Nonspecific benign pathological results on computed tomography-guided lung biopsy: A predictive model of true negatives

Machine-Learning-Based Classification Model to Address Diagnostic Challenges in Transbronchial Lung Biopsy

BACKGROUND

When obtaining specimens from pulmonary nodules in TBLB, distinguishing between benign samples and mis-sampling from a tumor presents a challenge. Our objective is to develop a machine-learning-based classifier for TBLB specimens.

METHODS

Three pathologists assessed six pathological findings, including interface bronchitis/bronchiolitis (IB/B), plasma cell infiltration (PLC), eosinophil infiltration (Eo), lymphoid aggregation (Ly), fibroelastosis (FE), and organizing pneumonia (OP), as potential histologic markers to distinguish between benign and malignant conditions. A total of 251 TBLB cases with defined benign and malignant outcomes based on clinical follow-up were collected and a gradient-boosted decision-tree-based machine learning model (XGBoost) was trained and tested on randomly split training and test sets.

RESULTS

Five pathological changes showed independent, mild-to-moderate associations (AUC ranging from 0.58 to 0.75) with benign conditions, with IB/B being the strongest predictor. On the other hand, FE emerged to be the sole indicator of malignant conditions with a mild association (AUC = 0.66). Our model was trained on 200 cases and tested on 51 cases, achieving an AUC of 0.78 for the binary classification of benign vs. malignant on the test set.

CONCLUSION

The machine-learning model developed has the potential to distinguish between benign and malignant conditions in TBLB samples excluding the presence or absence of tumor cells, thereby improving diagnostic accuracy and reducing the burden of repeated sampling procedures for patients.

Predictive model for the probability of malignancy in solitary pulmonary nodules: a meta-analysis

BACKGROUND

To date, multiple predictive models have been developed with the goal of reliably differentiating between solitary pulmonary nodules (SPNs) that are malignant and those that are benign. The present meta-analysis was conducted to assess the diagnostic utility of these predictive models in the context of SPN differential diagnosis.

METHODS

The PubMed, Embase, Cochrane Library, CNKI, Wanfang, and VIP databases were searched for relevant studies published through August 31, 2021. Pooled data analyses were conducted using Stata v12.0.

RESULTS

In total, 20 retrospective studies that included 5171 SPNs (malignant/benign: 3662/1509) were incorporated into this meta-analysis. Respective pooled sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), and diagnostic score values were 88% (95CI%: 0.84-0.91), 78% (95CI%: 0.74-0.80), 3.91 (95CI%: 3.42-4.46), 0.16 (95CI%: 0.12-0.21), and 3.21 (95CI%: 2.87-3.55), with an area under the summary receiver operating characteristic curve value of 86% (95CI%: 0.83-0.89). Significant heterogeneity among studies was detected with respect to sensitivity (I² = 89.07%), NLR (I² = 87.29%), and diagnostic score (I² = 72.28%). In a meta-regression analysis, sensitivity was found to be impacted by the standard reference in a given study (surgery and biopsy vs. surgery only, P = 0.02), while specificity was impacted by whether studies were blinded (yes vs. unclear, P = 0.01). Sensitivity values were higher when surgery and biopsy samples were used as a standard reference, while unclear blinding status was associated with increased specificity. No significant evidence of publication bias was detected for the present meta-analysis (P = 0.539).

CONCLUSIONS

The results of this meta-analysis demonstrate that predictive models can offer significant diagnostic utility when establishing whether SPNs are malignant or benign.

Calcification sign for prediction of benignity in pulmonary nodules: A meta-analysis

BACKGROUND

The calcification sign assessed by computed tomography appears to be a potential marker for benignities among patients diagnosed with pulmonary nodules (PNs). The following meta-analysis has been purposefully designed to figure-out the diagnostic value of the calcification signature as a way of identifying benignities from PNs.

METHODS

Cochrane Library, Embase and PubMed were considered as a reference to obtain the required data from January 2000 until October 2020. Stata v12.0 was used as a standard tool for statistical assessment.

RESULTS

Eleven retrospective studies were assessed via this meta-analysis, which included 6136 PNs (1827 benign and 4309 malignant). The pooled diagnostic odd ratios, positive likelihood ratio (PLR), negative likelihood ratio (NLR), sensitivity and specificity were 6.79, 6.06, 0.89, 13% and 98%, respectively. The value obtained for the area under the curve was 0.65, showing moderate overall diagnostic accuracy. A significant heterogeneity was found while calculating the pooled sensitivity (I²  = 85.5%), specificity (I²  = 75.0%), PLR (I²  = 59.0%), NLR (I²  = 79.5%) and DOR (I²  = 100.0%) in the current analysis. Sub-group analyses presented better PLR and specificity values for the study with a sample size ≥ 400. Deeks' funnel plot asymmetry test detected no potential evidence of significant publication bias (p = 0.091).

CONCLUSIONS

Calcification signs have been identified as moderate regulators corresponding to overall diagnostic performance (via marking a distinct differentiation between malignant and benign) for PNs. However, the manifestation of the calcification sign had a good directive property for benign PNs.