Aggressive surgical management of pT4 gastric carcinoma should be limited to patients without adverse prognostic factors such as advanced nodal involvement and pancreatic invasion.

Limited studies investigated clinicopathological and prognostic significance of histologic and molecular subgroups of gastric cancer concurrently. We retrospectively enrolled 1,248 patients with gastric cancer who received radical gastrectomy with lymphadenectomy and classified these cases into the Epstein–Barr virus (EBV)‐associated and microsatellite instability (MSI)‐associated subtypes by EBV‐encoded small RNA in situ hybridization and immunohistochemical stains for DNA mismatch repair proteins, respectively. The remaining cases were categorized as the Lauren intestinal and diffuse/mixed subtypes. The clinicopathological and prognostic significance of the subtypes was examined by statistical analysis. In total, 65 (5.2%), 116 (9.3%), 496 (39.7%), 431 (34.5%) and 140 (11.2%) cases were identified as EBV‐associated, MSI‐associated, intestinal, diffuse and mixed subtypes, respectively. The EBV‐associated, MSI‐associated, intestinal and diffuse/mixed subtypes exhibited distinctive clinicopathological characteristics, including differences in age, gender, stump cancer, gastric location, tumor size, TNM stage, margin involvement, lymphatic/perineural invasion, HER2 status and recurrence pattern. The log‐rank test showed survival discrimination (p < 0.001), and the multivariate analysis identified EBV‐associated and MSI‐associated cases demonstrated better outcomes than the diffuse/mixed subtype (EBV, HR 0.464, 95% CI 0.296–0.727, p = 0.001; MSI, HR 0.590, 95% CI 0.407–0.856, p = 0.005). EBV‐associated lymphoepithelioma‐like carcinoma cases had the most favorable outcome (HR 0.138, 95% CI 0.033–0.565, p = 0.006). In different clinical groups, the subtypes exhibited survival discrepancies. The EBV‐associated and diffuse/mixed cases exhibited more favorable response to chemotherapy. In conclusion, this combined classification, in parallel with the molecular subtypes specified in the Cancer Genome Atlas study, has implications for the clinical management of gastric cancer.

The PI3K/Akt/mTOR pathway is overactivated and heat shock protein (HSP) 90 is overexpressed in common cancers. We hypothesized that targeting both pathways can kill intrahepatic cholangiocarcinoma (CCA) cells. HSP90 and PTEN protein expression was evaluated by immunohistochemical staining of samples from 78 patients with intrahepatic CCA. CCA cell lines and a thioacetamide (TAA)-induced CCA animal model were treated with NVP-AUY922 (an HSP90 inhibitor) and NVP-BEZ235 (a PI3K/mTOR inhibitor) alone or in combination.Both HSP90 overexpression and loss of PTEN were poor prognostic factors in patients with intrahepatic CCA. The combination of the HSP90 inhibitor NVP-AUY922 and the PI3K/mTOR inhibitor NVP-BEZ235 was synergistic in inducing cell death in CCA cells. A combination of NVP-AUY922 and NVP-BEZ235 caused tumor regression in CCA rat animal model. This combination not only inhibited the PI3K/Akt/mTOR pathway but also induced ROS, which may exacerbate the vicious cycle of ER stress. Our data suggest simultaneous targeting of the PI3K/mTOR and HSP pathways for CCA treatment.

Pelvic radiograph (PXR) is essential for detecting proximal femur and pelvis injuries in trauma patients, which is also the key component for trauma survey. None of the currently available algorithms can accurately detect all kinds of trauma-related radiographic findings on PXRs. Here, we show a universal algorithm can detect most types of trauma-related radiographic findings on PXRs. We develop a multiscale deep learning algorithm called PelviXNet trained with 5204 PXRs with weakly supervised point annotation. PelviXNet yields an area under the receiver operating characteristic curve (AUROC) of 0.973 (95% CI, 0.960–0.983) and an area under the precision-recall curve (AUPRC) of 0.963 (95% CI, 0.948–0.974) in the clinical population test set of 1888 PXRs. The accuracy, sensitivity, and specificity at the cutoff value are 0.924 (95% CI, 0.912–0.936), 0.908 (95% CI, 0.885–0.908), and 0.932 (95% CI, 0.919–0.946), respectively. PelviXNet demonstrates comparable performance with radiologists and orthopedics in detecting pelvic and hip fractures.

Hip and pelvic fractures are serious injuries with life-threatening complications. However, diagnostic errors of fractures in pelvic X-rays (PXRs) are very common, driving the demand for computer-aided diagnosis (CAD) solutions. A major challenge lies in the fact that fractures are localized patterns that require localized analyses. Unfortunately, the PXRs residing in hospital picture archiving and communication system do not typically specify region of interests. In this paper, we propose a two-stage hip and pelvic fracture detection method that executes localized fracture classification using weakly supervised ROI mining. The first stage uses a large capacity fully-convolutional network, i.e., deep with high levels of abstraction, in a multiple instance learning setting to automatically mine probable true positive and definite hard negative ROIs from the whole PXR in the training data. The second stage trains a smaller capacity model, i.e., shallower and more generalizable, with the mined ROIs to perform localized analyses to classify fractures. During inference, our method detects hip and pelvic fractures in one pass by chaining the probability outputs of the two stages together. We evaluate our method on 4 410 PXRs, reporting an area under the ROC curve value of 0.975, the highest among state-of-the-art fracture detection methods. Moreover, we show that our two-stage approach can perform comparably to human physicians (even outperforming emergency physicians and surgeons), in a preliminary reader study of 23 readers.

Purpose: Intrahepatic cholangiocarcinoma is a fatal primary liver cancer resulting from diagnosis at an advanced stage. Understanding the mechanisms of drug resistance and metastasis of cholangiocarcinoma may improve the disease prognosis. Enhanced aldehyde dehydrogenase (ALDH) activity is suggested to be associated with increased drug resistance and the metastasis. This study aims to investigate the roles of the ALDH isoforms in cholangiocarcinoma. Experimental Design: Aldefluor assays, RT-PCR, and Western blot analysis were used to identify the major ALDH isoforms contributing to Aldefluor activity in human cholangiocarcinoma cell lines. We manipulated isoform expression in HuCCT1 cells to elucidate the role of ALDH1A3 in the malignant progression of these cells. Finally, we used immunohistochemical staining to evaluate the clinical significance of ALDH1A3 in 77 hepatectomized cholangiocarcinoma patients and an additional 31 patients with advanced cholangiocarcinoma who received gemcitabine-based therapy. Results: ALDHhigh cholangiocarcinoma cells not only migrated faster but were more resistant to gemcitabine. Among the 19 ALDH isoforms studied, ALDH1A3 was found to be the main contributor to Aldefluor activity. In addition, we also found that knockdown of ALDH1A3 expression in HuCCT1 cells markedly reduced not only their sensitivity to gemcitabine, which might be attributed to a decreased expression of ribonucleotide reductase M1, but also their migration. Most importantly, this enzyme was also identified as an independent poor prognostic factor for patients with intrahepatic cholangiocarcinoma, as well as a prognostic biomarker of gemcitabine-treated patients. Conclusions: ALDH1A3 plays an important role in enhancing malignant behavior of cholangiocarcinoma and serves as a new therapeutic target. Clin Cancer Res; 22(16); 4225–35. ©2016 AACR.

The transcriptional co-activator Yes-associated protein (YAP) has been implicated as an oncogene and is found to promote breast cancer metastasis. However, the pro-metastatic mechanism of YAP remains unclear. Here, we demonstrated that YAP functions as a transcriptional repressor of growth differentiation factor-15 (GDF15), a divergent member of the transforming growth factor superfamily, in several breast cancer cell lines. Functionally, knockdown of YAP decreased, whereas knockdown of GDF15 increased, the metastatic potential of breast cancer cells. More than that, the reduced metastasis in YAP-depleted cells could be reversed by simultaneous knockdown of GDF15. Mechanistically, the repressive effect of YAP on GDF15 requires its transcriptional factor TEAD (TEA domain family). In addition, YAP recruits polycomb repressive complex 2 (PRC2) to tri-methylate histone H3 lysine 27 in the promoter region of GDF15. Co-immunoprecipitation experiments demonstrated that YAP and enhancer of zeste 2 PRC2 subunit (EZH2) physically interact with each other. In conclusion, our data reveal that YAP promotes metastasis of breast cancer cells by repressing GDF15 transcription and present a novel molecular mechanism underlying the pro-metastasis function of YAP oncoprotein, with the implication of a therapeutic avenue for breast cancer treatment.

As the demand for more descriptive machine learning models grows within medical imaging, bottlenecks due to data paucity will exacerbate. Thus, collecting enough large-scale data will require automated tools to harvest data/label pairs from messy and real-world datasets, such as hospital picture archiving and communication systems (PACSs). This is the focus of our work, where we present a principled data curation tool to extract multi-phase computed tomography (CT) liver studies and identify each scan's phase from a real-world and heterogenous hospital PACS dataset. Emulating a typical deployment scenario, we first obtain a set of noisy labels from our institutional partners that are text mined using simple rules from DICOM tags. We train a deep learning system, using a customized and streamlined 3D squeeze and excitation (SE) architecture, to identify non-contrast, arterial, venous, and delay phase dynamic CT liver scans, filtering out anything else, including other types of liver contrast studies. To exploit as much training data as possible, we also introduce an aggregated cross entropy loss that can learn from scans only identified as "contrast". Extensive experiments on a dataset of 43K scans of 7680 patient imaging studies demonstrate that our 3DSE architecture, armed with our aggregated loss, can achieve a mean F1 of 0.977 and can correctly harvest up to 92.7% of studies, which significantly outperforms the text-mined and standard-loss approach, and also outperforms other, and more complex, model architectures.