Automatic segmentation and classification of Papanicolaou-stained cells and dataset for oral cancer detection.

BACKGROUND AND OBJECTIVE: Papanicolaou staining has been successfully used to assist early detection of cervix cancer for several decades. We postulate that this staining technique can also be used for assisting early detection of oral cancer, which is responsible for about 300,000 deaths every year. The rational for such claim includes two key observations: (i) nuclear atypia, i.e., changes in volume, shape, and staining properties of the cell nuclei can be linked to rapid cell proliferation and genetic instability; and (ii) Papanicolaou staining allows one to reliably segment cells' nuclei and cytoplasms. While Papanicolaou staining is an attractive tool due to its low cost, its interpretation requires a trained pathologist. Our goal is to automate the segmentation and classification of morphological features needed to evaluate the use of Papanicolaou staining for early detection of mouth cancer. METHODS: We built a convolutional neural network (CNN) for automatic segmentation and classification of cells in Papanicolaou-stained images. Our CNN was trained and evaluated on a new image dataset of cells from oral mucosa consisting of 1,563 Full HD images from 52 patients, annotated by specialists. The effectiveness of our model was evaluated against a group of experts. Its robustness was also demonstrated on five public datasets of cervical images captured with different microscopes and cameras, and having different resolutions, colors, background intensities, and noise levels. RESULTS: Our CNN model achieved expert-level performance in a comparison with a group of three human experts on a set of 400 Papanicolaou-stained images of the oral mucosa from 20 patients. The results of this experiment exhibited high Interclass Correlation Coefficient (ICC) values. Despite being trained on images from the oral mucosa, it produced high-quality segmentation and plausible classification for five public datasets of cervical cells. Our Papanicolaou-stained image dataset is the most diverse publicly available image dataset for the oral mucosa in terms of number of patients. CONCLUSION: Our solution provides the means for exploring the potential of Papanicolaou-staining as a powerful and inexpensive tool for early detection of oral cancer. We are currently using our system to detect suspicious cells and cell clusters in oral mucosa slide images. Our trained model, code, and dataset are available and can help practitioners and stimulate research in early oral cancer detection.

Population Dynamics is a Cancer Driver.

Most tissues are continuously renovated through the division of stem cells and the death of old or damaged cells, which is known as cell turnover rate (CTOR). Despite being in steady state, tissues have different population dynamics and leading to diverse clonality levels. Here, we propose and test that cell population dynamics can be a cancer driver. We employed the evolutionary software esiCancer to show that CTOR, within a range comparable to what is observed in human tissues, can amplify the risk of a mutation due to ancestral selection (ANSEL). In a high CTOR tissue, a mutated ancestral cell is likely to be selected and persist over generations, which leads to a scenario of elevated ANSEL profile, characterized by few niches of large clones, which does not occur in low CTOR. We found that CTOR is significantly associated with the risk of developing cancer, even when correcting for mutation load, indicating that population dynamics per se is a cancer driver. This concept is central to understanding cancer risk and for the design of new therapeutic interventions that minimize the contribution of ANSEL in cancer growth.

A CNN-based approach for joint segmentation and quantification of nuclei and NORs in AgNOR-stained images.

BACKGROUND AND OBJECTIVE: Oral cancer is the sixth most common kind of human cancer. Brush cytology for counting Argyrophilic Nucleolar Organizer Regions (AgNORs) can help early mouth cancer detection, lowering patient mortality. However, the manual counting of AgNORs still in use today is time-consuming, labor-intensive, and error-prone. The goal of our work is to address these shortcomings by proposing a convolutional neural network (CNN) based method to automatically segment individual nuclei and AgNORs in microscope slide images and count the number of AgNORs within each nucleus. METHODS: We systematically defined, trained and tested 102 CNNs in the search for a high-performing solution. This included the evaluation of 51 network architectures combining 17 encoders with 3 decoders and 2 loss functions. These CNNs were trained and evaluated on a new AgNOR-stained image dataset of epithelial cells from oral mucosa containing 1,171 images from 48 patients, with ground truth annotated by specialists. The annotations were greatly facilitated by a semi-automatic procedure developed in our project. Overlapping nuclei, which tend to hide AgNORs, thus affecting their true count, were discarded using an automatic solution also developed in our project. Besides the evaluation on the test dataset, the robustness of the best performing model was evaluated against the results produced by a group of human experts on a second dataset. RESULTS: The best performing CNN model on the test dataset consisted of a DenseNet-169 + LinkNet with Focal Loss (DenseNet-169 as encoder and LinkNet as decoder). It obtained a Dice score of 0.90 and intersection over union (IoU) of 0.84. The counting of nuclei and AgNORs achieved precision and recall of 0.94 and 0.90 for nuclei, and 0.82 and 0.74 for AgNORs, respectively. Our solution achieved a performance similar to human experts on a set of 291 images from 6 new patients, obtaining Intraclass Correlation Coefficient (ICC) of 0.91 for nuclei and 0.81 for AgNORs with 95% confidence intervals of [0.89, 0.93] and [0.77, 0.84], respectively, and p-values < 0.001, confirming its statistical significance. Our AgNOR-stained image dataset is the most diverse publicly available AgNOR-stained image dataset in terms of number of patients and the first for oral cells. CONCLUSIONS: CNN-based joint segmentation and quantification of nuclei and NORs in AgNOR-stained images achieves expert-like performance levels, while being orders of magnitude faster than the later. Our solution demonstrated this by showing strong agreement with the results produced by a group of specialists, highlighting its potential to accelerate diagnostic workflows. Our trained model, code, and dataset are available and can stimulate new research in early oral cancer detection.

Synthesizing Camera Noise Using Generative Adversarial Networks.

We present a technique for synthesizing realistic noise for digital photographs. It can adjust the noise level of an input photograph, either increasing or decreasing it, to match a target ISO level. Our solution learns the mappings among different ISO levels from unpaired data using generative adversarial networks. We demonstrate its effectiveness both quantitatively, using Kullback-Leibler divergence and Kolmogorov-Smirnov test, and qualitatively through a large number of examples. We also demonstrate its practical applicability by using its results to significantly improve the performance of a state-of-the-art trainable denoising method. Our technique should benefit several computer-vision applications that seek robustness to noisy scenarios.

Video Folding: Increased Framerate for Semi-Repetitive Sequences.

We introduce a technique to synthetically increase the framerate of semi-repetitive videos (i.e., videos of motion that repeats but not in an identical fashion) to aid in visualization. By reordering and combining frames from all repetitions, we produce a single non-repetitive sequence with much higher temporal resolution. Then, we use a novel frame warping technique based on a dense corrective flow to counteract differences between repetitions. The resulting video maintains smoothness of motion and additionally allows for seamless, infinite looping. We demonstrate the effectiveness of the proposed solution both quantitatively, by measuring the improvement over existing methods, and qualitatively, by performing a user evaluation and providing several examples in the article and accompanying video.

esiCancer: Evolutionary In Silico Cancer Simulator.

The evolution of cancer is inferred mainly from samples taken at discrete points that represent glimpses of the complete process. In this study, we present esiCancer as a cancer-evolution simulator. It uses a branching process, randomly applying events to a diploid oncogenome, altering probabilities of proliferation and death of the affected cells. Multiple events that occur over hundreds of generations may lead to a gradual change in cell fitness and the establishment of a fast-growing population. esiCancer provides a platform to study the impact of several factors on tumor evolution, including dominance, fitness, event rate, and interactions among genes as well as factors affecting the tumor microenvironment. The output of esiCancer can be used to reconstruct clonal composition and Kaplan-Meier-like survival curves of multiple evolutionary stories. esiCancer is an open-source, standalone software to model evolutionary aspects of cancer biology. SIGNIFICANCE: This study provides a customizable and hands-on simulation tool to model the effect of diverse types of genomic alterations on the fate of tumor cells.

Are Retinal Vessels Calibers Influenced by Blood Pressure Measured at the Time of Retinography Acquisition?

BACKGROUND: Retinal arterial narrowing is associated with higher office blood pressure (BP) and ambulatory blood pressure monitoring, and increased incidence of cardiovascular disease, but it is still unknown if the vessel caliber is associated with BP measured at the time of retinography acquisition. METHODS: Retinal arteriolar and venular calibers were measured by the microdensitometric method in 448 patients with hypertension. Participants underwent 24-hours ambulatory blood pressure (24-h ABP) monitoring simultaneously with the retinography acquisition. Association between arteriolar and venular calibers with increase of 10 mmHg in the mean 24-hours, daily, and nightly BP, and with BP measured at the time of retinography, was evaluated by ANOVA and multivariate analyses. RESULTS: Mean 24-hours, daytime and nighttime systolic and diastolic BP were inversely associated with the arteriolar caliber, but not with the venular caliber. Arteriolar caliber decreased -0.8 (95% CI -1.4 to -0.2) µm per 10-mmHg increase in 24-hours mean systolic BP, adjusted for age, gender, fellow vessel, and duration of hypertension (P = 0.01). The corresponding decreasing in arteriolar caliber by 10 mmHg of increasing in mean diastolic BP was -1.1 µm (-2.0 to -0.2, P = 0.02). The decrease of arteriolar caliber by the same increasing of BP measured at the time of retinography was lower and not statistically significant, particularly for mean diastolic BP and outer arterioles calibers: -1.0 (-1.8 to -0.2) µm in the daytime BP average versus -0.3 (-0.9 to 0.3) at the moment of retinography acquisition. CONCLUSIONS: These findings suggest that the caliber of arteriolar retinal vessels in patients with uncontrolled hypertension are not significantly influenced by blood pressure measured at the time of retinography acquisition.

Association between plasma adiponectin and arteriolar vessel caliber among elderly hypertensive subjects.

Circulating adiponectin has been related to vascular diseases, but few studies examined the relationship between plasma adiponectin and microvascular abnormalities among hypertensive individuals. We tested the association between plasma adiponectin level and retinal vessel calibers in patients with hypertension.This study included 172 patients with confirmed hypertension, aged 18-80 years. Subjects with recent cardiovascular events, advanced heart failure and end-stage renal disease were excluded. Arteriolar and venular calibers were measured in retinographies using a microdensitometric image-processing method. Blood pressure was measured using a validated oscillometric device. We observed a statistically significant inverse association between plasma adiponectin and arteriolar caliber among participants aged 60 years or older after controlling for confounders (Adjusted ß = -0.42; P = .001). In the final model, HbA1C and low-density lipoprotein also remained independently associated with arteriolar caliber. There was no association of adiponectin with venular caliber and retinal vessel calibers in participants <60 years old.Adiponectin is inversely associated with retinal arteriolar caliber in elderly hypertensive participants, suggesting that plasma adiponectin may be a marker of microvascular damage and of higher cardiovascular risk in this age stratum.

Association between carotid intima-media thickness and retinal arteriolar and venular diameter in patients with hypertension: a cross-sectional study.

OBJECTIVE: Abnormalities in retinal vessels are frequent, but its association with findings in other vascular sites has been scarcely investigated. This study aimed to examine the association between ultrasound-measured carotid intima-media thickness (IMT) and retinal arteriolar and venular diameters, in hypertensive patients. METHODS: In this cross-section study, 173 hypertensive patients had both retinography taken and digitized to determine vessel diameters by an edge-detecting computerized method and carotid ultrasound for semi-automated carotid IMT measurement. The association between the mean common carotid IMT and retinal arterioles and venules diameters was assessed by using multiple linear regression models. RESULTS: The mean (±SD) arteriolar and venular diameters were 102. 8 (±11.6) µm and 128.9 (±15.5) µm, respectively, and common carotid IMT was 0.87 (±0.19) mm. A significant and independent association was demonstrated for carotid IMT and retinal arteriolar caliber (adjusted ß -0.245, p = 0.001) and for carotid IMT and retinal venular caliber (adjusted ß 0.191, p = 0.009) after controlling for age, gender, systolic blood pressure, total cholesterol and HDL (high-density lipoprotein) cholesterol, prior cardiovascular disease, carotid plaque and the retinal fellow vessel. CONCLUSION: In patients with hypertension, carotid intima-media thickness, a marker of macrovascular damage, is significantly and independently associated with microvascular damage, determined by retinal arteriolar and venular calibers.

Using Patterns to Encode Color Information for Dichromats.

Color is one of the most common ways to convey information in visualization applications. Color vision deficiency (CVD) affects approximately 200 million individuals worldwide and considerably degrades their performance in understanding such contents by creating red-green or blue-yellow ambiguities. While several content-specific methods have been proposed to resolve these ambiguities, they cannot achieve this effectively in many situations for contents with a large variety of colors. More importantly, they cannot facilitate color identification. We propose a technique for using patterns to encode color information for individuals with CVD, in particular for dichromats. We present the first content-independent method to overlay patterns on colored visualization contents that not only minimizes ambiguities but also allows color identification. Further, since overlaying patterns does not compromise the underlying original colors, it does not hamper the perception of normal trichromats. We validated our method with two user studies: one including 11 subjects with CVD and 19 normal trichromats, and focused on images that use colors to represent multiple categories; and another one including 16 subjects with CVD and 22 normal trichromats, which considered a broader set of images. Our results show that overlaying patterns significantly improves the performance of dichromats in several color-based visualization tasks, making their performance almost similar to normal trichromats'. More interestingly, the patterns augment color information in a positive manner, allowing normal trichromats to perform with greater accuracy.

Nuclear morphometric analysis (NMA): screening of senescence, apoptosis and nuclear irregularities.

Several cellular mechanisms affect nuclear morphology which can therefore be used to assess certain processes. Here, we present an analytic tool to quantify the number of cells in a population that present characteristics of senescence, apoptosis or nuclear irregularities through nuclear morphometric analysis. The tool presented here is based on nuclear image analysis and evaluation of size and regularity of adhered cells in culture. From 46 measurements of nuclear morphometry, principal component analysis filtered four measurements that best separated regular from irregular nuclei. These measurements, namely aspect, area box, radius ratio and roundness were combined into a single nuclear irregularity index (NII). Normal nuclei are used to set the parameters for a given cell type, and different nuclear phenotypes are separated in an area versus NII plot. The tool was validated with ß-gal staining for senescence and annexin or caspases inhibitor for apoptosis as well as several treatments that induce different cellular phenotypes. This method provides a direct and objective way of screening normal, senescent, apoptotic and nuclear irregularities which may occur during failed mitosis or mitotic catastrophe, which may be very useful in basic and clinical research.

Computer-assisted methods to evaluate retinal vascular caliber: what are they measuring?

PURPOSE: Computer-assisted methods to measure retinal vessel diameters have been incorporated into research, but it is not clear which component of the vessels they are measuring. This study was conducted to compare measurements of retinal vessel diameter by using imaging-processing software on color fundus photographs (FPs) and fluorescein angiographs (FAs). METHODS: FP and FA images were taken simultaneously in 52 eyes of 31 patients referred for angiography for diagnosis of retinal disease. Arteriolar and venular calibers were measured in two concentric zones around the optic disc center. Pearson correlation coefficients and Bland-Altman plots were used to evaluate the agreement between the measurements made by FP and FA. RESULTS: The differences between the diameters measured by the microdensitometric method from FP and FA were 2.59 ± 8.67 µm in the inner arteriola, 4.93 ± 7.47 µm in the outer arteriola, -1,58 ± 8.49 µm in the inner venula, and -1.80 ± 7.28 µm in the outer venula. The differences plotted by the Bland-Altman method were slight. The Pearson correlation coefficients of measurements by FP and FA were 0.84 for inner zone and 0.87 for outer zone arterioles and 0.93 and 0.94 for the inner and outer zone venules, respectively. CONCLUSIONS: The very slight differences between measurements of retinal vessel diameter by the two methods demonstrate that the microdensitometric method mostly measures the vessel lumen. Differences in vessel diameters measured by the microdensitometric method observed in clinical conditions may therefore be ascribed to variation in wall thickness or vasoconstriction.

A physiologically-based model for simulation of color vision deficiency.

Color vision deficiency (CVD) affects approximately 200 million people worldwide, compromising the ability of these individuals to effectively perform color and visualization-related tasks. This has a significant impact on their private and professional lives. We present a physiologically-based model for simulating color vision. Our model is based on the stage theory of human color vision and is derived from data reported in electrophysiological studies. It is the first model to consistently handle normal color vision, anomalous trichromacy, and dichromacy in a unified way. We have validated the proposed model through an experimental evaluation involving groups of color vision deficient individuals and normal color vision ones. Our model can provide insights and feedback on how to improve visualization experiences for individuals with CVD. It also provides a framework for testing hypotheses about some aspects of the retinal photoreceptors in color vision deficient individuals.

An efficient naturalness-preserving image-recoloring method for dichromats.

We present an efficient and automatic image-recoloring technique for dichromats that highlights important visual details that would otherwise be unnoticed by these individuals. While previous techniques approach this problem by potentially changing all colors of the original image, causing their results to look unnatural to color vision deficients, our approach preserves, as much as possible, the image's original colors. Our approach is about three orders of magnitude faster than previous ones. The results of a paired-comparison evaluation carried out with fourteen color-vision deficients (CVDs) indicated the preference of our technique over the state-of-the-art automatic recoloring technique for dichromats. When considering information visualization examples, the subjects tend to prefer our results over the original images. An extension of our technique that exaggerates color contrast tends to be preferred when CVDs compared pairs of scientific visualization images. These results provide valuable information for guiding the design of visualizations for color-vision deficients.