Convolutional neural network deep learning model accurately detects rectal cancer in endoanal ultrasounds.

BACKGROUND: Imaging is vital for assessing rectal cancer, with endoanal ultrasound (EAUS) being highly accurate in large tertiary medical centers. However, EAUS accuracy drops outside such settings, possibly due to varied examiner experience and fewer examinations. This underscores the need for an AI-based system to enhance accuracy in non-specialized centers. This study aimed to develop and validate deep learning (DL) models to differentiate rectal cancer in standard EAUS images. METHODS: A transfer learning approach with fine-tuned DL architectures was employed, utilizing a dataset of 294 images. The performance of DL models was assessed through a tenfold cross-validation. RESULTS: The DL diagnostics model exhibited a sensitivity and accuracy of 0.78 each. In the identification phase, the automatic diagnostic platform achieved an area under the curve performance of 0.85 for diagnosing rectal cancer. CONCLUSIONS: This research demonstrates the potential of DL models in enhancing rectal cancer detection during EAUS, especially in settings with lower examiner experience. The achieved sensitivity and accuracy suggest the viability of incorporating AI support for improved diagnostic outcomes in non-specialized medical centers.

Software toolbox for analysis of the endometrial myometrial junction - a pilot study.

PURPOSE: To develop and evaluate an algorithm for computerized evaluation and measurement of the endometrial-myometrial junction (EMJ). MATERIALS AND METHODS: The advanced image processing toolbox of the Matlab software package was used for identificiation and quantitative analysis of the EMJ area on three-dimensional (3D) rendered coronal plane uterine images, with clear-cut borders of the EMJ. The algorithm was used to process the images and calculate the geometric parameters characterizing the identified EM The manual measurements of the maximum thickness of the EMJ were compared to automated measurements performed by the algorithm on the same images. RESULTS: For all three interfaces, the mean maximum manual measurement was less than the mean maximui computed measurement. The differences between the two measurements were not statistically significant (p = 0.275, 0.608 and 0.41 for the right wall, left wall, and fundus, respectively). The mean systematic and random errors ranged from 5.4% tol9.3% and 20.4 to 48.6%, respectively. Pearson correlations for the right wall, left wall and fundus (r = 0.642, p = 0.001; r = 0.730, p < 0.001, and r 0.694, p < 0.001, respectively) were good. CONCLUSIONS: Maximum EMJ thickness measurements performed by the innovative Matla software algorithm are as accurate as manual measurements, and have the potential to reduce inter-observer variability.

Impact of p53 Status on Radiosensitization of Tumor Cells by MET Inhibition-Associated Checkpoint Abrogation.

UNLABELLED: Signaling via the MET receptor tyrosine kinase has been implicated in crosstalk with cellular responses to DNA damage. Our group previously demonstrated that MET inhibition in tumor cells with deregulated MET activity results in radiosensitization via downregulation of the ATR-CHK1-CDC25 pathway, a major signaling cascade responsible for intra-S and G2-M cell-cycle arrest following DNA damage. Here we aimed at studying the potential therapeutic application of ionizing radiation in combination with a MET inhibitor, EMD-1214063, in p53-deficient cancer cells that harbor impaired G1-S checkpoint regulation upon DNA damage. We hypothesized that upon MET inhibition, p53-deficient cells would bypass both G1-S and G2-M checkpoints, promoting premature mitotic entry with substantial DNA lesions and cell death in a greater extent than p53-proficient cells. Our data suggest that p53-deficient cells are more susceptible to EMD-1214063 and combined treatment with irradiation than wild-type p53 lines as inferred from elevated γH2AX expression and increased cytotoxicity. Furthermore, cell-cycle distribution profiling indicates constantly lower G1 and higher G2-M population as well as higher expression of a mitotic marker p-histone H3 following the dual treatment in p53 knockdown isogenic variant, compared with the parental counterpart. IMPLICATIONS: The concept of MET inhibition-mediated radiosensitization enhanced by p53 deficiency is of high clinical relevance, as p53 is frequently mutated in numerous types of human cancer. The current data point for a therapeutic advantage for an approach combining MET targeting along with DNA-damaging agents for MET-positive/p53-negative tumors.

Correlation between the tumoral expression of beta3-integrin and outcome in cervical cancer patients who had undergone radiotherapy.

Integrins are cell-surface receptors, which mediate cell-to-cell and cell-to-extracellular matrix adhesion. Besides playing an important role in tumour angiogenesis, beta3-integrin is also expressed in several types of epithelial cancer cells. It was the purpose of the present study to evaluate the prognostic value of beta3-integrin expression in patients with cervical cancer. Biopsies were taken from 82 patients with squamous cell or adenocarcinomas of the uterine cervix who had undergone external-beam radiotherapy with or without brachytherapy. These tissue samples were analysed immunohistochemically for the expression of beta3-integrin. The impact of immunoreactivity for beta3-integrin on survival end points was assessed by univariate and multivariate analyses, and its correlation with clinicopathological characteristics evaluated by crosstabulations. beta3-integrin was expressed in 61% (50 of 82) of the patients. Kaplan-Meier curves revealed local progression-free survival, distant metastasis-free survival and cause-specific survival to be significantly shorter (P-values according to the log-rank test: 0.002, 0.04 and 0.01, respectively) in patients with beta3-integrin expression. The prognostic impact of this parameter was even higher than for other well-known prognostic parameters and remained statistically significant in the multivariate analyses. beta3-integrin, which is expressed in the majority of patients with advanced cervical cancer, has a significant prognostic impact on outcome according to univariate and multivariate analyses.

Genes induced by growth arrest in a pancreatic beta cell line: identification by analysis of cDNA arrays.

Pancreatic beta cell lines are a potentially attractive source of material for cell therapy of insulin-dependent diabetes mellitus. However, induction of proliferation in post-mitotic, differentiated beta cells is likely to affect the expression of multiple genes associated with cell function, resulting in dedifferentiation. We have developed a murine beta cell line by conditional transformation with the SV40 T antigen oncoprotein. These cells can undergo reversible induction of proliferation and growth arrest. Here we utilized this model to identify differences in gene expression between proliferating and quiescent beta cells, by analyzing known beta cell genes and differentially secreted proteins, as well as by a systematic survey of a mouse cDNA array. Our findings demonstrate that growth arrest stimulates expression of the insulin gene and genes encoding components of the insulin secretory vesicles. Screening of the cDNA array revealed the activation of multiple genes following growth arrest, many of them novel genes which may be related to beta cell function. Characterization of these genes is likely to contribute to our understanding of beta cell function and the ability to employ beta cell lines in cell therapy of diabetes.

Computerized quantification of structures within ovarian cysts using ultrasound images.

Ovarian cysts are a common type of ovarian mass. The morphology of cysts, as it appears in ultrasound images, is currently used for classification of ovarian pathologies. However, this classification process is based on human interpretation of the sonographic image. In this article, a semiautomatic algorithm for the quantification of ovarian cysts is presented. This algorithm categorizes the structures within a cyst and extracts their quantitative geometric properties (width, characteristic diameter). To assess the validity of the technique, its performance was compared to human classification and manual measurements made by an expert. The results show a good match between automatic evaluations made by a computer and those of an experienced observer, indicating a potential for clinical use.

Structure-function analysis of the trypanosomatid spliced leader RNA.

In trypanosomes, all mRNAs possess a spliced leader (SL) at their 5' end. SL is added to pre-mRNA via trans -splicing from a small RNA, the SL RNA. To examine structure-function aspects of the trypanosomatid SL RNA, an in vivo system was developed in the monogenetic trypanosomatid Leptomonas collosoma to analyze the function of chimeric and site-directed SL RNA mutants in trans -splicing. Stable cell lines expressing chimeric and mutated SL RNA from the authentic SL RNA regulatory unit were obtained. The chimeric RNA was expressed and assembled into an SL RNP particle, but could not serve as a substrate in splicing. Mutations in loop II and III of L.collosoma SL RNA formed the Y structure intermediate. In addition, a double SL RNA mutant in loop II, and positions 7 and 8 of the intron, also formed the Y structure intermediate, suggesting that these intron positions, although proposed to participate in the interaction of SL RNA with U5, may not be crucial for the first step of the trans -splicing reaction. A mutation in the exon located in loop I was not utilized in splicing, suggesting the importance of exon sequences for trans -splicing in trypanosomes. However, a double SL RNA mutant in loop II and exon position 31 was utilized in both steps of splicing; the mutant thus provides a model molecule for further analysis of positions essential for the function of the SL RNA.

Decay constant of Doppler flow waveform as a possible indicator of ovarian malignancy.

The objectives of this study were to analyze the decay constant (tau) of the Doppler flow waveform in ovarian tumors; to determine if differences in this constant can discriminate between malignant and benign ovarian tumors; and to compare the decay constant to the known resistive index (RI), in order to determine its potential prognostic application. Patients with ovarian masses (46) were evaluated in a retrospective study; 13 had malignant tumors, 7 showed tumors with low malignant potential (LMP), 11 had benign masses, 4 had secondary ovarian metastases and 11 had functional ovarian masses. Doppler flow waves measured in the ovary before operation were analyzed from archival videotapes. The RI was calculated preoperatively, and the decay constant of the flow waveform was analyzed retrospectively. We approximated the decaying portion of the flow waveform from the systolic peak to the diastolic level to an exponential curve. Then, the decay constant associated with the flow signal was compared for different types of ovarian pathology. Ovaries with malignancies showed significantly higher mean values for the decay constant (89.7; 95% confidence interval 60.0-119.3) than those with benign tumors (41.8; 25.7-57.9) (p < 0.007), where tau is provided in pixels (in this study each pixel equals approximately 11.4 ms). The mean RI value for malignant tumors was 0.44 +/- 0.12 whereas, in benign tumors, it was 0.622 +/- 0.11. For the benign tumors, both tau and RI did not differ significantly from the measured indices in LMP tumors, metastases and functional ovarian findings. In addition, when the cutoff value of tau was set at 48, 92.3% of all malignancies were identifiable using only tau. This preliminary study indicates that the decay constant of the Doppler flow waveform is able to discriminate between malignant and benign masses and may, thus, provide substantial assistance as an additional parameter in the diagnosis of malignant ovarian tumors in postmenopausal patients.

Stable transfection in the monogenetic trypanosomatid Leptomonas collosoma--transcription barrier of heterologous trypanosomatid SL RNA genes and expression of a chimeric SL RNA molecule.

A stable transfection system for the lower trypanosomatid Leptomonas collosoma was established using the Leishmania expression vector pX that was derived from an extrachromosomal amplified DNA carrying the dihydrofolate reductase-thymidylate synthase gene. Transformants harboring the pX vector were selected on Geneticin, and cell lines harboring as many as 200 copies per cell were obtained by increasing the drug concentration. The system was utilized to examine the expression of the SL RNA genes of Trypanosoma brucei and Leishmania mexicana amazonensis. Despite the high copy number of the foreign genes, no heterologous SL RNA was detected in steady-state RNA populations or by nascent transcription of cells made permeable by lysolecithin, suggesting the existence of a transcription barrier for this gene among the trypanosomatids. Such a barrier does not exist for the T. brucei 5S rRNA gene, since transcription of this gene was detected in permeable cells carrying the heterologous gene and in steady-state RNA population. To overcome the transcription barrier, the authentic regulatory region of the L. collosoma SL gene was identified. Chimeric constructs carrying 50 or 415 nt of the L. collosoma SL upstream sequence and 24 nt of the L. collosoma exon portion were fused to the T. brucei SL RNA gene at the SL portion. Expression of a chimeric SL RNA of 150 nt, composed of 24 nt L. collosoma RNA and 126 nt T. brucei RNA, was observed only in cell lines carrying the 415-nt upstream sequence. The efficient expression of the chimeric SL RNA, using the L. collosoma SL gene regulatory regions, may facilitate a structure-function analysis of chimeric and site-directed mutated SL RNA in trans-splicing.

An automatic contour extraction algorithm for short-axis cardiac magnetic resonance images.

In this article an automatic process is presented for the extraction of the left ventricular contours. The only operator-supplied information is the location of a single point within the left ventricle. The raw data for processing are short-axis cardiac MR images, obtained using various MRI protocols. In this study we considered only protocols for which the blood appears dark. The proposed algorithm performs smoothing and thresholding of the original MR images, and uses the segmented images to form edge images. From those edge images the endocardial curve is extracted using a modified contour following approach. The algorithm can accommodate moderate noise and small gaps in the contour. A smooth curve that approximates the epicardium is then obtained using a method combining radial search and FFT based smoothing. The quality of the obtained contours was visually assessed by three observers. Their evaluation shows that the algorithm is successful when images of reasonable quality are processed. Further evaluation work is required in order to fully assess the performance and limitation of the suggested method.

A two-dimensional extension of minimum cross entropy thresholding for the segmentation of ultrasound images.

Segmentation is often an important step in medical image analysis. The local entropy is a possible variable for segmenting ultrasound images containing fluid surrounded by a soft tissue. A commonly used tool for image segmentation is thresholding. Recently, a new thresholding technique, known as "minimum cross entropy thresholding" (MCE), has been proposed. We present a multivariate extension of MCE in which the segmented variable (gray level) is replaced by a weighted combination of several image parameters. We propose to use a bivariate extension of MCE, which uses a linear combination of the gray level and the local entropy. The results obtained are demonstrated for ultrasound images of ovarian cysts.

The distribution of the local entropy in ultrasound images.

In this article, a model for the amplitude statistics of the backscattered ultrasonic signal is presented. We propose to view a tissue as being composed of a large number of small units, each having slightly different characteristics. This variability within the tissue is expressed by fluctuations in the parameters of the local probability distribution function (PDF). Based on analogous expressions derived for radio propagation and optical scintillations, the local PDF is modulated by a lognormal distribution of the local standard deviation. Integrating the local contributions yields the amplitude PDF for the entire tissue. We show that the local entropy is a normal variable, since for four different local PDFs it is linearly related to the logarithm of the local standard deviation. When the local entropy histogram exhibits distinct and multiple peaks, the local entropy distribution can be used for region segmentation. This fact is demonstrated for ultrasonic images of ovarian cysts.

Ligand binding by fibroblast growth factor receptors investigated using chimeric receptor molecules.

In order to map in detail the ligand binding sites of fibroblast growth factor receptor 2 (FGFR2) and keratinocyte growth factor receptor (KGFR), we have generated receptor molecules that are chimeric within the membrane proximal sequence that varies between them. The chimeric molecules are found to bind aFGF with a greater than 5-fold difference in affinity, indicating that there is coupling between the chimeric regions with respect to aFGF binding. Further, binding of bFGF and KGF is abolished in the chimeras, showing that the binding site for these ligands requires the whole of the 48- or 50- amino acid variable sequence to be intact. Direct interactions between the different regions exchanged in the chimeras are most probably involved in forming KGF or bFGF binding sites.

Developmental localization of the splicing alternatives of fibroblast growth factor receptor-2 (FGFR2).

The gene for fibroblast growth factor receptor-2 (FGFR2) encodes two splice variants designated here as keratinocyte growth factor (KGFR) and bek. Their ligand-binding specificity is markedly different due to mutually exclusive alternative splicing. We asked whether alternative exon usage, in addition to influencing receptor specificity, could be correlated with transcriptional localization. This problem was studied by in situ hybridization and PCR, using probes and primers specific for the alternative exons of FGFR2. Transcripts of both variants were detected in all three germ layers within the embryonic and the extraembryonic areas of the primitive-streak embryo. The overall level of KGFR expression surpassed that of bek. The localized expression of both variant receptors was, however, more diffuse in the gastrula than later during organogenesis, when KGFR transcripts were evident mainly in epithelia, whereas bek was present in the corresponding mesenchymes. Our findings show the following: (1) Expression of both FGFR2 variants is concordant with their involvement in murine gastrulation. They may endow competence to multiple areas, which may be restricted by their more confined ligands. (2) KGFR and bek seem to have unique roles in the development of the skin and its derivatives, whereas bek is preferentially expressed during osteogenesis. The two variants share potential regions of trans regulation in the genome; hence, we suggest that alternative splicing is jointly responsible for ligand binding and spatial specificity. (3) Finally, we defined the binding specificity of KGFR and bek to various FGF. The possibility of identifying specific functional areas for certain ligand-receptor pairs is discussed.

Multiple structural elements determine ligand binding of fibroblast growth factor receptors. Evidence that both Ig domain 2 and 3 define receptor specificity.

The murine fibroblast growth factor receptor 2 (FGFR2) and keratinocyte growth factor receptor (KGFR) are two products of the same gene which display distinct binding specificities. We and others have shown that a major structural element underlying this functional divergence is a variable 50 amino acids long region constituting the C-terminal half of the third immunoglobulin (Ig)-like domain of the receptor. This region of the two receptors is encoded by two distinct exons which are alternatively used in cells of different tissues and origin. To further investigate the role of this confined variable region in determining ligand binding specificity we have generated a chimeric molecule between FGFR1 and KGFR where the variable segment of KGFR replaces the homologous region in FGFR1. Binding studies as well as chemical crosslinking of radiolabeled ligands revealed that the recombinant FGFR1/KGFR chimera has retained the binding affinity to acidic FGF and FGF4 (hst/kfgf) but lost the capacity to bind basic FGF (bFGF). This chimeric receptor bound keratinocyte growth factor (KGF), however, with significantly lower affinity as compared with KGFR. High affinity binding of KGF was acquired only when also domain 2 in this chimera was replaced by its homologous domain from FGFR2. These results demonstrate that ligand binding and specificity involves multiple receptor elements which are located at both Ig-like domain 2 and 3 of FGF receptors.

A confined variable region confers ligand specificity on fibroblast growth factor receptors: implications for the origin of the immunoglobulin fold.

Binding of cellular growth factors to their receptors constitutes a highly specific interaction and the basis for cell and tissue-type specific growth and differentiation. A unique feature of fibroblast growth factor (FGF) receptors is the multitude of structural variants and an unprecedented degree of cross-reactivity between receptors and their various ligands. To examine receptor-ligand specificity within these families of growth factors and receptors, we used genetic engineering to substitute discrete regions between Bek/FGFR2 and the closely related keratinocyte growth factor receptor (KGFR). We demonstrate that a confined, 50 amino acid, variable region within the third immunoglobulin-like domain of Bek and KGFR exclusively determines their ligand binding specificities. Replacing the variable region of Bek/FGFR2 with the corresponding sequence of KGFR resulted in a chimeric receptor which bound KGF and had lost the capacity to bind basic FGF. We present evidence that the two variable sequences are encoded by two distinct exons that map close together in the mouse genome and follow a constant exon, suggesting that the two receptors were derived from a common gene by mutually exclusive alternative mRNA splicing. These results identify the C-terminal half of the third immunoglobulin-like domain of FGF receptors as a major determinant for ligand binding and present a novel genetic mechanism for altering receptor-ligand specificity and generating receptor diversity.

Flg-2, a new member of the family of fibroblast growth factor receptors.

Two genes, flg and bek, have been recently shown to encode receptors for fibroblast growth factors (FGF). Here we report the molecular cloning and sequence of a new member of the FGF receptor family, denoted flg-2, which was isolated from a human keratinocyte cDNA library. The cDNA sequence predicts an extracellular region possessing three immunoglobulin-like domains, a single transmembrane region and a cytoplasmic portion containing the tyrosine kinase domain split by a short inter-kinase segment. The amino acid sequence of flg-2 shows 68% and 64% identity with bek and flg, respectively. The most variable domain among the three genes is the amino-terminal immunoglobulin-like domain. Comparison with the chicken FGF receptor genes suggests that flg-2 is homologous to cek-2, whereas flg and bek are homologous to cek-1 and cek-3, respectively. Analysis of mRNA from various tissues shows that flg-2 is expressed predominantly in skin, brain and lung.

Developmental expression of c-kit, a proto-oncogene encoded by the W locus.

Developmental expression of the c-kit proto-oncogene, a receptor tyrosine kinase encoded by the W locus, was investigated by in situ hybridization in normal mouse embryos. Early after implantation transcripts were detectable only in the maternal placenta (6 1/2-7 1/2 days p.c.). Subsequently (8 1/2 days p.c.) numerous ectodermal (neural tube, sensory placodes) and endodermal (embryonic gut) derivatives expressed c-kit. Later transcripts were detected also in the blood islands of the yolk sac and in the embryonic liver, the main sites of embryonic hemopoiesis. Around midgestation, transcripts accumulated in the branchial pouches and also in primordial germ cells of the genital ridges. This complex pattern of expression remained characteristic also later in gestation, when c-kit was expressed in highly differentiated structures of the craniofacial area, in presumptive melanoblasts and in the CNS. In the adult ovary, maternal c-kit transcripts were detected. They were present in the oocytes of both immature and mature ovarian follicles, but not in the male germ line, where c-kit expression may be down regulated. Thus, c-kit activity is complex and appears in multiple tissues including those that also display defects in mutations at the W locus where c-kit is encoded. Correlation between W phenotypes and c-kit expression, as well as the regulation of the complex and multiple expression of polypeptide growth factors and receptors, is discussed.