Figure - available from: BMC Medical Imaging
This content is subject to copyright. Terms and conditions apply.
Adenocarcinoma in situ (AIS) (a) and minimally invasive adenocarcinoma (MIA) (b) arising from a non-smoker’s lung. The MIA arising from a non-smoker’s lung shows relatively typical radiological findings with GGO at the periphery and a solid component in the center. Conversely, AIS (c) and MIA (d) arising from a smoker’s lung show relatively low density and are difficult to differentiate
Source publication
Purpose
We aimed to examine the characteristics of imaging findings of adenocarcinoma in situ (AIS) and minimally invasive adenocarcinoma (MIA) in the lungs of smokers compared with those of non-smokers.
Materials and methods
We included seven cases of AIS and 20 cases of MIA in lungs of smokers (pack-years ≥ 20) and the same number of cases of AI...
Citations
... 5,6 Other relevant works relied on histopathological images rather than CT scans, or simply regarded the adenocarcinoma as one category of NSCLC. 7 In other machine visionbased studies, many of them requested complicated data as inputs, such as multiple features. 8 Most studies only paid attention to the invasiveness or growth patterns rather than generating comprehensive knowledge on the pathological diagnosis. ...
Background
The latest international multidisciplinary histopathological classification of lung cancer indicates that a deeper study of the lung adenocarcinoma requires a comprehensive multidisciplinary platform. However, in the traditional pathological examination or previous computer‐vision‐based research, the entire lung is not considered in a comprehensive manner.
Purpose
The study aims to develop a deep learning model proposed for diagnosing the lung adenocarcinoma histopathologically based on CT scans. Instead of just classifying the lung adenocarcinoma, the pathological report should be inferred based on both the invasiveness and growth pattern of the tumors.
Methods
A self‐distillation trained multitask dense‐attention network (SD‐MdaNet) is proposed and validated based on 2412 labeled CT scans from 476 patients and 845 unlabeled scans. Inferring the pathological report is divided into two tasks, predicting the invasiveness of the lung tumor and inferring growth patterns of tumor cells in a comprehensive histopathological subtyping manner with excellent accuracy. In the proposed method, the dense‐attention module is introduced to better extract features from a small dataset in the main branch of the MdaNet. Next, task‐specific attention modules are utilized in different branches and finally integrated as a multitask model. The second task is a blend of classification and regression tasks. Thus, a specialized loss function is developed. In the proposed knowledge distillation (KD) process, the MdaNet as well as its main branch trained for solving two single tasks, respectively, are treated as multiple teachers to produce a student model. A novel KD loss function is developed to take the advantage of all the models as well as data with labels and without labels.
Results
SD‐MdaNet achieves an AUC of 98.7±0.4%$98.7\pm 0.4\%$ on invasiveness prediction, and 91.6±1.0%$91.6\pm 1.0\%$ on predominant growth pattern prediction on our dataset. Moreover, the average mean squared error in inferring growth pattern proportion reaches 0.0217±0.0019$0.0217\pm 0.0019$, and the AUC for predominant growth pattern proportion reaches 91.6±1.0%$91.6\pm 1.0\%$. The proposed SD‐MdaNet is significantly better than all other benchmarking methods (FDR<0.05$FDR<0.05$).
Conclusions
Experimental results demonstrate that the proposed SD‐MdaNet can significantly improve the performance of the lung adenocarcinoma pathological diagnosis using only CT scans. Analyses and discussions are conducted to interpret the advantages of the SD‐MdaNet.
Lung cancer is the most common cause of cancer-related deaths globally. Although smoking-related lung cancers continue to account for the majority of diagnoses, smoking rates have been decreasing for several decades. Lung cancer in individuals who have never smoked (LCINS) is estimated to be the fifth most common cause of cancer-related deaths worldwide in 2023, preferentially occurring in women and Asian populations. As smoking rates continue to decline, understanding the aetiology and features of this disease, which necessitate unique diagnostic and treatment paradigms, will be imperative. New data have provided important insights into the molecular and genomic characteristics of LCINS, which are distinct from those of smoking-associated lung cancers and directly affect treatment decisions and outcomes. Herein, we review the emerging data regarding the aetiology and features of LCINS, particularly the genetic and environmental underpinnings of this disease as well as their implications for treatment. In addition, we outline the unique diagnostic and therapeutic paradigms of LCINS and discuss future directions in identifying individuals at high risk of this disease for potential screening efforts.
Background:
In addition to CT images and pathological features, many other molecular characteristics remain unknown about multiple primary lung cancer (MPLC) from intrapulmonary metastatic lung cancer.
Case presentation:
In this study, we reported a patient with an early-stage MPLC with both adenocarcinoma in situ (AIS) subtype and minimally invasive adenocarcinoma (MIA) subtype. The patient was diagnosed with more than 10 nodules and underwent precise surgery assisted by three-dimensional (3D) reconstruction at the left upper lung lobe. Whole-exome sequencing (WES) and multiple immunohistochemistry (mIHC) were performed to reveal the genomic profiling and tumor microenvironments of multiple nodules in this patient with MPLC. Based on 3D reconstruction location information, we found that the genomic and pathological results of adjacent lymph nodes were quite different. On the other hand, PD-L1 expression and the proportion of infiltrating lymphocytes in tumor microenvironments were all at a low status and did not vary in adjacent lymph nodes. Additionally, maximum diameter and tumor mutational burden levels were found to be significantly associated with CD8+ T cell proportion (p<0.05). Besides, CD163+ macrophages and CD4+ T cell proportion were higher in MIA nodules than in AIS nodules (p<0.05). This patient reached a recurrence-free survival of 39 months.
Conclusion:
Generally, in addition to CT imaging and pathological results, genomic profiling and tumor microenvironments may facilitate identifying the potential molecular mechanisms and clinical outcomes in patients with early-stage MPLC.
