Lung Neoplasm
Conditions
Keywords
lung neoplasm nodule image processing computer-assisted
Brief summary
A new software product takes two chest radiographs, aligns them, and then subtracts one image from the other. The resulting image represents an image showing any differences between them. The study is to determine whether radiologists using this new software perform better with it than when they do not use it.
Interventions
None listed
Sponsors
Riverain Medical Group, Miamisburg, OH
BioStat Solutions, Inc., Mt. Airy, MD
Georgetown University
Eligibility
Sex/Gender
ALL
Age
35 Years to 100 Years
Healthy volunteers
No
Inclusion criteria
For Radiologists: American Board of Radiology Certification and live within the Baltimore, MD-Washington, DC Metropolitan areas For chest radiographs, evidence of the presence or absence of lung nodule confirmed by expert panel; adequate image quality \-
Exclusion criteria
Radiologists who assisted by providing cases for review For chest radiographs: poor image quality
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Localized Receiver Operating Characteristic (LROC) Comparison | 1 day | The area under the LROC curve will be compared for the chest radiograph interpretations done without the new software and those done with the new software. Improvement will be demonstrated if the improvement with the new software is statistically significant at the p=\<0.05. There were 422 cases in the total study. 20 of these were inserted as noise cases, not to be analyzed. Thus there were 402 cases to be analyzed. There were 120 cases with nodules and 282 without a nodule. LROC is a method for measuring the success or failure of a method where there is a tradeoff between the detection of lung nodules that are there (true positives) and the detection that the radiologist considers to be a nodule where no nodule is present (false positive). It yields a single number that done not have a unit of measurement. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Sensitivity and Specificity | 1 day | Sensitivity and specificity will be measured. If the radiologists using the new software have higher sensitivity, statistically significant at the p=\< 0.05, the use of the new software will be considered to have resulted in improvement. A decrease in specificity is expected. |
| False Positive Decisions of Radiologists | 1 day | This is a comparison of the radiologists working without and with the software. The false positive rate is the percentage of cases in which the radiologists identified a lesions/location suspected of being cancer at a location where cancer was not present. . A false positive represents a location selected on a chest image without cancer and, also, a mark on a chest image where cancer was present, but a different location, one without cancer, was marked.The radiologists could mark up to five locations on an image and had to provide a confidence rating for each. This analysis is of the single mark with the highest confidence level. |
Countries
United States
Participant flow
Recruitment details
Recruited by email 15 radiologists practicing in the DC Metro area. Used prior subjects and those referred by prior research subjects.
Pre-assignment details
discussion of project was done with each participant. Any questions were answered and signed consent was obtained. There were no dropouts of those recruited.
Participants by arm
| Arm | Count |
|---|---|
| Radiologists Board Certified Radiologists working in the Washington, DC, Baltimore region | 15 |
| Total | 15 |
Baseline characteristics
| Characteristic | Radiologists |
|---|---|
| Age, Categorical <=18 years | 0 Participants |
| Age, Categorical >=65 years | 0 Participants |
| Age, Categorical Between 18 and 65 years | 15 Participants |
| Region of Enrollment United States | 15 participants |
| Sex: Female, Male Female | 5 Participants |
| Sex: Female, Male Male | 10 Participants |
Adverse events
| Event type | EG000 affected / at risk |
|---|---|
| deaths Total, all-cause mortality | — / — |
| other Total, other adverse events | 0 / 15 |
| serious Total, serious adverse events | 0 / 15 |
Outcome results
Primary
Localized Receiver Operating Characteristic (LROC) Comparison
The area under the LROC curve will be compared for the chest radiograph interpretations done without the new software and those done with the new software. Improvement will be demonstrated if the improvement with the new software is statistically significant at the p=\<0.05. There were 422 cases in the total study. 20 of these were inserted as noise cases, not to be analyzed. Thus there were 402 cases to be analyzed. There were 120 cases with nodules and 282 without a nodule. LROC is a method for measuring the success or failure of a method where there is a tradeoff between the detection of lung nodules that are there (true positives) and the detection that the radiologist considers to be a nodule where no nodule is present (false positive). It yields a single number that done not have a unit of measurement.
Time frame: 1 day
Population: All participants were included for calculation of A-LROC and Sensitivity-Specificity. All radiographs were interpreted both without and with software assistance
| Arm | Measure | Value (MEAN) | Dispersion |
|---|---|---|---|
| Board Certified Radiologists Working Without Software | Localized Receiver Operating Characteristic (LROC) Comparison | 0.477 unitless | Standard Error 0.037 |
| Radiologists Working With Software | Localized Receiver Operating Characteristic (LROC) Comparison | 0.536 unitless | Standard Error 0.037 |
Comparison: Measure is the difference between the radiologists working without the software less the value for the radiologists working with the software. Thus a negative value would indicate that the the radiologists showed better results when using the software.p-value: <0.0595% CI: [-0.086, -0.031]mixed model:Dorfman, Berbaum, Metz
Secondary
False Positive Decisions of Radiologists
This is a comparison of the radiologists working without and with the software. The false positive rate is the percentage of cases in which the radiologists identified a lesions/location suspected of being cancer at a location where cancer was not present. . A false positive represents a location selected on a chest image without cancer and, also, a mark on a chest image where cancer was present, but a different location, one without cancer, was marked.The radiologists could mark up to five locations on an image and had to provide a confidence rating for each. This analysis is of the single mark with the highest confidence level.
Time frame: 1 day
| Arm | Measure | Value (MEAN) | Dispersion |
|---|---|---|---|
| Board Certified Radiologists Working Without Software | False Positive Decisions of Radiologists | 9.7 percentage of marks not on cancers | Standard Error 1.9 |
| Radiologists Working With Software | False Positive Decisions of Radiologists | 11.2 percentage of marks not on cancers | Standard Error 1.9 |
Secondary
Sensitivity and Specificity
Sensitivity and specificity will be measured. If the radiologists using the new software have higher sensitivity, statistically significant at the p=\< 0.05, the use of the new software will be considered to have resulted in improvement. A decrease in specificity is expected.
Time frame: 1 day
| Arm | Measure | Group | Value (MEAN) | Dispersion |
|---|---|---|---|---|
| Board Certified Radiologists Working Without Software | Sensitivity and Specificity | Sensitivity: correct detection of cancer | 43.8 Percentage of cases | Standard Error 0.36 |
| Board Certified Radiologists Working Without Software | Sensitivity and Specificity | Specificity: cancer free cases correctly identifie | 94.1 Percentage of cases | Standard Error 0.12 |
| Radiologists Working With Software | Sensitivity and Specificity | Sensitivity: correct detection of cancer | 50.1 Percentage of cases | Standard Error 0.36 |
| Radiologists Working With Software | Sensitivity and Specificity | Specificity: cancer free cases correctly identifie | 92.6 Percentage of cases | Standard Error 0.12 |