AI-driven Characterization of Solid Pulmonary Nodules on CT Imaging for Enhanced Malignancy Prediction in Small-sized Lung Adenocarcinoma

OBJECTIVES

Distinguishing solid nodules from nodules with ground-glass lesions in lung cancer is a critical diagnostic challenge, especially for tumors ≤2 cm. Human assessment of these nodules is associated with high inter-observer variability, which is why an objective and reliable diagnostic tool is necessary. This study focuses on artificial intelligence (AI) to automatically analyze such tumors and to develop prospective AI systems that can independently differentiate highly malignant nodules.

MATERIALS AND METHODS

Our retrospective study analyzed 246 patients who were diagnosed with negative clinical lymph node metastases (cN0) using positron emission tomography-computed tomography (PET/CT) imaging and underwent surgical resection for lung adenocarcinoma. AI detected tumor sizes ≤2 cm in these patients. By utilizing AI to classify these nodules as solid (AI_solid) or non-solid (non-AI_solid) based on confidence scores, we aim to correlate AI determinations with pathological findings, thereby advancing the precision of preoperative assessments.

RESULTS

Solid nodules identified by AI with a confidence score ≥0.87 showed significantly higher solid component volumes and proportions in patients with AI_solid than in those with non-AI_solid, with no differences in overall diameter or total volume of the tumors. Among patients with AI_solid, 16% demonstrated lymph node metastasis, and a significant 94% harbored invasive adenocarcinoma. Additionally, 44% were upstaging postoperatively. These AI_solid nodules represented high-grade malignancies.

CONCLUSION

In small-sized lung cancer diagnosed as cN0, AI automatically identifies tumors as solid nodules ≤2 cm and evaluates their malignancy preoperatively. The AI classification can inform lymph node assessment necessity in sublobar resections, reflecting metastatic potential.

Preoperative evaluation of visceral pleural invasion in peripheral lung cancer utilizing deep learning technology

PURPOSE

This study aimed to assess the efficiency of artificial intelligence (AI) in the detection of visceral pleural invasion (VPI) of lung cancer using high-resolution computed tomography (HRCT) images, which is challenging for experts because of its significance in T-classification and lymph node metastasis prediction.

METHODS

This retrospective analysis was conducted on preoperative HRCT images of 472 patients with stage I non-small cell lung cancer (NSCLC), focusing on lesions adjacent to the pleura to predict VPI. YOLOv4.0 was utilized for tumor localization, and EfficientNetv2 was applied for VPI prediction with HRCT images meticulously annotated for AI model training and validation.

RESULTS

Of the 472 lung cancer cases (500 CT images) studied, the AI algorithm successfully identified tumors, with YOLOv4.0 accurately localizing tumors in 98% of the test images. In the EfficientNet v2-M analysis, the receiver operating characteristic curve exhibited an area under the curve of 0.78. It demonstrated powerful diagnostic performance with a sensitivity, specificity, and precision of 76.4% in VPI prediction.

CONCLUSION

AI is a promising tool for improving the diagnostic accuracy of VPI for NSCLC. Furthermore, incorporating AI into the diagnostic workflow is advocated because of its potential to improve the accuracy of preoperative diagnosis and patient outcomes in NSCLC.

A spatio-temporal image analysis for growth of indeterminate pulmonary nodules detected by CT scan

The objective is to evaluate the performance of computational image classification for indeterminate pulmonary nodules (IPN) chronologically detected by CT scan. Total 483 patients with 670 abnormal pulmonary nodules, who were taken chest thin-section CT (TSCT) images at least twice and resected as suspicious nodules in our hospital, were enrolled in this study. Nodular regions from the initial and the latest TSCT images were cut manually for each case, and approached by Python development environment, using the open-source cv2 library, to measure the nodular change rate (NCR). These NCRs were statistically compared with clinico-pathological factors, and then, this discriminator was evaluated for clinical performance. NCR showed significant differences among the nodular consistencies. In terms of histological subtypes, NCR of invasive adenocarcinoma (ADC) were significantly distinguishable from other lesions, but not from minimally invasive ADC. Only for cancers, NCR was significantly associated with loco-regional invasivity, p53-immunoreactivity, and Ki67-immunoreactivity. Regarding Epidermal Growth Factor Receptor gene mutation of ADC-related nodules, NCR showed a significant negative correlation. On staging of lung cancer cases, NCR was significantly increased with progression from pTis-stage 0 up to pT1b-stage IA2. For clinical shared decision-making (SDM) whether urgent resection or watchful-waiting, receiver operating characteristic (ROC) analysis showed that area under the ROC curve was 0.686. For small-sized IPN detected by CT scan, this approach shows promise as a potential navigator to improve work-up for life-threatening cancer screening and assist SDM before surgery.

Deep learning-powered 3D segmentation derives factors associated with lymphovascular invasion and prognosis in clinical T1 stage non-small cell lung cancer

Background

Lymphovascular invasion (LVI) is an invasive biologic behavior that affects the treatment and prognosis of patients with early-stage lung cancer. This study aimed to identify LVI diagnostic and prognostic biomarkers using deep learning-powered 3D segmentation with artificial intelligence (AI) technology.

Methods

Between January 2016 and October 2021, we enrolled patients with clinical T1 stage non-small cell lung cancer (NSCLC). We used commercially available AI software (Dr. Wise system, Deep-wise Corporation, China) to extract quantitative AI features of pulmonary nodules automatically. Dimensionality reduction was achieved through least absolute shrinkage and selection operator regression; subsequently, the AI score was calculated.Then, the univariate and multivariate analysis was further performed on the AI score and patient baseline parameters.

Results

Among 175 enrolled patients, 22 tested positive for LVI at pathology review. Based on the multivariate logistic regression results, we incorporated the AI score, carcinoembryonic antigen, spiculation, and pleural indentation into the nomogram for predicting LVI. The nomogram showed good discrimination (C-index = 0.915 [95% confidence interval: 0.89-0.94]); moreover, calibration of the nomogram revealed good predictive ability (Brier score = 0.072). Kaplan-Meier analysis revealed that relapse-free survival and overall survival were significantly higher among patients with a low-risk AI score and without LVI than those among patients with a high-risk AI score (p = 0.008 and p = 0.002, respectively) and with LVI (p = 0.013 and p = 0.008, respectively).

Conclusions

Our findings indicate that a high-risk AI score is a diagnostic biomarker for LVI in patients with clinical T1 stage NSCLC; accordingly, it can serve as a prognostic biomarker for these patients.

Artificial intelligence-based radiomics for the prediction of nodal metastasis in early-stage lung cancer

We aimed to investigate the value of computed tomography (CT)-based radiomics with artificial intelligence (AI) in predicting pathological lymph node metastasis (pN) in patients with clinical stage 0-IA non-small cell lung cancer (c-stage 0-IA NSCLC). This study enrolled 720 patients who underwent complete surgical resection for c-stage 0-IA NSCLC, and were assigned to the derivation and validation cohorts. Using the AI software Beta Version (Fujifilm Corporation, Japan), 39 AI imaging factors, including 17 factors from the AI ground-glass nodule analysis and 22 radiomics features from nodule characterization analysis, were extracted to identify factors associated with pN. Multivariate analysis showed that clinical stage IA3 (p = 0.028), solid-part size (p < 0.001), and average solid CT value (p = 0.033) were independently associated with pN. The receiver operating characteristic analysis showed that the area under the curve and optimal cut-off values of the average solid CT value relevant to pN were 0.761 and -103 Hounsfield units, and the threshold provided sensitivity, specificity, and negative predictive values of 69%, 65%, and 94% in the entire cohort, respectively. Measuring the average solid-CT value of tumors for pN may have broad applications such as guiding individualized surgical approaches and postoperative treatment.

Radiomics with Artificial Intelligence for the Prediction of Early Recurrence in Patients with Clinical Stage IA Lung Cancer

BACKGROUND

We seek to explore the ability of computed tomography (CT)-based radiomics coupled with artificial intelligence (AI) to predict early recurrence (< 2 years after surgery) in patients with clinical stage 0-IA non-small cell lung cancer (c-stage 0-IA NSCLC).

PATIENTS AND METHODS

Data of 642 patients were collected for early recurrence and assigned to the derivation and validation cohorts at a ratio of 2:1. Using the AI software Beta Version (Fujifilm Corporation, Japan), 39 AI imaging factors, including 17 factors from the AI ground-glass nodule analysis and 22 radiomic features from nodule characterization analysis, were extracted.

RESULTS

Multivariate analysis showed that male sex (p = 0.016), solid part size (p < 0.001), CT value standard deviation (p = 0.038), solid part volume ratio (p = 0.016), and bronchus translucency (p = 0.007) were associated with recurrence-free survival (RFS). Receiver operating characteristics analysis showed that the area under the curve and optimal cutoff values relevant to recurrence were 0.707 and 1.49 cm for solid part size, and 0.710 and 22.9% for solid part volume ratio, respectively. The 5-year RFS rates for patients in the validation set with solid part size ≤ 1.49 cm and > 1.49 cm were 92.2% and 70.4% (p < 0.001), whereas those for patients with solid part volume ratios ≤ 22.9% and > 22.9% were 97.8% and 71.7% (p < 0.001), respectively.

CONCLUSIONS

CT-based radiomics coupled with AI contributes to the noninvasive prediction of early recurrence in patients with c-stage 0-IA NSCLC.

Characterization of Microscopic Multicellular Foci in Grossly Normal Renal Parenchyma of Von Hippel-Lindau Kidney

Background and Objectives: This study aims to describe the earliest renal lesions in patients with von Hippel-Lindau (VHL) disease, especially the multicellular microscopic pathologic events, to get information into the genesis of renal neoplasms in this condition. Materials and Methods: Multicellular events were identified, and 3dimensional reconstruction was performed in grossly normal kidney parenchyma from VHL disease patients by using H&E-stained slides previously prepared. Results: The lesions were measured and the volume of clusters was calculated. Immunohistochemistry was performed for downstream HIF-target protein carbonic anhydrase 9 (CAIX) as well as CD34 for assessment of angiogenesis. We divided lesions into four types according to lesion height/size. The number of lesions was markedly decreased from lesion 1 (smallest) to lesion 2, then from lesions 2 to 3, and again from lesion 3 to 4. Distribution was highly consistent in the four cases, and the same decrement pattern was seen in all blocks studied. The volumes of clusters were measured and divided into three categories according to their volume. The most frequent pathologic event in VHL kidneys was category 1 (smallest volume), then category 2, and then category 3. Conclusion: We demonstrate that tracking histologic and morphologic changes in 3 dimensions of multicellular microscopic pathologic events enabled us to confirm a protracted sequence of events from smaller to larger cellular amplification events in VHL kidney.