A Fully Automated Artificial Intelligence System to Assist Pathologists' Diagnosis to Predict Histologically High-grade Urothelial Carcinoma from Digitized Urine Cytology Slides Using Deep Learning.

BACKGROUND: Urine cytology, although a useful screening method for urothelial carcinoma, lacks sensitivity. As an emerging technology, artificial intelligence (AI) improved image analysis accuracy significantly. OBJECTIVE: To develop a fully automated AI system to assist pathologists in the histological prediction of high-grade urothelial carcinoma (HGUC) from digitized urine cytology slides. DESIGN, SETTING, AND PARTICIPANTS: We digitized 535 consecutive urine cytology slides for AI use. Among these slides, 181 were used for AI development, 39 were used as AI test data to identify HGUC by cell-level classification, and 315 were used as AI test data for slide-level classification. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: Out of the 315 slides, 171 were collected immediately prior to bladder biopsy or transurethral resection of bladder tumor, and then outcomes were compared with the histological presence of HGUC in the surgical specimen. The primary aim was to compare AI prediction of the histological presence of HGUC with the pathologist's histological diagnosis of HGUC. Secondary aims were to compare the time required for AI evaluation and concordance between the AI's classification and pathologist's cytology diagnosis. RESULTS AND LIMITATIONS: The AI capability for predicting the histological presence of HGUC was 0.78 for the area under the curve. Comparing the AI predictive performance with pathologists' diagnosis, the AI sensitivity of 63% for histological HGUC prediction was superior to a pathologists' cytology sensitivity of 46% (p = 0.0037). On the contrary, there was no significant difference between the AI specificity of 83% and pathologists' specificity of 89% (p = 0.13), and AI accuracy of 74% and pathologists' accuracy of 68% (p = 0.08). The time required for AI evaluation was 139 s. With respect to the concordance between the AI prediction and pathologist's cytology diagnosis, the accuracy was 86%. Agreements with positive and negative findings were 92% and 84%, respectively. CONCLUSIONS: We developed a fully automated AI system to assist pathologists' histological diagnosis of HGUC using digitized slides. This AI system showed significantly higher sensitivity than a board-certified cytopathologist and may assist pathologists in making urine cytology diagnoses, reducing their workload. PATIENT SUMMARY: In this study, we present a deep learning-based artificial intelligence (AI) system that classifies urine cytology slides according to the Paris system. An automated AI system was developed and validated with 535 consecutive urine cytology slides. The AI predicted histological high-grade urothelial carcinoma from digitized urine cytology slides with superior sensitivity than pathologists, while maintaining comparable specificity and accuracy.

Maintenance of HCT116 colon cancer cell line conforms to a stochastic model but not a cancer stem cell model.

The cancer stem cell (CSC) model, in which a small population of cells within a tumor possesses the ability to self-renew and reconstitute the phenotype of primary tumor, has gained wide acceptance based on evidence over the past decade. It has also been reported that cancer cell lines contain a CSC subpopulation. However, phenotypic differences between CSCs and non-CSCs in cancer cell lines are not better defined than in primary tumors. Furthermore, some cell lines do not have a CSC population, revealed as a side population and expression of CD133. Thus, the identification of CSCs in cancer cell lines remains elusive. Here, we investigated the CSC hierarchy within HCT116 colon cancer cells, which do not have a CD133-positive subpopulation. We examined the expression of alternative CSC markers epithelial specific antigen (ESA) and CD44 in floating-sphere-derived cells, which are known to be the cells of enriching CSCs. Sphere-derived HCT116 cells exhibited heterogeneous expression of ESA and CD44. The two major subpopulations of HCT116 sphere cells (ESA(low)CD44(-/low) and ESA(high)CD44(high)) exhibited a biological/proliferative hierarchy of sphere-forming and soft agar colony-forming activity. However, there was no difference between the two subpopulations in the incidence of xenograft tumors. When ESA(low)CD44(-/low) cells were allowed to aggregate and re-form floating-spheres, the biological/proliferative hierarchy of parental HCT116 spheres was reconstituted, in terms of ESA and CD44 expression. Thus, HCT116 cells have plasticity when they are set in floating-spheres, suggesting that maintenance of the HCT116 cell line conforms to a stochastic model, not a CSC model.

Microtubule-associated protein 2-positive cells derived from microglia possess properties of functional neurons.

Microglia are believed to play an important role in the regulation of phagocytosis, neuronal survival, neuronal cell death, and inflammation. Recent studies have demonstrated that microglia are multipotential stem cells that give rise to neurons, astrocytes, and oligodendrocytes. However, the functional properties of neurons derived from microglia are poorly understood. In this study, we investigated the possibility that microglia differentiate into functional neurons. Immunocytochemical study demonstrated that microtubule-associated protein 2 (MAP2)-positive cells were derived from microglia under differentiation conditions. Intracellular Ca(2+) imaging study demonstrated that KCl caused no significant changes in [Ca(2+)](i) in microglia, whereas it caused a remarkable increase in [Ca(2+)](i) in microglia-derived cells. Furthermore, electrophysiological study demonstrated that the spike waveform, firing rate, and tetrodotoxin sensitivity of extracellular action potentials evoked by 4-aminopyridine from microglia-derived MAP2-positive cells were nearly identical to those from cultured cortical neurons. These results suggest that microglia-derived MAP2-positive cells possess properties of functional neurons.

Invasion precedes tumor mass formation in a malignant brain tumor model of genetically modified neural stem cells.

Invasiveness, cellular atypia, and proliferation are hallmarks of malignant gliomas. To effectively target each of these characteristics, it is important to understand their sequence during tumorigenesis. However, because most gliomas are diagnosed at an advanced stage, the chronology of gliomagenesis milestones is not well understood. The aim of the present study was to determine the onset of these characteristics during tumor development. Brain tumor-initiating cells (BTICs) were established by overexpressing H-Ras(V12) in normal neural stem/progenitor cells isolated from the subventricular zone of adult mice harboring a homozygous deletion of the Ink4a/Arf locus. High-grade malignant brain tumors were then created by orthotopic implantation of 10(5) BTICs into the forebrain of 6-week-old wild-type mice. Micewere killed every week for 5 weeks, and tumors were assessed for cellular atypia, proliferation, hemorrhage, necrosis, and invasion. All mice developed highly invasive, hypervascular glioblastoma-like tumors. A 100% penetrance rate and a 4-week median survival were achieved. Tumor cell migration along fiber tracts started within days after implantation and was followed by perivascular infiltration of tumor cells with marked recruitment of reactive host cells. Next, cellular atypia became prominent. Finally, mass proliferation and necrosis were observed in the last stage of the disease. Video monitoring of BTICs in live brain slices confirmed the early onset of migration, as well as the main cell migration patterns. Our results showed that perivascular and intraparenchymal tumor cell migration precede tumor mass formation in the adult brain, suggesting the need for an early and sustained anti-invasion therapy.

Microglia-derived interleukin-6 and leukaemia inhibitory factor promote astrocytic differentiation of neural stem/progenitor cells.

Neural stem/progenitor cells (NSPCs) proliferate and differentiate depending on their intrinsic properties and local environment. It has been recognized that astrocytes promote neurogenic differentiation of NSPCs, suggesting the importance of cell-cell interactions between glial cells and NSPCs. Recent studies have demonstrated that microglia, one type of glial cells, play an important role in neurogenesis. However, little is known about how activated microglia control the proliferation and differentiation of NSPCs. In this study, we investigated the possibility that microglia-derived soluble factors regulate the behaviour of NSPCs. To this end, NSPCs and microglial cultures were obtained from rat embryonic day 16 subventricular zone (SVZ) and rat postnatal 1 day cortex, respectively, and the conditioned medium from microglia was prepared. Microglial-conditioned medium had no significant effect on the proliferation of NSPCs. In contrast, it increased the percentage of cells positive for a marker of astrocytes, glial fibrillary acidic protein (GFAP) during differentiation. The induction of astrocytic differentiation by microglial-conditioned medium was reduced by the inhibition of the Janus kinase/signal transducer and activation of transcription (JAK/STAT) and mitogen-activated protein kinase (MAPK) pathways. Furthermore, microglia-derived interleukin (IL)-6 and leukaemia inhibitory factor (LIF) were identified as essential molecules for this astrocytic differentiation using neutralizing antibodies and recombinant cytokines. Our results suggest that microglia as well as astrocytes contribute to the integrity of the local environment of NSPCs, and at least IL-6 and LIF released by activated microglia promote astrocytic differentiation of NSPCs via the activation of the JAK/STAT and MAPK pathways.