Search

ObjectivesThe objective of this study was to develop and validate a state-of-the-art, deep learning (DL)-based model for detecting breast cancers on mammography.MethodsMammograms in a hospital development dataset, a hospital test dataset, and a clinic test dataset were retrospectively collected from January 2006 through December 2017 in Osaka City University Hospital and Medcity21 Clinic. The hospital development dataset and a publicly available digital database for screening mammography (DDSM) dataset were used to train and to validate the RetinaNet, one type of DL-based model, with five-fold cross-validation. The model's sensitivity and mean false positive indications per image (mFPI) and partial area under the curve (AUC) with 1.0 mFPI for both test datasets were externally assessed with the test datasets.ResultsThe hospital development dataset, hospital test dataset, clinic test dataset, and DDSM development dataset included a total of 3179 images (1448 malignant images), 491 images (225 malignant images), 2821 images (37 malignant images), and 1457 malignant images, respectively. The proposed model detected all cancers with a 0.45-0.47 mFPI and had partial AUCs of 0.93 in both test datasets.ConclusionsThe DL-based model developed for this study was able to detect all breast cancers with a very low mFPI. Our DL-based model achieved the highest performance to date, which might lead to improved diagnosis for breast cancer.

It was reported that the chemokine (C-X-C motif) ligand 1 (CXCL1) from cancer cells stimulated the recruitment of bone marrow-derived mesenchymal cells (BM-MCs) into tumor stroma via chemokine (C-X-C motif) receptor 2 (CXCR2) signaling. We conducted this retrospective study to determine the clinicopathologic significance of the CXCL1-CXCR2 axis in human gastric cancer.The correlations between the clinicopathological features of 270 primary gastric carcinomas and CXCL1 in cancer cells and CXCR2 in stromal cells were analyzed in immunohistochemical studies. The effect of gastric cancer cells on the expression of CXCR2 in BM-MCs was examined using diffuse-type gastric cancer cell lines in vitro.The expression of CXCL1 in cancer cells was correlated with T invasion (T2-T4), lymph node metastasis, lymphatic invasion, venous invasion, peritoneal cytology, peritoneal metastasis and CXCR2 expression in stromal cells. The expression of CXCR2 in stromal cells was correlated with macroscopic type-4 cancers, histological type, T invasion (T2-T4), lymph node metastasis, lymphatic invasion, infiltration, peritoneal cytology, peritoneal metastasis and CD271 expression in stromal cells. The overall survival of patients with CXCL1 and CXCR2-positive cancer was poorer than that of the patients with negative cancer. Both CXCL1 expression in cancer cells and CXCR2 expression in stromal cells were independent prognostic factors for gastric cancer patients.The expressions of CXCL1 in cancer cells and CXCR2 in stromal cells are useful prognostic factors for gastric cancer patients.

Pancreatic ductal adenocarcinoma (PDAC) is characterized by its hypovascularity, with an extremely poor prognosis because of its highly invasive nature. PDAC proliferates with abundant stromal cells, suggesting that its invasive activity might be controlled by intercellular interactions between cancer cells and fibroblasts. Using four PDAC cell lines and two pancreas cancer-associated fibroblasts (CAFs), the expression of insulin-like growth factor-1 (IGF1) and IGF1 receptor (IGF1R) was evaluated by RT-PCR, FACScan, western blot, or ELISA. Correlation between IGF1R and the hypoxia marker carbonic anhydrase 9 (CA9) was examined by immunohistochemical staining of 120 pancreatic specimens. The effects of CAFs, IGF1, and IGF1R inhibitors on the motility of cancer cells were examined by wound-healing assay or invasion assay under normoxia (20% O2) and hypoxia (1% O2). IGF1R expression was significantly higher in RWP-1, MiaPaCa-2, and OCUP-AT cells than in Panc-1 cells. Hypoxia increased the expression level of IGF1R in RWP-1, MiaPaCa-2, and OCUP-AT cells. CA9 expression was correlated with IGF1R expression in pancreatic specimens. CAFs produced IGF1 under hypoxia, but PDAC cells did not. A conditioned medium from CAFs, which expressed αSMA, stimulated the migration and invasion ability of MiaPaCa-2, RWP-1, and OCUP-AT cells. The motility of all PDAC cells was greater under hypoxia than under normoxia. The motility-stimulating ability of CAFs was decreased by IGF1R inhibitors. These findings might suggest that pancreas CAFs stimulate the invasion activity of PDAC cells through paracrine IGF1/IGF1R signaling, especially under hypoxia. Therefore the targeting of IGF1R signaling might represent a promising therapeutic approach in IGF1R-dependent PDAC.

Cancer stem cells (CSCs) are responsible for cancer progression, metastasis, and recurrence. To date, the specific markers of CSCs remain undiscovered. The aim of this study was to identify novel biomarkers of gastric CSCs for clinical diagnosis using proteomics technology. CSC-like SP cells, OCUM-12/SP cells, OCUM-2MD3/SP cells, and their parent OCUM-12 cells and OCUM-2MD3 cells were used in this study. Protein lysates from each cell line were analyzed using QSTAR Elite Liquid Chromatography with Tandem Mass Spectrometry, coupled with isobaric tags for relative and absolute quantitation technology. Candidate proteins detected by proteomics technology were validated by immunohistochemical analysis of 300 gastric cancers. Based on the results of LC-MS/MS, eight proteins, including RBBP6, GLG1, VPS13A, DCTPP1, HSPA9, HSPA4, ALDOA, and KRT18, were up-regulated in both OCUM-12/SP cells and OCUM-2MD3/SP cells when compared to their corresponding parent cells. RT-PCR analysis indicated that the expression level of RBBP6, HSPA4, DCTPP1, HSPA9, VPS13A, ALDOA, GLG1, and CK18 was high in OCUM-12/SP and OCUM-2MD3/SP, in compared with the control of parent OCUM-12 and OCUM-2MD3. These proteins were significantly associated with advanced invasion depth, lymph node metastasis, distant metastasis, or advanced clinical stage. RBBP6, DCTPP1, HSPA4, and ALDOA expression in particular were significantly associated with a poor prognosis in the 300 gastric cancer patients. RBBP6 was determined to be an independent prognostic factor. The motility-stimulating ability of OCUM-12/SP cells and OCUM-2MD3/SP cells was inhibited by RBBP6 siRNA. These findings might suggest that the eight proteins, RBBP6, GLG1, VPS13A, DCTPP1, HSPA9, HSPA4, ALDOA, and KRT18, utilizing comparative proteomics analysis, were perceived to be potential CSC markers of gastric cancer. Of the eight candidate proteins, RBBP6 was suggested to be a promising prognostic biomarker and a therapeutic target for gastric cancer.