Methods for improving colorectal cancer annotation efficiency for artificial intelligence-observer training

BACKGROUND Missing occult cancer lesions accounts for the most diagnostic errors in retrospective radiology reviews as early cancer can be small or subtle,making the lesions difficult to detect.Secondobserver is the most effective technique for reducing these events and can be economically implemented with the advent of artificial intelligence(AI).AIM To achieve appropriate AI model training,a large annotated dataset is necessary to train the AI models.Our goal in this research is to compare two methods for decreasing the annotation time to establish ground truth:Skip-slice annotation and AI-initiated annotation.METHODS We developed a 2D U-Net as an AI second observer for detecting colorectal cancer(CRC)and an ensemble of 5 differently initiated 2D U-Net for ensemble technique.Each model was trained with 51 cases of annotated CRC computed tomography of the abdomen and pelvis,tested with 7 cases,and validated with 20 cases from The Cancer Imaging Archive cases.The sensitivity,false positives per case,and estimated Dice coefficient were obtained for each method of training.We compared the two methods of annotations and the time reduction associated with the technique.The time differences were tested using Friedman’s two-way analysis of variance.RESULTS Sparse annotation significantly reduces the time for annotation particularly skipping 2 slices at a time(P<0.001).Reduction of up to 2/3 of the annotation does not reduce AI model sensitivity or false positives per case.Although initializing human annotation with AI reduces the annotation time,the reduction is minimal,even when using an ensemble AI to decrease false positives.CONCLUSION Our data support the sparse annotation technique as an efficient technique for reducing the time needed to establish the ground truth.

Differences in characteristics of patients with and without known risk factors for hepatocellular carcinoma in the United States

AIM: To examine the clinical characteristics of a subgroup of patients with hepatocellular carcinoma (HCC) and compare them to those with known risk factors.METHODS: We used the HCC database of 306 patients seen at our institution from January 1, 1995 to December 31, 2001. Of the 306 patients, 63 (20%, group 1) had no known risk factors (hepatitis C virus, hepatitis B virus, alcohol, hemochromatosis or cirrhosis from any cause) and 243 (group 2) had one or more risk factors.RESULTS: The median age was similar in both groups, but there were disproportionate numbers of younger (< 30 years old), older (> 80 years) patients, women (33% vs 18%), and Caucasians (81% vs 52%) in group 1 as compared to group 2. There were fewer Asians (2% vs 11%) and African Americans (13% vs 27%) in group 1. Abdominal pain (70% vs 37%) was more common while gastrointestinal bleeding (0% vs 11%) and ascites (4% vs 17%) were less common in group 1 compared to group 2. Group 1 had larger tumor burden (median size 9.4 cm vs 5.7 cm) at the time of presentation, but there were no differences in the site (right, left or bilateral lesions), or number of tumors between the two groups.CONCLUSION: HCC patients without identifiable risk factors have different characteristics and clinical presentation compared to those with known risk factors.Absence of cirrhosis and larger tumor burden may explain the differences in the presenting symptoms.