Computer-aided detection is increasingly being touted to help find colorectal polyps in CT colonography screening. Researchers from the U.K. have found that commercially available CAD software is equally as good at spotting colorectal cancer.
Computer-aided detection is increasingly being touted to help find colorectal polyps in CT colonography screening. Researchers from the U.K. have found that commercially available CAD software is equally as good at spotting colorectal cancer.
Dr. Steve Halligan and colleagues at University Hospital London reviewed CT colonography data sets from 59 patients with 59 proven cancers who were imaged in the prone and supine positions at six different centers. Patients underwent full bowel preparation.
Researchers used ColonCAD (Medicsight) software and applied four different filter settings: 0, 50, 75, and 100. Computer prompts were categorized by a single observer as either true positive if the detection boundary overlapped tumor boundary or false negative if located elsewhere.
The 0 filter setting was the most sensitive and least specific. CAD detected 54 of 59 cancers at the filter setting of 0, dropping down to 47 detected cancers at the setting of 100. The median number of false-positive CAD marks per patient was 10 at the filter setting of 0, decreasing to 3 at the setting of 100.
Of the 54 cancers detected by CAD, 22 were only detected on either the prone (16) or supine (six) data set. One-third of cancers were detected on both scans at all filter settings.
The CAD system used was developed to detect polyps. Cancer detection could possibly increase with software that is specifically developed to detect cancer, Halligan said.
"It's serendipitous that this CAD system detects cancer. Many polyps look like cancer. Polypoid cancer is more likely to be detected rather than annular," he said.
Researchers concluded that CAD is effective to detect cancer, and both prone and supine CT acquisitions are required for optimal results.
In another study, Luca Bogoni, Ph.D., and colleagues from Siemens Medical Solutions found their prototype CAD system detected polyps between 6 mm and 20 mm comparably to other systems.
Researchers tested the CAD on more than 1000 data sets, nearly evenly split between those who had complete bowel preparation and those who did not. Data came from 14 centers that used various CT scanners from different manufacturers. One-third of cases were used to train the system, and the rest were used for testing. Validation of CAD was mostly with optical colonoscopy
All data were acquired in both prone and supine positions. Overall sensitivity for the bowel-prepped cases was 88%, with an average of 2.2 false positives per volume (436 volumes). Overall sensitivity for the tagged cases was 84%, with an average of 3.3 false positives per volume (873).
"This gives confirmatory evidence that automatic polyp detection can operate at a high level of sensitivity with a low false-positive rate in a second reader role," Bogoni said.
MRI-Based AI Radiomics Model Offers 'Robust' Prediction of Perineural Invasion in Prostate Cancer
July 26th 2024A model that combines MRI-based deep learning radiomics and clinical factors demonstrated an 84.8 percent ROC AUC and a 92.6 percent precision-recall AUC for predicting perineural invasion in prostate cancer cases.
Breast MRI Study Examines Common Factors with False Negatives and False Positives
July 24th 2024The absence of ipsilateral breast hypervascularity is three times more likely to be associated with false-negative findings on breast MRI and non-mass enhancement lesions have a 4.5-fold likelihood of being linked to false-positive results, according to new research.
Can Polyenergetic Reconstruction Help Resolve Streak Artifacts in Photon Counting CT?
July 22nd 2024New research looking at photon-counting computed tomography (PCCT) demonstrated significantly reduced variation and tracheal air density attenuation with polyenergetic reconstruction in contrast to monoenergetic reconstruction on chest CT.