Can digital tomosynthesis occupy a middle ground between CT and conventional radiography in lung imaging? A new study published September 8 in the Journal of Digital Imaging indicates that tomo may have a tough time finding a role in the chest that's not already well-served by the other two modalities.
Researchers from the Mayo Clinic in Rochester, MN, compared digital tomosynthesis with chest x-ray in a population of 82 individuals, while also comparing both technologies to CT as the gold standard. They found that tomo was much better than standard radiography in several areas, such as for detecting nodules larger than 4 mm -- considered the threshold for clinical significance.
But tomo didn't come close to detecting the number of lesions that CT found. And the extra time required to read tomo studies means that many facilities may stick to x-ray for basic chest applications, while sending screening studies and more complex exams to CT (September 8, 2015, JDI).
Radiography's shortcomings
Radiography has always had a number of shortcomings, but its saving grace has been its familiarity, low cost, and low radiation dose. Among radiography's most limiting characteristics is the problem of superimposed tissue that can obscure pathology.
First developed over a decade ago, digital tomosynthesis offers the promise of improved diagnostic performance by eliminating the tissue superimposition problem. Indeed, tomo is revolutionizing the field of breast imaging, with multiple studies demonstrating the technology's contribution to improved sensitivity and lower recall rates.
Tomo has developed more slowly outside of breast imaging, however. It was originally thought that tomo could occupy a middle ground between x-ray and CT for nodule detection, but the technology just hasn't been able to touch CT's sensitivity.
But what if tomo were found to be useful for characterizing previously detected lesions to see if they are cancerous?
"If you have a chest x-ray with an abnormality, is tomo a reasonable second test before going to CT?" senior author Dr. Brian Bartholmai said in an interview with AuntMinnie.com.
The Mayo team decided to test the question in a prospective, blinded study of 82 patients. No nodules were found in 11 patients, leaving the researchers with 71 patients in whom 360 nodules were detected.
Conventional radiography and tomosynthesis exams were interpreted by three radiologists who were new to tomo imaging. A fourth radiologist read the CT scans that served as the gold standard. The researchers then calculated sensitivity on a per-nodule basis, while specificity was calculated on a per-patient basis; the performance of each technique was also assessed based on lesion size and location.
For nodules considered clinically significant (larger than 4 mm), tomosynthesis had a sensitivity of 32%, compared with 17% for radiography (p < 0.001). Tomo maintained its edge with larger nodules, as indicated below.
| Sensitivity of digital tomo vs. chest radiography | ||
| Nodule size | Radiography sensitivity | Tomo sensitivity |
| ≥ 4 mm | 17% | 32% |
| ≥ 6 mm | 25% | 42% |
| ≥ 8 mm | 30% | 49% |
The researchers also analyzed whether the sensitivity of tomosynthesis varied based on the location of nodules in the lung. The modality had double the sensitivity of conventional x-ray in the right upper lobe, right middle lobe, and left lower lobe, but its sensitivity edge was lowest in the right lower lobe. Also, there was no statistically significant difference between the modalities within 3 cm of the hilum.
Regarding nodule composition, digital tomosynthesis recorded sensitivity of 41% in identifying calcified nodules larger than 4 mm, compared with 24% for chest radiography, and it had sensitivity of 30% for noncalcified nodules, compared with 16% for radiography. There was no statistically significant difference between the two techniques in correctly characterizing nodule composition -- a key factor in determining whether a nodule is malignant or benign.
The use of tomosynthesis also affected workflow. While there was no real difference between chest radiography and tomo in terms of clinical throughput, the data processing required for tomo did take a "moderate" amount of time, which is an important consideration in today's busy imaging facilities.
Finally, the researchers measured the radiation dose delivered by each of the three modalities. As expected, chest radiography had the lowest mean effective dose at 0.10 mSv, but digital tomosynthesis wasn't much higher, at a mean effective dose of 0.21 mSv. CT's mean effective dose was 6.8 mSv -- some 37 times higher than the dose from tomosynthesis.
Neither here nor there
In analyzing the results of the study, the authors noted that neither chest radiography nor tomosynthesis approached the sensitivity or specificity of CT for identifying and characterizing lung nodules. Also, tomo's edge over radiography in terms of sensitivity disappeared in nodules 1 cm and larger.
Given these factors, how would digital tomosynthesis fit into the armamentarium of the average imaging facility for chest imaging? Bartholmai said the study has left him skeptical; indeed, the tomosynthesis unit employed in the study is only used for research purposes at Mayo, not for clinical imaging.
Instead, tomosynthesis is probably better employed in other clinical applications, such as breast imaging, or for orthopedic use as a way to see around artifacts created by metal implants.
"I was surprised that CT was so much better for edges and density of nodules," Bartholmai said. "Tomo is not giving us all of the information we need in the chest."
