This study considers the use of deep learning to diagnose osteoporosis from hip radiographs, and whether adding clinical data improves diagnostic performance over the image mode alone. For objective labeling, we collected a dataset containing 1131 images from patients who underwent both skeletal bone mineral density measurement and hip radiography at a single general hospital between 2014 and 2019. Osteoporosis was assessed from the hip radiographs using five convolutional neural network (CNN) models. We also investigated ensemble models with clinical covariates added to each CNN. The accuracy, precision, recall, specificity, negative predictive value (npv), F1 score, and area under the curve (AUC) score were calculated for each network. In the evaluation of the five CNN models using only hip radiographs, GoogleNet and EfficientNet b3 exhibited the best accuracy, precision, and specificity. Among the five ensemble models, EfficientNet b3 exhibited the best accuracy, recall, npv, F1 score, and AUC score when patient variables were included. The CNN models diagnosed osteoporosis from hip radiographs with high accuracy, and their performance improved further with the addition of clinical covariates from patient records.

It is necessary to accurately identify dental implant brands and the stage of treatment to ensure efficient care. Thus, the purpose of this study was to use multi-task deep learning to investigate a classifier that categorizes implant brands and treatment stages from dental panoramic radiographic images. For objective labeling, 9767 dental implant images of 12 implant brands and treatment stages were obtained from the digital panoramic radiographs of patients who underwent procedures at Kagawa Prefectural Central Hospital, Japan, between 2005 and 2020. Five deep convolutional neural network (CNN) models (ResNet18, 34, 50, 101 and 152) were evaluated. The accuracy, precision, recall, specificity, F1 score, and area under the curve score were calculated for each CNN. We also compared the multi-task and single-task accuracies of brand classification and implant treatment stage classification. Our analysis revealed that the larger the number of parameters and the deeper the network, the better the performance for both classifications. Multi-tasking significantly improved brand classification on all performance indicators, except recall, and significantly improved all metrics in treatment phase classification. Using CNNs conferred high validity in the classification of dental implant brands and treatment stages. Furthermore, multi-task learning facilitated analysis accuracy.

The tumor organoid (tumoroid) model in three-dimensional (3D) culture systems has been developed to reflect more closely the in vivo tumors than 2D-cultured tumor cells. Notably, extracellular vesicles (EVs) are efficiently collectible from the culture supernatant of gel-free tumoroids. Matrix metalloproteinase (MMP) 3 is a multi-functional factor playing crucial roles in tumor progression. However, roles of MMP3 within tumor growth and EVs have not unveiled. Here, we investigated the protumorigenic roles of MMP3 on integrities of tumoroids and EVs. We generated MMP3-knockout (KO) cells using the CRISPR/Cas9 system from rapidly metastatic LuM1 tumor cells. Moreover, we established fluorescent cell lines with palmitoylation signal-fused fluorescent proteins (tdTomato and enhanced GFP). Then we confirmed the exchange of EVs between cellular populations and tumoroids. LuM1-tumoroids released large EVs (200–1000 nm) and small EVs (50–200 nm) while the knockout of MMP3 resulted in the additional release of broken EVs from tumoroids. The loss of MMP3 led to a significant reduction in tumoroid size and the development of the necrotic area within tumoroids. MMP3 and CD9 (a category-1 EV marker tetraspanin protein) were significantly down-regulated in MMP3-KO cells and their EV fraction. Moreover, CD63, another member of the tetraspanin family, was significantly reduced only in the EVs fractions of the MMP3-KO cells compared to their counterpart. These weakened phenotypes of MMP3-KO were markedly rescued by the addition of MMP3-rich EVs or conditioned medium (CM) collected from LuM1-tumoroids, which caused a dramatic rise in the expression of MMP3, CD9, and Ki-67 (a marker of proliferating cells) in the MMP3-null/CD9-low tumoroids. Notably, MMP3 enriched in tumoroids-derived EVs and CM deeply penetrated recipient MMP3-KO tumoroids, resulting in a remarkable enlargement of solid tumoroids, while MMP3-null EVs did not. These data demonstrate that EVs can mediate molecular transfer of MMP3, resulting in increasing the proliferation and tumorigenesis, indicating crucial roles of MMP3 in tumor progression.

Tumor growth, progression, and therapy resistance are crucial factors in the prognosis of cancer. The properties of three-dimensional (3D) tumor-like organoids (tumoroids) more closely resemble in vivo tumors compared to two-dimensionally cultured cells and are therefore effectively used for assays and drug screening. We here established a repurposed drug for novel anticancer research and therapeutics using a 3D tumoroid-based screening system. We screened six pharmacologically active compounds by using an original tumoroid-based multiplex phenotypic screening system with a matrix metalloproteinase 9 (MMP9) promoter-driven fluorescence reporter for the evaluation of both tumoroid formation and progression. The antiparkinson drug benztropine was the most effective compound uncovered by the screen. Benztropine significantly inhibited in vitro tumoroid formation, cancer cell survival, and MMP9 promoter activity. Benztropine also reduced the activity of oncogenic signaling transducers and trans-activators for MMP9, including STAT3, NF-κB, and β-catenin, and the properties of cancer stem cells/cancer-initiating cells. Benztropine and GBR-12935 directly targeted the dopamine transporter DAT/SLC6A3, whose genetic alterations such as amplification were correlated with poor prognosis for cancer patients. Benztropine also inhibited the tumor growth, circulating tumor cell (CTC) number, and rate of metastasis in a tumor allograft model in mice. In conclusion, we propose the repurposing of benztropine for anticancer research and therapeutics that can suppress tumor progression, CTC, and metastasis of aggressive cancers by reducing key pro-tumorigenic factors.

Background and Objectives: A few deep learning studies have reported that combining image features with patient variables enhanced identification accuracy compared with image-only models. However, previous studies have not statistically reported the additional effect of patient variables on the image-only models. This study aimed to statistically evaluate the osteoporosis identification ability of deep learning by combining hip radiographs with patient variables. Materials andMethods: We collected a dataset containing 1699 images from patients who underwent skeletal-bone-mineral density measurements and hip radiography at a general hospital from 2014 to 2021. Osteoporosis was assessed from hip radiographs using convolutional neural network (CNN) models (ResNet18, 34, 50, 101, and 152). We also investigated ensemble models with patient clinical variables added to each CNN. Accuracy, precision, recall, specificity, F1 score, and area under the curve (AUC) were calculated as performance metrics. Furthermore, we statistically compared the accuracy of the image-only model with that of an ensemble model that included images plus patient factors, including effect size for each performance metric. Results: All metrics were improved in the ResNet34 ensemble model compared with the image-only model. The AUC score in the ensemble model was significantly improved compared with the image-only model (difference 0.004; 95% CI 0.002–0.0007; p = 0.0004, effect size: 0.871). Conclusions: This study revealed the additional effect of patient variables in identification of osteoporosis using deep CNNs with hip radiographs. Our results provided evidence that the patient variables had additive synergistic effects on the image in osteoporosis identification.

Accumulating evidence has shown that cancer stroma and BM-derived cells (BMDCs) in the tumor microenvironment (TME) play vital roles in tumor progression. However, the mechanism by which oral cancer stroma recruits any particular subset of BMDCs remains largely unknown. Here, we sought to identify the subset of BMDCs that is recruited by cancer stroma. We established a sequential transplantation model in BALB/c nude mice, including (a) BM transplantation of GFP-expressing cells and (b) coxenografting of patient-derived stroma (PDS; 2 cases, designated PDS1 and PDS2) with oral cancer cells (HSC-2). As controls, xenografting was performed with HSC-2 alone or in combination with normal human dermal fibroblasts (HDF). PDS1, PDS2, and HDF all promoted BMDC migration in vitro and recruitment in vivo. Multicolor immunofluorescence revealed that the PDS coxenografts recruited Arginase-1 ⁺ CD11b ⁺ GR1 ⁺ GFP ⁺ cells, which are myeloid-derived suppressor cells (MDSCs), to the TME, whereas the HDF coxenograft did not. Screening using microarrays revealed that PDS1 and PDS2 expressed CCL2 mRNA (encoding C-C motif chemokine ligand 2) at higher levels than did HDF. Indeed, PDS xenografts contained significantly higher proportions of CCL2 ⁺ stromal cells and CCR2 ⁺ Arginase-1 ⁺ CD11b ⁺ GR1 ⁺ MDSCs (as receiver cells) than the HDF coxenograft. Consistently, a CCL2 synthesis inhibitor and a CCR2 antagonist significantly inhibited the PDS-driven migration of BM cells in vitro. Furthermore, i.p. injection of the CCR2 antagonist to the PDS xenograft models significantly reduced the CCR2 ⁺ Arginase-1 ⁺ CD11b ⁺ GR1 ⁺ MDSC infiltration to the TME. In conclusion, oral cancer stroma–secreted CCL2 is a key signal for recruiting CCR2 ⁺ MDSCs from BM to the TME.