Artificial Neural Network and HCC features
Tumor histological features of hepatocellular carcinoma (HCC) strongly influence not only recurrence but also patient survival: nuclear grading and vascular invasion are considered the two most important prognostic factors but they are accurately assessable only on the pathological examination of surgical specimens and only a minority of cirrhotic patients with HCC are eligible for resection or transplantation. Percutaneous fine-needle aspiration (FNA) biopsy can be used to characterize HCC grade but its usefulness remains controversial as the heterogeneity of the histological grade, within a single tumor, may limit the overall accuracy; in addition, it can not assess microscopic vascular invasion (MVI).Tumor size and number of lesions, assessed on surgical pathologic specimens, as well as serum alfa-fetoprotein (AFP) are the variables that were most frequently found to be associated to both tumor grade and vascular invasion. In addition, given the strict relationship reported between tumor grade and microscopic vascular invasion, histological grade is currently used as a surrogate marker of MVI. However, the interaction among these factors is complex and are known to interact in a non-linear fashion, thus, making it difficult to distinguish between classes when using the conventional linear discriminant analysis. When classes are separated by a non-linear boundary, as in the present situation, an artificial neural network (ANN) has been demonstrated to perform better than conventional discriminant analysis. Read more »
What is an Artificial Neural Network?
Artificial neural network uses computer technology to model a biologic neural system both structurally and functionally. Like its biologic counterpart, an artificial neural network consists of a set of highly interconnected processing units (neurons) tied together with weighted connections. The network itself consists of an input layer, an output layer and one or more hidden layers. The input layer comprises the data available for the analysis (e.g. various laboratory tests) and the output layer comprises the outcome (e.g. diagnosis). One of the basic characteristics of the ANN is that it learns through examples: learning is achieved through exposure of paired input-output data (training). An ANN learns to associate each input with the corresponding output, by modifying the weights of the connections between neurons. Once an input has been applied as a stimulus to the first layer of neurons it is propagated through each upper layer until an output is generated. This output pattern is then compared to the desired output and an error signal is generated: the error signal is then transmitted backwards across the net and the connection weight between neurons is updated in order to decrease the overall error of the network. As learning proceeds, the error between ANN output and the desired output decreases until a minimum is reached.
Data necessary to ANN:
of clinical, radiographic or intraoperative evidence of
extra-hepatic disease or involvement of a major branch of the portal
veins and attempt of curative surgery;
2) absence of preoperative treatment such as transarterial-chemoembolization, radiofrequency ablation or percutaneous ethanol injection and absence of previous systemic chemotherapy;
3) availability of computed tomography scan (CT) or magnetic resonance imaging (MRI) not earlier than 3 months before surgery.
For further informations please visit http://www.jhep-elsevier.com; the ANN requires Microsoft Excel 2000 (or later); macro option must be activated.The software is self-installing:
Logistic regression Model
Logistic regression model showed a significant lower accuracy (P<0.001) in predicting both tumor grade and presence of microscopic vascular invasion in comparison to arificial neural network; however, the logistic regression model correctly identifies 81% of tumor grades and 85% of MVI.
Proposal for a multicentric study
The pilot study showed a very good accuracy of the ANN in the prediction of HCC tumor grade and MVI, superior to that of conventional discriminant analysis. The limitation of the pilot study, and thus the proposal for a multicentric study, is that the ANN of the was built and tested on an internal cohort and it could thus be argued that data originating from other centers may lead to a wrong diagnosis: however, we feel that this should not be considered a limitation since the distinctive characteristic of the artificial neural networks is that they can learn through examples making the prediction of histological features feasible on datasets never seen before.
Statistical analysis: development of an ANN and comparison of performance with that of a linear model; ANN is built on a training group including a cross validation, necessary since neural networks can be over-trained to recognize specific cases in a training set rather than learning general predictive characteristics; linear model is constructed on the basis of multivariate logistic regression coefficients of predictors of tumor grade on the basis of the following formula p = 1 / (1+e –logit(p)) were logit (p) = costant + a*b0 + b*b1 + c*b2 + n*bn. The performance of ANN and linear model in predicting tumor grade and microscopic vascular invasion is tested on the training group using receiver operating characteristic (ROC) curve analysis and expressed in terms of overall accuracy (sum of correct predictions divided by total predictions), positive predictive value (PPV) and log-likelihood ratio (PLR). Agreement between ANN prediction and surgical pathology in the testing group is reported using Cohen’s k coefficient: (Pr(a) – Pr(e)) / (1 – Pr(e)) where Pr(a) is the relative observed agreement and Pr(e) is the proportion of agreement expected to occur by chance alone. Agreement is considered excellent if k is >0.80, good if k ranges from 0.60 to 0.80, fair if k ranges from 0.40 to 0.60 and poor if it is <0.39.
Contact BLOG member: Dott. Cucchetti MD, Fax:(+39) 051 6363721; e-mail: firstname.lastname@example.org
Copyright 2010. Bologna Liver Oncology Group (BLOG) - Ospedale S.Orsola - Malpighi - Università di Bologna - Alma Mater Studiorum - ITALY