Virtual Library
Start Your Search
Rowena Yip
Author of
-
+
P2.16 - Surgery (ID 717)
- Event: WCLC 2017
- Type: Poster Session with Presenters Present
- Track: Surgery
- Presentations: 1
- Moderators:
- Coordinates: 10/17/2017, 09:30 - 16:00, Exhibit Hall (Hall B + C)
-
+
P2.16-014 - Deconstructing Surgical Decision Making (ID 9543)
09:30 - 09:30 | Presenting Author(s): Rowena Yip
- Abstract
Background:
With the increase in number of individuals undergoing CT screening, lung cancers are now being detected at an earlier stage. Curative treatment can thus be performed on these patients, resulting in better lung cancer survival. Effective surgical decision making depends upon the degree of knowledge and experience the treating surgeon has about the outcome of actions, ability to assess risk and its subsequent impact. Use of a gnostic expert system would increase cost-effectiveness and efficiency. Our objective is to garner experts’ tacit knowledge about surgical decision making in a form of probability function.
Method:
Nine surgeons with extensive experiences in thoracic surgery were presented with a set of hypothetical cases, specified by indicators for surgical treatment (lobectomy or limited resection). Their choice of surgery and probability of performing limited resection were recorded for each case. Probabilities were translated into a logistic probability function for limited resection by 1) taking logits of the probabilities: Y=log[P/(1-P)], then 2) applying a general linear model for the mean of Y, Ŷ=β~1~+ β~2~X~2~+ β~3~X~3~+ β~4~X~4~+ β~5~X~5~+ β~6~X~6~+ β~7~X~7~ + ε. Standardized coefficients were computed and ranked to determine the effect of each indicator on limited resection.
Result:
Across the 24 cases, the median probabilities of limited resection among experts ranged from 0.0% to 100.0%, their case-specific IQR had values from 5 to 90 (Q3-Q1) percentage points, and ranges had values from 10-100(max-min) percentage points. Considering the expert-specific median probabilities, five out of eight experts favored lobectomy (median probabilities of limited resection ≤12.5%). Two other experts had median probabilities of 42.5% and 49% while the remaining expert favored limited resection (median probability 65%). The effect of each indicator on preferring limited resection over lobectomy varied between surgeons. Overall, distance from relevant pleura and nodule size were important factors for considering limited resection.
Conclusion:
There was great inter-surgeons variability on surgical decision making. Garnering experts’tacit knowledge on surgical decision making will enhance efficiency of health care and potentially change surgical practice.
-
+
P3.13 - Radiology/Staging/Screening (ID 729)
- Event: WCLC 2017
- Type: Poster Session with Presenters Present
- Track: Radiology/Staging/Screening
- Presentations: 2
- Moderators:
- Coordinates: 10/18/2017, 09:30 - 16:00, Exhibit Hall (Hall B + C)
-
+
P3.13-028 - Controversies on Lung Cancers Manifesting as Part-Solid Nodules (ID 10074)
09:30 - 09:30 | Presenting Author(s): Rowena Yip
- Abstract
Background:
Questions have been raised about the appropriate treatment of lung cancers manifesting as subsolid nodules (nonsolid nodules (NSNs) and part-solid nodules (PSNs)), as these have very high reported survival rates and have been observed in up to 10% of screening participants. Our goal in this report is to summarize the publications on survival of patients with resected lung cancers manifesting as PSNs and to further the development of consensus definitions of the CT appearance and the workup of such nodules.
Method:
PubMed/MEDLINE and EMBASE databases were searched for all studies/ clinical trials on CT-detected lung cancer in English before Dec 21, 2015 to identify surgically-resected lung cancers manifesting as PSNs. Outcome measures were lung cancer-specific survival (LCS), overall survival (OS), or disease free survival (DFS). All PSNs were classified by the percentage of solid component to the entire nodule diameter into: Category PSNs < 80% or Category PSNs ≥ 80%.
Result:
Twenty studies reported on PSNs < 80%: 7 reported DFS and 2 OS of 100%, 6 DFS 96.3-98.7%, and 11 OS 94.7-98.9% (median DFS 100% and OS 97.5%). Twenty-seven studies reported on PSNs ≥ 80%: 1 DFS and 2 OS of 100%, 19 DFS 48.0%-98.0% (median 82.6%), and 16 reported OS 43.0%-98.0% (median DFS 82.6%, OS 85.5%). Both DFS and OS were always higher for PSNs<80%.
Conclusion:
A clear definition of the upper limit of solid component of a PSN is needed to avoid misclassification because cell-types and outcomes are different for PSN and solid nodules. The workup should be based on the size of the solid component.
-
+
P3.13-035 - Automatic Estimation of Measurement Error on CT Imaging (ID 10333)
09:30 - 09:30 | Author(s): Rowena Yip
- Abstract
Background:
There has been increasing recognition that lung nodule measurement on CT scans is imprecise and that an understanding of the extent of this imprecision is necessary when trying to determine whether actual change in volume has occurred. The various factors that influence this are numerous with two of the most prominent being the overall quality of the CT scan (including all of the adjustable parameters) and the size of the nodule.
Method:
We have developed an automated system whereby a calibration device is scanned on a given scanner with a given protocol and then the system can automatically predict the extent of measurement error for a given size solid nodule. We compared this approach to empirically derived results obtained from a database of 117 screen-detected stable nodule ranging in size from 2.2 to 18.7 mm that were scanned twice on the same CT scanner using the same protocol. Automated volumetric analysis was performed using commercial software. This allowed us to determine the relationship between standard deviation of the measurements versus nodule size. We then scanned our calibration device using the same scanning protocol as was used on those nodules to automatically calculate the size and standard deviation relationship.
Result:
Predicted solid nodule volume standard deviation compared with empirically derived values across a range of nodule sizes was within 20% (see figure)Figure 1
Conclusion:
Results from our automated approach were highly correlated with results obtained from scans obtained in actual clinical practice. The ability to predict extent of error specific to a given scanner and scanning protocol is an essential step in understanding whether change has occurred and has implications for both diagnosis and therapy assessment, including predicting when a follow up scan should be obtained. This type of information will ultimately become a necessary component of all quantitative imaging programs.
