ABSTRACT
OBJECTIVE: Pneumothorax is an acute thoracic disease caused by abnormal air collection between the lungs and chest wall. Recently, artificial intelligence (AI), especially deep learning (DL), has been increasingly employed for automating the diagnostic process of pneumothorax. To address the opaqueness often associated with DL models, explainable artificial intelligence (XAI) methods have been introduced to outline regions related to pneumothorax. However, these explanations sometimes diverge from actual lesion areas, highlighting the need for further improvement. METHOD: We propose a template-guided approach to incorporate the clinical knowledge of pneumothorax into model explanations generated by XAI methods, thereby enhancing the quality of the explanations. Utilizing one lesion delineation created by radiologists, our approach first generates a template that represents potential areas of pneumothorax occurrence. This template is then superimposed on model explanations to filter out extraneous explanations that fall outside the template's boundaries. To validate its efficacy, we carried out a comparative analysis of three XAI methods (Saliency Map, Grad-CAM, and Integrated Gradients) with and without our template guidance when explaining two DL models (VGG-19 and ResNet-50) in two real-world datasets (SIIM-ACR and ChestX-Det). RESULTS: The proposed approach consistently improved baseline XAI methods across twelve benchmark scenarios built on three XAI methods, two DL models, and two datasets. The average incremental percentages, calculated by the performance improvements over the baseline performance, were 97.8% in Intersection over Union (IoU) and 94.1% in Dice Similarity Coefficient (DSC) when comparing model explanations and ground-truth lesion areas. We further visualized baseline and template-guided model explanations on radiographs to showcase the performance of our approach. CONCLUSIONS: In the context of pneumothorax diagnoses, we proposed a template-guided approach for improving model explanations. Our approach not only aligns model explanations more closely with clinical insights but also exhibits extensibility to other thoracic diseases. We anticipate that our template guidance will forge a novel approach to elucidating AI models by integrating clinical domain expertise.
ABSTRACT
Background: Osimertinib is a third-generation epidermal growth factor receptor tyrosine kinase inhibitor (EGFR-TKI) approved for use in EGFR-mutant lung cancer. We examined its performance in the second/subsequent line after resistance to first- and second-generation (1/2G) EGFR-TKI. Methods: We reviewed electronic records of 202 patients who received osimertinib from July 2015 to January 2019 in the second/subsequent line after progression on prior EGFR-TKI. Of these, complete data from 193 patients were available. Clinical data including patient characteristics, primary EGFR mutation, T790M mutation status, presence of baseline brain metastases (BM), first-line EGFR-TKI use, and survival outcomes were extracted, and results retrospectively analyzed. Results: Of 193 evaluable patients, 151 (78.2%) were T790M+ (T790M positive) with 96 (49.2%) tissue confirmed; 52% of patients received osimertinib in the second line setting. After median follow up of 37 months, median progression-free survival (PFS) of the entire cohort was 10.3 [95% confidence interval (CI): 8.64-11.50] months and median overall survival (OS) was 20 (95% CI: 15.61-23.13) months. Overall response rate (ORR) to osimertinib was 43% (95% CI: 35.9-50.3%); 48.3% in T790M+ vs. 20% in T790M- (T790M negative) patients. OS in T790M+ patients was 22.6 vs. 7.9 months in T790M- patients (HR 0.43, P=0.001), and PFS was 11.2 vs. 3.1 months respectively (HR 0.52, P=0.01). Tumour T790M+ was significantly associated with longer PFS (P=0.007) and OS (P=0.01) compared to tumour T790M- patients, however this association was not seen with plasma T790M+. Of the 22 patients with paired tumor/plasma T790M testing, response rate (RR) to osimertinib was 30% for those plasma T790M+/tumour T790M-, compared to 63% and 67% for those who were plasma T790M+/tumour T790M+ and plasma T790M-/tumour T790M+, respectively. By multivariable analysis (MVA), Eastern Cooperative Oncology Group (ECOG) performance status ≥2 was associated with shorter OS (HR 2.53, P<0.001) and PFS (HR 2.10, P<0.001), whereas presence of T790M+ was associated with longer OS (HR 0.50, P=0.008) and PFS (HR 0.57, P=0.027). Conclusions: This cohort demonstrated the efficacy of osimertinib in second line/beyond for EGFR+ (EGFR mutation-positive) non-small cell lung cancer (NSCLC). Tissue T790M result appeared more predictive of osimertinib efficacy compared to plasma, highlighting potential T790M heterogeneity and the advantage with paired tumor-plasma T790M testing at TKI resistance. T790M- disease at resistance remains an unmet treatment need.