Many nodules remain indeterminate in etiology after comprehensive noninvasive radiological assessment. At this point, a decision to observe, biopsy, or resect the nodule is made. This decision is usually made based on the subjective perception of probability of malignacy using clinical parameters such as patient age and cigarette-smoking history as well as the radiologic features of the nodule . Unfortunately, clinical judgment incorrectly classifies a high proportion of malignant nodules as benign [57,58]. Accordingly, there have been ongoing attempts to develop more accurate, objective methods to optimize decision making in this regard.
Computer-assisted decision analytical models may improve the management of patients with indeterminate nodules. These techniques factor in risks of biopsy and surgical resection, accuracy of biopsy results, morbidity-adjusted life expectancies of patients, and cost data to predict the optimal management strategy for an indeterminate nodule based on estimates of probability of malignancy [59-61]. Such techniques have been used to compare three different management strategies: radiological observation, immediate resection, and transtho-racic needle aspiration biopsy. These studies suggest that the most cost-effective strategy is observation when the probability of cancer is low (pCa < 0.05), surgical resection when the probability of cancer is high (pCa > 0.60), and biopsy when the probability of cancer is between 0.05 and 0.6 [34,38,62,63].
In patients with an indeterminate nodule, decision making can also be assisted by such methodologies as Bayesian analysis, artificial neural network analysis, or multivariate logistic-regression models [7,57,62-64]. When these methods are used to predict the likelihood of malignancy for a given nodule, they typically perform slightly better than human observers given the same clinical and radiologic information. However, the clinical utility and applicability of this slight degree of improvement remains to be shown.
Was this article helpful?