AR: Temporal subtraction improves chest x-ray CAD
An early trial of a CAD program improved all 12 participating radiologists’ detection of lung nodules on x-ray, while averaging 2.6 false-positives per image, offering potentially promising results for the temporal subtraction method for CAD, according to a study published in the August issue of Academic Radiology.
CAD has moved forward in radiology as a complementary tool for improving the detection of nodules and other diagnostic abnormalities. The technology’s application has been limited, however, by inordinately low levels of specificity to achieve satisfactory sensitivity.
The temporal subtraction technique of CAD, in which a patient’s previous x-ray is subtracted from a follow-up to highlight changes, has offered some promise in improving detection, although this newer method has shown a tendency to create artifacts. In addition, the requirement for at least one previous patient image poses an added limitation, offered Takatoshi Aoki, MD, from the department of radiology at the University of Occupational and Environmental Health in Kitakyushu, Japan, and co-authors.
Aoki and colleagues developed and tested a temporal subtraction CAD system using x-rays from 51 healthy individuals and 51 cases with lung nodules (17 lung carcinomas, 17 metastatic lung tumors and 17 benign nodules). Nodules were smaller than 3 cm and were interpreted by 12 radiologists, including four residents, four fellows and four attending radiologists, with and without CAD.
All radiologists showed improved detection with CAD, with an average area under the curve of 0.849 without CAD and 0.950 with the technology. The improvements were significant for each group (residents, fellows and attendings), and were most dramatic for residents.
Across a range of diagnostic confidence levels for the interpreting physicians, the average radiologist experienced 4.0 clinically relevant improvements in confidence, compared with 0.6 detrimental effects. Each interpretation averaged 2.6 false-positive findings, a number the authors indicated was lower than previous studies.
“Because of the use of this CAD system, significant improvements in diagnostic accuracy for pulmonary nodule detection were experienced by all 12 radiologists in the observer performance study,” Aoki and colleagues reflected. “Although all observer categories benefited from this CAD system, those with less experience tended to benefit more.”
The authors stressed that the system would be most practical for follow-up examinations of patients at high risk for lung tumors.
“Even though computed tomographic techniques continue to develop, chest radiography remains the first and the most common examination for the detection of lung nodules,” Aoki and colleagues stated.
“Given the increasingly digital nature of chest radiography, CAD will most certainly be an integral part of clinical practice,” the authors concluded.
CAD has moved forward in radiology as a complementary tool for improving the detection of nodules and other diagnostic abnormalities. The technology’s application has been limited, however, by inordinately low levels of specificity to achieve satisfactory sensitivity.
The temporal subtraction technique of CAD, in which a patient’s previous x-ray is subtracted from a follow-up to highlight changes, has offered some promise in improving detection, although this newer method has shown a tendency to create artifacts. In addition, the requirement for at least one previous patient image poses an added limitation, offered Takatoshi Aoki, MD, from the department of radiology at the University of Occupational and Environmental Health in Kitakyushu, Japan, and co-authors.
Aoki and colleagues developed and tested a temporal subtraction CAD system using x-rays from 51 healthy individuals and 51 cases with lung nodules (17 lung carcinomas, 17 metastatic lung tumors and 17 benign nodules). Nodules were smaller than 3 cm and were interpreted by 12 radiologists, including four residents, four fellows and four attending radiologists, with and without CAD.
All radiologists showed improved detection with CAD, with an average area under the curve of 0.849 without CAD and 0.950 with the technology. The improvements were significant for each group (residents, fellows and attendings), and were most dramatic for residents.
Across a range of diagnostic confidence levels for the interpreting physicians, the average radiologist experienced 4.0 clinically relevant improvements in confidence, compared with 0.6 detrimental effects. Each interpretation averaged 2.6 false-positive findings, a number the authors indicated was lower than previous studies.
“Because of the use of this CAD system, significant improvements in diagnostic accuracy for pulmonary nodule detection were experienced by all 12 radiologists in the observer performance study,” Aoki and colleagues reflected. “Although all observer categories benefited from this CAD system, those with less experience tended to benefit more.”
The authors stressed that the system would be most practical for follow-up examinations of patients at high risk for lung tumors.
“Even though computed tomographic techniques continue to develop, chest radiography remains the first and the most common examination for the detection of lung nodules,” Aoki and colleagues stated.
“Given the increasingly digital nature of chest radiography, CAD will most certainly be an integral part of clinical practice,” the authors concluded.