机器学习算法在肝细胞癌微血管侵犯术前预测中的应用价值

目的探讨机器学习算法在肝细胞癌微血管侵犯(MVI)术前预测中的应用价值。方法采用回顾性描述性研究方法。收集2015年5月至2018年12月福建医科大学孟超肝胆医院收治的277例肝细胞癌患者的临床病理资料;男235例,女42例;年龄为(56±10)岁,年龄范围为33~80岁。患者术前均行磁共振成像检查。227例肝细胞癌患者通过计算机产生随机数方法以7∶3比例分为训练集193例和验证集84例。应用逻辑回归列线图,支持向量机(SVM)、随机森林(RF)、人工神经网络(ANN)和轻量级梯度提升机(LightGBM)机器学习算法构建MVI术前预测模型。观察指标:(1)训练集及验证集患者临床病理资料分析。(2)影响训练集患者肿瘤MVI危险因素分析。(3)机器学习算法预测模型构建及其术前预测肿瘤MVI准确性比较。正态分布的计量资料以±s表示,组间比较采用配对t检验。计数资料以绝对数表示,组间比较采用χ2检验。单因素和多因素分析采用Logistic回归模型。结果(1)训练集及验证集患者临床病理资料分析:训练集和验证集患者性别(男,女)分别为157、36例和78、6例,两组比较,差异有统计学意义(χ2=6.028,P<0.05)。(2)影响训练集患者肿瘤MVI危险因素分析:训练集193例患者中,MVI阳性108例,MVI阴性85例。单因素分析结果显示:年龄、肿瘤数目、肿瘤直径、卫星病灶、肿瘤边界、甲胎蛋白(AFP)、碱性磷酸酶(ALP)和纤维蛋白原水平是影响肿瘤MVI的相关因素(比值比=0.971,2.449,1.368,4.050,2.956,4.083,2.532,1.996,95%可信区间为0.943~1.000,1.169~5.130,1.180~1.585,1.316~12.465,1.310~6.670,2.214~7.532,1.016~6.311,1.323~3.012,P<0.05)。多因素分析结果显示:AFP>20μg/L、肿瘤多发、肿瘤直径越大、肿瘤边界不光滑是影响肿瘤MVI的独立危险因素(比值比=3.680,3.100,1.438,3.628,95%可信区间为1.842~7.351,1.334~7.203,1.201~1.721,1.438~9.150,P<0.05),而年龄越大,MVI发生风险越低(比值比=0.958,95%可信区间为0.923~0.994,P<0.05)。(3)机器学习算法预测模型构建及其术前预测肿瘤MVI准确性比较:①应用多因素分析结果筛选指标,包括年龄、AFP、肿瘤数目、肿瘤直径、肿瘤边界,构建逻辑回归列线图,SVM、RF、ANN及LightGBM机器学习算法预测模型,一致性分析结果显示逻辑回归列线图预测模型稳定性较好。逻辑回归列线图、SVM、RF、ANN、LightGBM机器学习算法预测模型训练集和验证集曲线下面积(AUC)分别为0.812、0.794、0.807、0.814、0.810和0.784、0.793、0.783、0.803、0.815,SVM、RF、ANN、LightGBM机器学习算法AUC分别与逻辑回归列线图AUC比较,差异均无统计学意义[(95%可信区间为0.731~0.849,0.744~0.860,0.752~0.867,0.747~0.862,Z=0.995,0.245,0.130,0.102,P>0.05)和(95%可信区间为0.690~0.873,0.679~0.865,0.702~0.882,0.715~0.891,Z=0.325,0.026,0.744,0.803,P>0.05)]。②应用RF、LightGBM机器学习算法自行筛选临床病理因素指标构建预测模型。根据指标对预测模型重要度排序,选择重要度>0.01的指标,包括年龄、肿瘤直径、AFP、白细胞(WBC)、血小板、总胆红素、天冬氨酸氨基转移酶、γ-谷氨酰转移酶、ALP和纤维蛋白原,构建RF机器学习算法预测模型;挑选重要度>5.0的指标,包括年龄、肿瘤直径、AFP、WBC、ALP和纤维蛋白原,构建LightGBM机器学习算法预测模型;由于ANN及SVM机器学习算法不具备筛选指标能力,应用单因素分析结果筛选指标,包括年龄、肿瘤数目、肿瘤直径、卫星病灶、肿瘤边界、AFP、ALP和纤维蛋白原水平,构建SVM、ANN机器学习算法预测模型。SVM、RF、ANN、LightGBM机器学习算法预测模型训练集和验证集AUC分别为0.803、0.838、0.793、0.847和0.810、0.802、0.802、0.836,分别与逻辑回归列线图AUC比较,差异均无统计学意义[(95%可信区间为0.740~0.857,0.779~0.887,0.729~0.848,0.789~0.895,Z=0.421,0.119,0.689,1.517,P>0.05)和(95%可信区间为0.710~0.888,0.700~0.881,0.701~0.881,0.740~0.908,Z=0.856,0.458,0.532,1.306,P>0.05)]。结论机器学习算法可用于术前预测肝细胞癌MVI,但其应用价值尚需多中心大样本数据进一步验证。     

心率减速力测定预测室性心律失常患者心脏猝死的价值

目的对176例经Holter心电图监护所检出的室性心律失常患者行心率减速力(DC)检测,评价其对心脏猝死的临床预测价值。方法采用美国DMS公司的12导联同步描记Holter心电监护仪,对门诊及病房就诊的受检者进行24⁷2 h的Holter心电图监护,除常规心律失常及ST段监护指标外,同时检测窦性心率震荡(HRT)及心率变异(HRV)的心搏间RR间期标准差指标(SDNN)。将Holter心电监护直接记录出室性心动过速(VT)者列为观察组,共176例,男93例,女83例,年龄3972 h的Holter心电图监护,除常规心律失常及ST段监护指标外,同时检测窦性心率震荡(HRT)及心率变异(HRV)的心搏间RR间期标准差指标(SDNN)。将Holter心电监护直接记录出室性心动过速(VT)者列为观察组,共176例,男93例,女83例,年龄39⁹2岁,平均53岁;随机选取性别和年龄与观察组相仿的182例为对照组进行对比分析。结果以心率减速力<4.5 ms为异常和<2.5 ms为显著异常的标准,两组的心率减速力值异常率分别为50.6%(89/176)和4.4%(8/182),观察组明显高于对照组(χ2=96.6,P<0.05);两组的心率减速力显著异常率分别为33.0%(58/176)和0%(0/182),两组间差异显著(χ2=75.6,P<0.05);室性心动过速患者的心率减速力异常检出率明显高于HRT和HRV检测指标(χ2=106.4,P<0.05;χ2=58.6,P<0.05)。结论室性心动过速患者的心率减速力测定,有较高的异常检出率。     

Blood- and tissue-based biomarkers for prediction of outcomes in urothelial carcinoma of the bladder

ObjectivesUrothelial carcinoma of the bladder (UCB) is a highly heterogeneous malignancy that causes significant morbidity and mortality. Standard pathologic features (stage, grade, and nodal status) are insufficient to predict accurately a patient's outcome. Biomarkers could help clinicians provide individualized prognostications and allow risk-stratified clinical decision making regarding surgical and medical treatment. This review summarizes the existing tissue- and blood-based biomarkers in UCB.Material and methodsA PubMed/Medline search was conducted to identify original articles regarding molecular biomarkers and UCB. Searches were limited to papers published in English. Keywords included urothelial carcinoma, bladder cancer, transitional cell, biomarker, marker, staining, cystectomy, recurrence or progression, survival, prediction, and prognosis.ResultsThe articles with the highest level of evidence were selected and reviewed, with the consensus of all the authors of this paper.ConclusionsThere is no doubt that a panel of biomarkers would eventually improve our clinical decision making regarding treatment and follow-up. However, to date, no biomarker panel is yet validated for daily clinical practice.     

Investigating predictability of in vitro toxicological assessments of cigarettes: Analysis of 7 years of regulatory submissions to Canadian regulatory authorities

A wealth of in vitro toxicological information on different types of tobaccos and tobacco products has been acquired and published, although the link between in vitro data and impact on human health remains elusive. The present study investigates the possibility of establishing quantitative models for the in vitro toxicological endpoint responses to cigarette smoke. To this end, it relies on information submitted to Canadian health authorities during the period 2006–2012. To our knowledge, this is the first time that published results concerning the influence of such factors as cigarette blend, diameter and filter type on in vitro toxicity are confirmed at the level of a representative range of products on a market. Taking these cigarette design features into account and adding a limited amount of quantitative mainstream smoke composition information, it is shown that, within the boundaries of the considered cigarette design parameters, the in vitro toxicological response can be effectively predicted. In vitro tests of tobacco products are an invaluable initial comparative product assessment tool. The present results reveal the limited value of data from repeated tests on products which do not undergo significant modifications.     

Prediction of pancreatic fistula after pancreatectomy by virtual touch tissue imaging and quantification (VTIQ) technology

ObjectivesThe aim of this study was to quantitatively evaluate the stiffness of pancreatic parenchyma and solid focal pancreatic lesions (FPLs) by virtual touch tissue imaging and quantification (VTIQ) technique and to investigate the potential usefulness of VTIQ method in the prediction of post-operative pancreatic fistula (POPF) after pancreatectomy.MethodsIn this prospective study, patients who scheduled to undergo pancreatectomy were initially enrolled and received VTIQ assessment within one week before surgery. VTIQ elastography (Siemens ACUSON Sequoia, 5C-1 transducer) was used to measure the shear wave velocity (SWV) value of FPLs and the body part pancreatic parenchyma. The palpation stiffness of pancreas was qualitatively evaluated during operation by surgeons. POPF was finally diagnosed and graded through a three-weeks post-operative follow-up according to international study group of pancreatic fistula (ISGPF). SWV values were compared between POPF positive and negative group. Receiver operating characteristic (ROC) analysis was used to evaluate the diagnostic efficacy of SWV value in predicting POPF.ResultsFrom December 2020 to June 2021, 44 patients were finally enrolled in this study, among which, 26 patients were identified to develop POPF after pancreatectomy. The SWV value of pancreatic parenchyma in POPF positive group was significantly lower than that in POPF negative group (P = 0.001). However, there was no significant difference in palpation stiffness between the two groups (P = 0.124). Besides, neither the SWV value of FPL nor the SWV ratio between FPL to surrounding pancreatic parenchyma differ significantly between POPF positive and negative group (P > 0.05). Taking SWV value of pancreatic parenchyma >1.10 m/s as a cut-off value for predicting POPF, area under the receiver operating characteristic curve (AUROC) was 0.864 with 72.2% sensitivity, 92.3% specificity, 86.7% positive predictive value (PPV) and 82.8% negative predictive value (NPV), respectively.ConclusionsVTIQ technique might be a potential non-invasive imaging method to predict POPF before pancreatectomy in future clinical practice.     

Deep Neural Network-Based Prediction of the Risk of Advanced Colorectal Neoplasia

Background/AimsRisk prediction models using a deep neural network (DNN) have not been reported to predict the risk of advanced colorectal neoplasia (ACRN). The aim of this study was to compare DNN models with simple clinical score models to predict the risk of ACRN in colorectal cancer screening.MethodsDatabases of screening colonoscopy from Kangbuk Samsung Hospital (n=121,794) and Kyung Hee University Hospital at Gangdong (n=3,728) were used to develop DNN-based prediction models. Two DNN models, the Asian-Pacific Colorectal Screening (APCS) model and the Korean Colorectal Screening (KCS) model, were developed and compared with two simple score models using logistic regression methods to predict the risk of ACRN. The areas under the receiver operating characteristic curves (AUCs) of the models were compared in internal and external validation databases.ResultsIn the internal validation set, the AUCs of DNN model 1 and the APCS score model were 0.713 and 0.662 (p<0.001), respectively, and the AUCs of DNN model 2 and the KCS score model were 0.730 and 0.667 (p<0.001), respectively. However, in the external validation set, the prediction performances were not significantly different between the two DNN models and the corresponding APCS and KCS score models (both p>0.1).ConclusionsSimple score models for the risk prediction of ACRN are as useful as DNN-based models when input variables are limited. However, further studies on this issue are warranted to predict the risk of ACRN in colorectal cancer screening because DNN-based models are currently under improvement.     

Predictors of upper limb spasticity after stroke? A systematic review and meta-analysis

ObjectiveTo determine the predictive markers for the occurrence of upper limb spasticity in the first 12 months after stroke.Data sourcesA systematic review was undertaken of the databases MEDLINE, EMBASE, CINAHL and PEDRO to 31st December 2017.Study selectionNon-experimental or experimental studies that included a control group with spasticity who did not receive an experimental intervention which investigated at least one variable (explanatory variable) measured at baseline against the development (or not) of spasticity at a future time point within 12 months post stroke were selected independently by two reviewers. Eleven papers met the selection criteria.Study appraisalData were extracted into tabular format using predefined data fields by two reviewers. Study quality was evaluated using the modified Downs and Black tool. Data were analysed using a meta-analysis or narrative review.ResultsTen studies, including 856 participants were analysed. The predictive markers of upper limb spasticity at one month post stroke were: motor 11.25 (odds ratio, OR); [95% CI:2.48, 51.04] and sensory impairments 4.91 (OR); [1.24, 19.46]; haemorrhagic stroke 3.70 (OR); [1.05, 12.98] and age 0.01 (OR) [0.00, 69.89]. Only motor impairment was found as a significant predictor at six months post stroke 30.68 (OR); [1.60, 587.06].LimitationsLow number of studies exploring biomechanical and neurophysiological in addition to behavioural predictors of spasticity were included.Conclusion and implications of key findingsUsing the results, the identified predictive markers have potential to better inform clinical decision-making and to plan specific rehabilitation interventions by physiotherapists for stroke survivors with upper limb spasticity.
Systematic Review Registration Number PROSPERO (ID: CRD42016027642).     

Placental Elasticity as a New Non-invasive Predictive Marker of Pre-eclampsia

Point shear wave elastography is an ultrasonography technique used to evaluate tissue elasticity. We examined whether placental elasticity is useful for predicting the onset of pre-eclampsia. Two hundred twenty-one participants were divided into two groups: one group at low risk (n?=?185) and the other at high risk (n?=?36) for pre-eclampsia. The two groups were compared with respect to shear wave velocity (SWV) of the placenta. Use of SWV as a predictor of pre-eclampsia was also investigated by creating a receiver operating characteristic (ROC) curve. The ROC curve was used to set a cutoff SWV value for predicting pre-eclampsia. The SWV of the high-risk group was significantly higher than that of the low-risk group (p < 0.001). Thirteen participants developed pre-eclampsia after SWV measurements, and the SWVs of these participants were significantly higher than those of participants in who pre-eclampsia did not develop. The cutoff value and area under the ROC curve were 1.188 m/s and 0.9118, respectively. Placental elasticity was significantly increased even before the onset of pre-eclampsia onset and, thus, may be a parameter used to predict the onset of pre-eclampsia.     

Injury and the Role of Alcohol: County-Wide Emergency Room Data

The purpose of this study is to describe variables associated with injury in an emergency room (ER) sample that is representative of an entire U.S. county. A probability sample ( n = 3717) of ER patients from the county hospital, 3 of the 0 community hospitals, and the three health maintenance organization hospitals in a single Northern California county were breathalyzed and interviewed at the time of the ER visit. Injured were most likely to consume only 1 or 2 drinks within el hr of injury occurrence. Twenty-three percent reported feeling drunk at the time of the event, and of these, 45% felt the event would not have happened if they had not been drinking. Breathalyzer reading, feeling drunk at the time of the event, and quantity-frequency (Q-F) of usual drinking were found to be predictive of admission to the ER with an injury, whereas breathalyzer reading, Q-F, and being injured in someone's home were predictive of reporting drinking prior to injury. Although feeling drunk at the time of the event and usual drinking patterns are predictive of injury occurrence, drinking prior to the event may not entail large quantities of alcohol consumed, but relatively small amounts consumed in close proximity to the injury event. These alcohol consumption variables may vary, however, depending on the type, cause, and severity of injury.     

Prediction of regioselectivity and preferred order of metabolisms on CYP1A2-mediated reactions part 3: Difference in substrate specificity of human and rodent CYP1A2 and the refinement of predicting system

Differences in CYP1A2-mediated metabolisms in human, rat and mouse have been analyzed with Template of human CYP1A2 established in our previous studies (Drug Metab Pharmacokinet 31:363, 2016 and 32:229, 2017). Using more than 25 chemicals including phenanthrene, MeIQx, PhIP, caffeine and furafylline, Template for human CYP1A2 was found to be applicable for rat and mouse CYP1A2 reactions with the consideration of five distinct regional interactions: 1) Expanded use of Ring D region of pro-metabolized molecules and also of trigger molecules, 2) acceptance of secondary amino groups at Position 31 of Ring eC1, 3) overlapping of pro-metabolized and trigger molecules at Ring eC4, 4) restricted maneuvering of substrates into Bay 1 region, and 5) allowance of passage of slightly large ligands in Thin-Area. These distinction points were found to be mutual for both substrates and inhibitors. In the present study, the decision-tree for substrates entering from Thin-Area has been reevaluated in consideration of species differences in human and rodent CYP1A2 forms. As the results, five steps of verification procedures have been introduced to predict the occurrence order of the regioselective metabolisms.