Chest X-Ray Diagnostic Quality Assessment: How Much Is Pixel-Wise Supervision Needed?

Chest X-ray is an important imaging method for the diagnosis of chest diseases. Chest radiograph diagnostic quality assessment is vital for the diagnosis of the disease because unqualified radiographs have negative impacts on doctors' diagnosis and thus increase the burden on patients due to the re-acquirement of the radiographs. So far no algorithms and public data sets have been developed for chest radiograph diagnostic quality assessment. Towards effective chest X-ray diagnostic quality assessment, we analyze the image characteristics of four main chest radiograph diagnostic quality issues, i.e. Scapula Overlapping Lung, Artifact, Lung Field Loss, and Clavicle Unflatness. Our experiments show that general image classification methods are not competent for the task because the detailed information used for quality assessment by radiologists cannot be fully exploited by deep CNNs and image-level annotations. Then we propose to leverage a multi-label semantic segmentation framework to find the problematic regions, and then classify the quality issues based on the results of segmentation. However, subsequent classification is often negatively affected by certain small segmentation errors. Therefore, we propose to estimate a distance map that measures the distance from a pixel to its nearest segment, and use it to force the prediction of semantic segmentation more holistic and suitable for classification. Extensive experiments validate the effectiveness of our semantic-segmentation-based solution for chest X-ray diagnostic quality assessment. However, general segmentation-based algorithms requires fine pixel-wise annotations in the era of deep learning. In order to reduce reliance on fine annotations and further validate how important pixel-wise annotations are, weak supervision for segmentation is applied, and demonstrates its ability close to that of full supervision. Finally, we present ChestX-rayQuality, a chest radiograph data set, which comprises 480 frontal-view chest radiographs with semantic segmentation annotations and four labels of quality issue. Also, other 1212 chest radiographs with limited annotations are imported to validate our algorithms and arguments on larger data set. These two data set will be made publicly available.

Arid1a regulates response to anti-angiogenic therapy in advanced hepatocellular carcinoma.

BACKGROUND & AIMS: AT-rich interaction domain 1a (Arid1a), a component of the chromatin remodeling complex, has emerged as a tumor suppressor gene. It is frequently mutated in hepatocellular carcinoma (HCC). However, it remains unknown how Arid1a suppresses HCC development and whether Arid1a deficiency could be exploited for therapy, we aimed to explore these questions. METHODS: The expression of Arid1a in human and mouse HCCs was determined by immunohistochemical (IHC) staining. Gene expression was determined by quantitative PCR, ELISA or western blotting. Arid1a knockdown HCC cell lines were established by lentiviral-based shRNA. Tumor angiogenesis was quantified based on vessel density. The regulation of angiopoietin (Ang2) expression by Arid1a was identified by chromatin immunoprecipitation (ChIP) assay. The tumor promoting function of Arid1a loss was studied with a xenograft model in nude mice and diethylnitrosamine (DEN)-induced HCC in Arid1a conditional knockout mice. The therapeutic values of Ang2 antibody and sorafenib treatment were evaluated both in vitro and in vivo. RESULTS: We demonstrate that Arid1a deficiency, occurring in advanced human HCCs, is associated with increased vessel density. Mechanistically, loss of Arid1a causes aberrant histone H3K27ac deposition at the angiopoietin-2 (Ang2) enhancer and promoter, which eventually leads to ectopic expression of Ang2 and promotes HCC development. Ang2 blockade in Arid1a-deficient HCCs significantly reduces vessel density and tumor progression. Importantly, sorafenib treatment, which suppresses H3K27 acetylation and Ang2 expression, profoundly halts the progression of Arid1a-deficient HCCs. CONCLUSIONS: Arid1a-deficiency activates Ang2-dependent angiogenesis and promotes HCC progression. Loss of Arid1a in HCCs confers sensitivity to Ang2 blockade and sorafenib treatment. LAY SUMMARY: AT-rich interaction domain 1a (Arid1a), is a tumor suppressor gene. Arid1a-deficiency promotes Ang2-dependent angiogenesis leading to hepatocellular carcinoma progression. Arid1a-deficiency also sensitizes tumors to Ang2 blockade by sorafenib treatment.

An inducible hepatocellular carcinoma model for preclinical evaluation of antiangiogenic therapy in adult mice.

The limited availability of experimental tumor models that faithfully mimic the progression of human tumors and their response to therapy remains a major bottleneck to the clinical translation and application of novel therapeutic principles. To address this challenge in hepatocellular carcinoma (HCC), one of the deadliest and most common cancers in the world, we developed and validated an inducible model of hepatocarcinogenesis in adult mice. Tumorigenesis was triggered by intravenous adenoviral delivery of Cre recombinase in transgenic mice expressing the hepatocyte-specific albumin promoter, a loxP-flanked stop cassette, and the SV40 large T-antigen (iAST). Cre recombinase-mediated excision of the stop cassette led to a transient viral hepatitis and resulted in multinodular tumorigenesis within 5 to 8 weeks. Tumor nodules with histologic characteristics of human HCC established a functional vasculature by cooption, remodeling, and angiogenic expansion of the preexisting sinusoidal liver vasculature with increasing signs of vascular immaturity during tumor progression. Treatment of mice with sorafenib rapidly resulted in the induction of vascular regression, inhibition of tumor growth, and enhanced overall survival. Vascular regression was characterized by loss of endothelial cells leaving behind avascular type IV collagen-positive empty sleeves with remaining pericytes. Sorafenib treatment led to transcriptional changes of Igf1, Id1, and cMet over time, which may reflect the emergence of potential escape mechanisms. Taken together, our results established the iAST model of inducible hepatocarcinogenesis as a robust and versatile preclinical model to study HCC progression and validate novel therapies.