Virtual Library
Start Your Search
S. Lu
Author of
-
+
P2.05 - Early Stage NSCLC (ID 706)
- Event: WCLC 2017
- Type: Poster Session with Presenters Present
- Track: Early Stage NSCLC
- Presentations: 1
- Moderators:
- Coordinates: 10/17/2017, 09:30 - 16:00, Exhibit Hall (Hall B + C)
-
+
P2.05-021 - Occult Nodal Metastasis Following Lobectomy for Clinical Stage I Lung Adenocarcinoma: Implications for Sublobar Resection (ID 9436)
09:30 - 09:30 | Author(s): S. Lu
- Abstract
Background:
We investigated the incidence and location of occult nodal metastasis (ONM) in patients who had undergone lobectomy and lymph node dissection for clinical stage I lung adenocarcinoma (ADC). We performed a risk regression analysis to identify any associated radiologic and pathologic factors.
Method:
Clinical stage I lung ADC patients (stage II and III were excluded by CT and FDG-PET/CT scans) who underwent lobectomy and systematic lymph node dissection (N=715, 2005-2011) were included in the analysis. ONM were defined as pathologically diagnosed metastatic lymph nodes that are not suspected to be involved by cancer on both CT and PET scans.
Result:
Among 715 patients, 75 (10.5%) ONM were identified: 64 (85%) hilar or peribronchial and 32 (43%) mediastinal. Multivariable risk regression analysis identified tumor diameter, SUVmax, and lymphovascular invasion as risk factors (P<0.01). The incidence of subcarinal lymph node (LN) metastasis was very low among patients whose primary tumors were in the right upper lobe or left upper division (N=1/439, 0.2%). Lower mediastinal LN metastasis was rarely identified only when the primary tumor was located in the right lower or left lower lobe (N=2/210, 1.0%).
Conclusion:
One in ten patients with clinical stage I lung adenocarcinoma showed occult nodal metastases, with the highest incidence in hilar lymph nodes; this observation may be relevant for clinicians when considering sublobar resection for these patients. Figure 1
-
+
P3.13 - Radiology/Staging/Screening (ID 729)
- Event: WCLC 2017
- Type: Poster Session with Presenters Present
- Track: Radiology/Staging/Screening
- Presentations: 1
- Moderators:
- Coordinates: 10/18/2017, 09:30 - 16:00, Exhibit Hall (Hall B + C)
-
+
P3.13-037 - Deep Learning System for Lung Nodule Detection (ID 9503)
09:30 - 09:30 | Author(s): S. Lu
- Abstract
Background:
Ever since its first success in large scale image recognition problem, deep convolutional neural networks (DCNNs) have shown their capabilities in solving many challenging visual perceptual tasks, such as image classification, segmentation, object detection. In some cases, DCNNs have already achieved near-human performance. In medical image analysis, DCNNs have also been successfully applied to lesion detection, segmentation, and diagnosis. The power of DCNN lies in its ability to learn a hierarchical representation of raw input data, without hand-crafted features.
Method:
The focus of this study was to improve the performance of DCNN in automatic detection of pulmonary nodule on CT scan. In particular, nodules whose size is less than 4mm were considered. A total of 171 scans were collected at Zhongshan Hospital Fudan University, which was first process by the proposed DCNN system (12Sigma). The detection results were then carefully reviewed by an experienced physician, all false positives and missed nodules were manually labeled. To refine the system, all false positives and true nodules of sizes 4mm and under were selected.
Result:
The data set was randomly split into training and test sets, where the training set consists of 90% (154 scans) and the test set consists of 10% of data (17 scans). Figure 1 shows the FROC curves on the test set before and after re-training. Overall, the detection sensitivity were improved at all false positive levels, but the improvement was most significant at low FP rate region. For example, when FP is 0.5 per scan, the detection sensitivity increased from 0.36 to 0.49. Figure 1
Conclusion:
This improvement suggests that, even with limited data, the deep neural network can learn from its mistakes and be easily tuned to be more sensitive to small nodules. We believe that when more data is colllected, more significant improve in high FP rate region will be observed.
