer seems to be more severe with higher mortality rates especially in patients with other comorbidity and in metastatic cases. Treatment modifications during the pandemic era sound to be logical in decreasing the infection rate.

Sign in / Sign up

Export Citation Format

Share Document