Application of deep learning to predict advanced neoplasia using big clinical data in colorectal cancer screening of asymptomatic adults.
Ontology highlight
ABSTRACT: Background/aims
We aimed to develop a deep learning model for the prediction of the risk of advanced colorectal neoplasia (ACRN) in asymptomatic adults, based on which colorectal cancer screening could be customized.Methods
We collected data on 26 clinical and laboratory parameters, including age, sex, smoking status, body mass index, complete blood count, blood chemistry, and tumor marker, from 70,336 first-time colonoscopy screening recipients. For reference, we used a logistic regression (LR) model with nine variables manually selected from the 26 variables. A deep neural network (DNN) model was developed using all 26 variables. The area under the receiver operating characteristic curve (AUC), sensitivity, and specificity of the models were compared in a randomly split validation group.Results
In comparison with the LR model (AUC, 0.724; 95% confidence interval [CI], 0.684 to 0.765), the DNN model (AUC, 0.760; 95% CI, 0.724 to 0.795) demonstrated significantly improved performance with respect to the prediction of ACRN (p < 0.001). At a sensitivity of 90%, the specificity significantly increased with the application of the DNN model (41.0%) in comparison with the LR model (26.5%) (p < 0.001), indicating that the colonoscopy workload required to detect the same number of ACRNs could be reduced by 20%.Conclusion
The application of DNN to big clinical data could significantly improve the prediction of ACRNs in comparison with the LR model, potentially realizing further customization by utilizing large quantities and various types of biomedical information.
SUBMITTER: Yang HJ
PROVIDER: S-EPMC8273821 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:Eliciting patient preferences within the context of shared decision making has been advocated for colorectal cancer screening. Risk stratification for advanced colorectal neoplasia (ACN) might facilitate more effective shared decision making when selecting an appropriate screening option. Our objective was to develop and validate a clinical index for estimating the probability of ACN at screening colonoscopy.We conducted a cross-sectional analysis of 3,543 asymptomatic, mostly average-risk patients 50-79 years of age undergoing screening colonoscopy at two urban safety net hospitals. Predictors of ACN were identified using multiple logistic regression. Model performance was internally validated using bootstrapping methods.The final index consisted of five independent predictors of risk (age, smoking, alcohol intake, height, and a combined sex/race/ethnicity variable). Smoking was the strongest predictor (net reclassification improvement (NRI), 8.4%) and height the weakest (NRI, 1.5%). Using a simplified weighted scoring system based on 0.5 increments of the adjusted odds ratio, the risk of ACN ranged from 3.2% (95% confidence interval (CI), 2.6-3.9) for the low-risk group (score ?2) to 8.6% (95% CI, 7.4-9.7) for the intermediate/high-risk group (score 3-11). The model had moderate to good overall discrimination (C-statistic, 0.69; 95% CI, 0.66-0.72) and good calibration (P=0.73-0.93).A simple 5-item risk index based on readily available clinical data accurately stratifies average-risk patients into low- and intermediate/high-risk categories for ACN at screening colonoscopy. Uptake into clinical practice could facilitate more effective shared decision-making for CRC screening, particularly in situations where patient and provider test preferences differ.
| S-EPMC4705553 | biostudies-literature
Project description:BACKGROUND:Colorectal cancer (CRC) is a common and potentially preventable malignancy. Evidence has emerged that coronary artery disease patients are at increased risk for developing CRC by shared risk factors. Here we investigated an association between CRC and atrial fibrillation (AF), a surrogate marker of cardiovascular risk, in the setting of routine screening colonoscopy. METHODS:We investigated 1949 asymptomatic participants (median age 61 [54-67] years, 49% females) undergoing screening colonoscopy within the SAKKOPI registry (Salzburg Colon Cancer Prevention Initiative). Forty-six participants with AF (2.4%) were identified, and colonoscopy findings were compared to non-AF participants. Propensity Score Matching (PSM) was used to create 1:1 and 3:1 age- and gender-matched couples. RESULTS:Abnormal findings on screening colonoscopy (any form of adenoma or carcinoma) were more common in AF participants with an odds ratios (OR) of 2.4 [1.3-4.3] in the unmatched analysis, and 2.6 [1.1-6.3] and 2.0 [1.1-4.0] in the 1:1 and 3:1 matched groups, respectively. Correspondingly, the odds of finding advanced adenomas or carcinomas was elevated about three-fold across the different matched and unmatched analyses (OR 3.3 [1.1-10.8] for 3:1 matched participants). At the same time, the prevalence and number of colonic lesions were significantly higher in AF participants (63.0% vs. 33.4% for 3:1 matched participants, p < 0.001). Non-CRC related findings on colonoscopy, like diverticulosis, were non-different between groups. CONCLUSION:Participants with AF had a higher burden of advanced premalignant adenomas and CRC in routine colonoscopy screening. Our data suggest that practitioners should monitor the CRC screening status, especially in AF patients.
| S-EPMC6678897 | biostudies-other
Project description:Following colonoscopic polypectomy, US Multisociety Task Force (USMSTF) guidelines stratify patients based on risk of subsequent advanced neoplasia (AN) using number, size, and histology of resected polyps, but have only moderate sensitivity and specificity. We hypothesized that a state-of-the-art statistical prediction model might improve identification of patients at high risk of future AN and address these challenges.Data were pooled from seven prospective studies which had follow-up ascertainment of metachronous AN within 3-5 years of baseline polypectomy (combined n = 8,228). Pooled data were randomly split into training (n = 5,483) and validation (n = 2,745) sets. A prognostic model was developed using best practices. Two risk cut-points were identified in the training data which achieved a 10 percentage point improvement in sensitivity and specificity, respectively, over current USMSTF guidelines. Clinical benefit of USMSTF versus model-based risk stratification was then estimated using validation data.The final model included polyp location, prior polyp history, patient age, and number, size and histology of resected polyps. The first risk cut-point improved sensitivity but with loss of specificity. The second risk cut-point improved specificity without loss of sensitivity (specificity 46.2 % model vs. 42.1 % guidelines, p < 0.001; sensitivity 75.8 % model vs. 74.0 % guidelines, p = 0.64). Estimated AUC was 65 % (95 % CI: 62-69 %).This model-based approach allows flexibility in trading sensitivity and specificity, which can optimize colonoscopy over- versus underuse rates. Only modest improvements in prognostic power are possible using currently available clinical data. Research considering additional factors such as adenoma detection rate for risk prediction appears warranted.
| S-EPMC5495135 | biostudies-literature
Project description:BACKGROUND:Several methods are recommended equally strongly for colorectal cancer screening in average-risk persons. Risk stratification would enable tailoring of screening within this group, with less invasive tests (sigmoidoscopy or occult blood tests) for lower-risk persons and colonoscopy for higher-risk persons. OBJECTIVE:To create a risk index for advanced neoplasia (colorectal cancer and adenomas or serrated polyps ?1.0 cm, villous histology, or high-grade dysplasia) anywhere in the colorectum, using the most common risk factors for colorectal neoplasia. DESIGN:Cross-sectional study. SETTING:Multiple endoscopy units, primarily in the Midwest. PATIENTS:Persons aged 50 to 80 years undergoing initial screening colonoscopy (December 2004 to September 2011). MEASUREMENTS:Derivation and validation of a risk index based on points from regression coefficients for age, sex, waist circumference, cigarette smoking, and family history of colorectal cancer. RESULTS:Among 2993 persons in the derivation set, prevalence of advanced neoplasia was 9.4%. Risks for advanced neoplasia in persons at very low, low, intermediate, and high risk were 1.92% (95% CI, 0.63% to 4.43%), 4.88% (CI, 3.79% to 6.18%), 9.93% (CI, 8.09% to 12.0%), and 24.9% (CI, 21.1% to 29.1%), respectively (P < 0.001). Sigmoidoscopy to the descending colon in the low-risk groups would have detected 51 of 70 (73% [CI, 61% to 83%]) advanced neoplasms. Among 1467 persons in the validation set, corresponding risks for advanced neoplasia were 1.65% (CI, 0.20% to 5.84%), 3.31% (CI, 2.08% to 4.97%), 10.9% (CI, 8.26% to 14.1%), and 22.3% (CI, 16.9% to 28.5%), respectively (P < 0.001). Sigmoidoscopy would have detected 21 of 24 (87.5% [CI, 68% to 97%]) advanced neoplasms. LIMITATIONS:Split-sample validation; results apply to first-time screening. CONCLUSION:This index stratifies risk for advanced neoplasia among average-risk persons by identifying lower-risk groups for which noncolonoscopy strategies may be effective and efficient and a higher-risk group for which colonoscopy may be preferred. PRIMARY FUNDING SOURCE:National Cancer Institute, Walther Cancer Institute, Indiana University Simon Cancer Center, and Indiana Clinical and Translational Sciences Institute.
| S-EPMC4840411 | biostudies-literature
Project description:Chinese translationBlack persons are more likely than white persons to be diagnosed with colorectal cancer and to die from it. The extent to which genetic or biological factors versus disparities in screening rates explain this variance remains controversial.To define the prevalence and location of presymptomatic advanced colorectal neoplasia (ACN) among white and black persons undergoing screening colonoscopy, controlling for other epidemiologic risk factors.Cross-sectional survey between 22 March 2005 and 31 January 2012.Urban, open-access, academic, safety-net hospital in Massachusetts.Asymptomatic, average-risk white (n = 1172) and black (n = 1681) persons aged 50 to 79 years undergoing screening colonoscopy.Adjusted prevalence and location of ACN, defined as a tubular adenoma 10 mm or more in size, any adenoma with villous features or high-grade dysplasia, any dysplastic serrated lesion, or invasive cancer.The prevalence of ACN was higher among white patients than black patients (6.8% vs. 5.0%; P = 0.039) but varied by sex (white vs. black men, 9.3% vs. 5.7%; white vs. black women, 3.5% vs. 4.3%; interaction P = 0.034). After controlling for many risk factors, black men were 41% less likely than white men (adjusted odds ratio [AOR], 0.59 [95% CI, 0.39 to 0.89]) to have ACN. No statistically significant difference was seen for women (AOR, 1.32 [CI, 0.73 to 2.40]). Black patients with ACN had a higher percentage of proximal disease (52% vs. 39%) after adjustment for age and sex (P = 0.055).Single-institution study with inadequate statistical power for subgroup analyses and recall bias.Black men are less likely than white men to have ACN at screening colonoscopy in a safety-net health care setting. Disparities in access to screening and differential exposure to modifiable risk factors rather than genetic or biological factors may be largely responsible for the higher incidence of CRC among black men. Genetic or biological factors may explain the predilection for proximal disease.
| S-EPMC4218749 | biostudies-other
Project description:ObjectivesTo assess detection rates for colorectal cancer (CRC) and advanced adenomas in asymptomatic CRC screening participants and bowel symptoms in association with CRC and advanced adenoma.DesignCross-sectional study.SettingTwo screening centres.Participants42 554 men and women, aged 50-74 years, participating in a randomised CRC screening trial. 36 059 participants underwent a sigmoidoscopy (and follow-up colonoscopy if positive sigmoidoscopy) and 6495 underwent a colonoscopy after a positive faecal immunochemical test (FIT).Primary and secondary outcome measuresProportion of asymptomatic participants diagnosed with CRC or advanced adenomas. Prevalence of bowel symptoms (rectal bleeding, change in bowel habits, diarrhoea, constipation, bloating, alternating bowel habits, general symptoms, other bowel symptoms) recorded by the endoscopist and their association with CRC and advanced adenomas.ResultsAmong sigmoidoscopy participants, 7336 (20.3%) reported at least one symptom. 120 (60%) out of 200 individuals with screen-detected CRC and 1301 (76.5%) out of 1700 with advanced adenoma were asymptomatic. Rectal bleeding was associated with detection of CRC and advanced adenoma (OR 4.3, 95% CI 3.1 to 6.1 and 1.8, 95% CI 1.5 to 2.1, respectively), while change in bowel habits only with CRC detection (OR 3.8, 95% CI 2.4 to 6.1). Among the FIT positives, 2173 (33.5%) reported at least one symptom. Out of 299 individuals with screen-detected CRC and 1639 with advanced adenoma, 167 (55.9%) and 1 175 (71.7%) were asymptomatic, respectively. Detection of CRC was associated with rectal bleeding (OR 1.8, 95% CI 1.4 to 2.3), change in bowel habits (OR 2.2, 95% CI 1.4 to 3.5) and abdominal pain (OR 1.8, 95% CI 1.2 to 2.7).ConclusionsSome bowel symptoms increased the likelihood of being diagnosed with CRC or advanced adenoma. However, the majority of individuals with these findings were asymptomatic. Asymptomatic individuals should be encouraged to participate in CRC screening.Trial registration numberClinicaltrials.gov Identifier: NCT01538550.
| S-EPMC8252874 | biostudies-literature
Project description:BACKGROUND: It is uncertain whether synchronous colorectal cancers (S-CRCs) preferentially develop through widespread DNA methylation and whether they have a prognosis worse than solitary CRC. As tumours with microsatellite instability (MSI) may confound the effect of S-CRC methylation on outcome, we addressed this issue in a series of CRC characterised by BRAF and MS status. METHODS: Demographics, clinicopathological records and disease-specific survival (DSS) were assessed in 881 consecutively resected CRC undergoing complete colonoscopy. All tumours were typed for BRAF(c.1799T>A) mutation and MS status, followed by search of germ-line mutation in patients with MSI CRC. RESULTS: Synchronous colorectal cancers (50/881, 5.7%) were associated with stage IV microsatellite-stable (MSS) CRC (19/205, 9.3%, P=0.001) and with HNPCC (9/32, 28%, P<0.001). BRAF mutation (60/881, 6.8%) was associated with sporadic MSI CRC (37/62, 60%, P<0.001) but not with S-CRC (3/50, 6.0%, P=0.96). Synchronous colorectal cancer (HR 1.82; 95% CI 1.15-2.87; P=0.01), synchronous advanced adenoma (HR 1.81; 95% CI 1.27-2.58; P=0.001), and BRAF(c.1799T>A) mutation (HR 2.16; 95% CI 1.25-3.73; P=0.01) were stage-independent predictors of death from MSS CRC. Disease-specific survival of MSI CRC patients was not affected by S-CRC (HR 0.74; 95% CI 0.09-5.75; P=0.77). CONCLUSION: Microsatellite-stable CRCs have a worse prognosis if S-CRC or synchronous advanced adenoma are diagnosed. The occurrence and the enhanced aggressiveness of synchronous MSS advanced neoplasia are not associated with BRAF mutation.
| S-EPMC3950856 | biostudies-other
Project description:Infectious disease occurs when a person is infected by a pathogen from another person or an animal. It is a problem that causes harm at both individual and macro scales. The Korea Center for Disease Control (KCDC) operates a surveillance system to minimize infectious disease contagions. However, in this system, it is difficult to immediately act against infectious disease because of missing and delayed reports. Moreover, infectious disease trends are not known, which means prediction is not easy. This study predicts infectious diseases by optimizing the parameters of deep learning algorithms while considering big data including social media data. The performance of the deep neural network (DNN) and long-short term memory (LSTM) learning models were compared with the autoregressive integrated moving average (ARIMA) when predicting three infectious diseases one week into the future. The results show that the DNN and LSTM models perform better than ARIMA. When predicting chickenpox, the top-10 DNN and LSTM models improved average performance by 24% and 19%, respectively. The DNN model performed stably and the LSTM model was more accurate when infectious disease was spreading. We believe that this study's models can help eliminate reporting delays in existing surveillance systems and, therefore, minimize costs to society.
| S-EPMC6121625 | biostudies-literature
Project description:Histopathology is the gold standard method for staging and grading human tumors and provides critical information for the oncoteam's decision making. Highly-trained pathologists are needed for careful microscopic analysis of the slides produced from tissue taken from biopsy. This is a time-consuming process. A reliable decision support system would assist healthcare systems that often suffer from a shortage of pathologists. Recent advances in digital pathology allow for high-resolution digitalization of pathological slides. Digital slide scanners combined with modern computer vision models, such as convolutional neural networks, can help pathologists in their everyday work, resulting in shortened diagnosis times. In this study, 200 digital whole-slide images are published which were collected via hematoxylin-eosin stained colorectal biopsy. Alongside the whole-slide images, detailed region level annotations are also provided for ten relevant pathological classes. The 200 digital slides, after pre-processing, resulted in 101,389 patches. A single patch is a 512 × 512 pixel image, covering 248 × 248 μm2 tissue area. Versions at higher resolution are available as well. Hopefully, HunCRC, this widely accessible dataset will aid future colorectal cancer computer-aided diagnosis and research.
| S-EPMC9240013 | biostudies-literature
Project description:PURPOSE:Colorectal cancer (CRC) screening guidelines recommend increased surveillance of individuals with sessile serrated adenomas/polyps (SSA/Ps), but there is uncertainty about the risk associated with SSA/Ps. We aimed to determine the association between SSA/Ps and subsequent advanced colorectal neoplasia. METHODS:This case-control study included Kaiser Permanente Washington (KPWA) members who received an index colonoscopy between 1/1/1998 and 12/31/2007, and had hyperplastic polyps (HPs) or SSA/Ps but no conventional adenomas according to study pathologist histologic review. Subsequent pathology reports and biopsies through 1/1/2013 were reviewed for advanced colorectal neoplasia. We linked to the Seattle-Puget Sound Surveillance Epidemiology and End Results (SEER) registry to identify additional CRC cases. We used generalized estimating equations with a logit link to estimate adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for advanced colorectal neoplasia, comparing those with SSA/Ps to those with HPs. RESULTS:There were 161 individuals with index SSA/Ps, 548 with HPs, and 918 subsequent endoscopies included in analyses. Of those with index SSA/Ps, 19 had subsequent advanced colorectal neoplasia; 39 with HPs had subsequent advanced colorectal neoplasia. Compared to those with HPs, those with SSA/Ps were not statistically significantly more likely to have subsequent advanced colorectal neoplasia (adjusted OR 1.79; CI 0.98-3.28). Polyp size???10 mm, right colon location, and the presence of multiple serrated polyps were also not associated with advanced colorectal neoplasia. CONCLUSIONS:Our results suggest that there is not a strong association between SSA/Ps and subsequent advanced colorectal neoplasia during the 5 years following SSA/P removal.
| S-EPMC6736607 | biostudies-literature