Bone Suppression Software with Chest X-Ray Equivalent to Dual Energy Subtraction Imaging in the Detection of Potential Lung Cancer

New software considerably enhances the radiologists’ ability to detect potentially cancerous lung nodules in X-ray images, and may become a cost-effective option to dual energy subtraction (DES) imaging, according to two new studies.

The studies’ findings were presented at the 20th anniversary meeting of the European Society of Thoracic Imaging (ESTI), June 22-24, 2012, in London (UK). Riverain Technologies’ (Dayton, OH, USA) ClearRead bone suppression software utilizes machine learning algorithms to convert any conventional chest X-ray image into an enhanced, soft tissue image without the ribs and clavicles that sometimes obscure early lung cancer. DES also creates a soft tissue image but requires a dedicated dual energy imaging system to form it.

Depending on the methodology, DES may require two separate scans, thereby exposing patients to more radiation than conventional X-ray. Riverain’s computer-aided detection (CAD) software, ClearRead +Detect, provides additional support in decision making by circling suspicious areas on a bone-suppressed image that may be lung cancer. Final determination is made by the radiologist.

The first study, conducted at the Institute of Diagnostic, Interventional and Pediatric Radiology at the University Hospital Bern (Switzerland), compared Riverain’s bone suppression and DES alone and in combination with CAD and revealed that the bone suppression software is as good as DES at detecting lung nodules while producing superior image quality.

In the retrospective study, three radiologists independently reviewed chest images of 143 patients: 101 patients with 155 lung nodules between 5-29 mm earlier validated using CT, and 42 subjects with no lung nodules. Each radiologist tagged suspected nodules on each patient’s original chest X-ray image, and individually on DES and bone-suppressed images with and without CAD.

The radiologists detected the most lung nodules in the bone-suppressed image with CAD markings. Their mean sensitivities--the percentage of the 155 lung nodules that were accurately identified--were: 46.9% using traditional X-ray only; 49.2% using a single-shot DES system; 49.7% using SoftView 2.0 (an earlier version of Riverain’s ClearRead bone suppression); and 51.6% using SoftView 2.0 plus OnGuard 5.1 (an earlier version Riverain’s ClearRead +Detect software). The overall diagnostic performance with the modalities was not significantly different.
“These findings are compelling results for hospitals, radiology practices and patients,” said Steve Worrell, Riverain’s chief technology officer. “Radiologists detected as many lung nodules using Riverain bone suppression software on conventional X-ray images as they detected using a dedicated piece of imaging equipment that is more expensive, may expose patients to more radiation, and can only be used in the single location where it is housed. Our software immediately enhances any standard chest X-ray image, after capture, and can be used throughout entire healthcare systems without additional imaging equipment, staff or space requirements, and without any additional tests or radiation dose for patients.”

The radiologists also gave the bone-suppressed images a significantly higher overall quality rating than the DES images. The true-positive and false-positive rates of these two modalities were statistically equivalent. “Electronic bone suppression provides equivalent detection rates for lung nodules as DES, with better image quality, and might be a cost-effective alternative to DES chest radiography in the detection of lung nodules,” said Zsolt Szucs-Farkas, MD, PhD, chief investigator.

Dr. Szucs-Farkas also compared CAD markings before radiologist interpretation to the radiologists’ findings. Whereas CAD and the radiologists detected many of the same lung nodules, each also detected nodules the other did not find. CAD, on its own without any radiologist interpretation (evaluated for research purposes only), accurately circled approximately one in four nodules that the radiologists missed, and the radiologists found approximately one in three nodules that CAD missed.

“Working together, radiologists and our CAD with bone suppression software bring different strengths to the table and significantly improve the detection of nodules that may be lung cancer using conventional chest X-ray,” Dr. Worrell said.

A second study presented at ESTI 2012 also confirmed Riverain’s bone suppression software considerably improves radiologists’ ability to detect lung nodules in chest X-ray images. Eight radiologists from the Radboud University Nijmegen Medical Center (RUNMC; The Netherlands) and the Meander Medical Center (Amersfoort, The Netherlands), reviewed chest X-ray images for 108 patients with only one CT-proven lung nodule and 192 patients without nodules. On average, they found 14.4% of lung nodules using Riverain’s bone suppression technology (SoftView 2.4, now called ClearRead bone suppression) that were missed when they used conventional X-ray alone, without an increase in false-positives. All individual readers improved detection with the help of the bone suppression software. Individual reader results ranged from as low as 52% without bone suppression to a high of 81% with the software. Average detection overall was 67% using X-ray alone, and 72% with bone suppression software.

Riverain Technologies applies proprietary pattern recognition and machine-learning technologies in the creation of software applications for use globally in the healthcare industry. The company’s ClearRead software increases the expert skills of radiologists to improve patient outcomes using standard chest X-ray, without additional radiation dose or procedures for patients.

Related Links:
Riverain Technologies
Radboud University Nijmegen Medical Center
Meander Medical Center