Deep transfer learning based model for colorectal cancer histopathology segmentation: A comparative study of deep pre-trained models.

Colorectal cancer is one of the leading causes of cancer-related death, worldwide. Early detection of suspicious tissues can significantly improve the survival rate. In this study, the performance of a wide variety of deep learning-based architectures is evaluated for automatic tumor segmentation of colorectal tissue samples. The proposed approach highlights the utility of incorporating convolutional neural network modules and transfer learning in the encoder part of a segmentation architecture for histopathology image analysis. A comparative and extensive experiment was conducted on a challenging histopathological segmentation task to demonstrate the effectiveness of incorporating deep modules in the segmentation encoder-decoder network as well as the contributions of its components. Experimental results demonstrate that shared DenseNet and LinkNet architecture is promising, achieves the state-of-the-art performance, and outperforms other methods with a dice similarity index of 82.74%±1.77, accuracy of 87.07%±1.56, and f1-score value of 82.79%±1.79.

Tikhonov gamma variate adaptive regularization applied to technetium Tc 99m diethylenetriamine pentaacetic acid plasma clearance, compared with three other methods, for measuring glomerular filtration rate in cats.

OBJECTIVE: To evaluate agreement of 4 methods (Tikhonov gamma variate adaptive regularization of plasma concentration-time curve fitting applied to technetium Tc 99m diethylenetriamine pentaacetic acid [99mTc-DTPA] plasma clearance [Tk-GV], plasma clearance of exogenous creatinine [CrCL], Gates gamma camera-based measurement method with 99mTc-DTPA renal clearance and dynamic scintigraphy [GTS], and iohexol renal clearance assessed with dynamic CT with Patlak plotting [CT-Pp]) for measuring glomerular filtration rates (GFR) in healthy cats. ANIMALS: 7 healthy, laboratory-raised cats. PROCEDURES: Each method for measuring GFR was performed twice in 7 cats at 24-day intervals. The Wilcoxon signed-rank sum test was used to compare the results obtained from the 14 studies for each method. Results from the 4 methods were assessed for agreement and correlation. RESULTS: The median GFR values were 2.75, 2.83, 3.14, and 4.26 mL/min/kg, for Tk-GV, CT-Pp, plasma CrCL, and GTS, respectively. Analysis with Wilcoxon signed-rank sum tests identified significant pairwise differences between results obtained with the Tk-GV versus the plasma CrCL method, the Tk-GV versus the GTS method, and the plasma CrCL versus the GTS method. The least variable method was Tk-GV, with an SD of 1.27 (mL/min/kg). CONCLUSIONS AND CLINICAL RELEVANCE: Findings indicated that Tk-GV yielded GFR measurements comparable with those obtained with CT-Pp, plasma CrCL, and GTS; however, the Tk-GV method yielded the tightest range of results among the methods evaluated.

A gamma-distribution convolution model of 99mTc-MIBI thyroid time-activity curves.

BACKGROUND: The convolution approach to thyroid time-activity curve (TAC) data fitting with a gamma distribution convolution (GDC) TAC model following bolus intravenous injection is presented and applied to 99mTc-MIBI data. The GDC model is a convolution of two gamma distribution functions that simultaneously models the distribution and washout kinetics of the radiotracer. The GDC model was fitted to thyroid region of interest (ROI) TAC data from 1 min per frame 99mTc-MIBI image series for 90 min; GDC models were generated for three patients having left and right thyroid lobe and total thyroid ROIs, and were contrasted with washout-only models, i.e., less complete models. GDC model accuracy was tested using 10 Monte Carlo simulations for each clinical ROI. RESULTS: The nine clinical GDC models, obtained from least counting error of counting, exhibited corrected (for 6 parameters) fit errors ranging from 0.998% to 1.82%. The range of all thyroid mean residence times (MRTs) was 212 to 699 min, which from noise injected simulations of each case had an average coefficient of variation of 0.7% and a not statistically significant accuracy error of 0.5% (p = 0.5, 2-sample paired t test). The slowest MRT value (699 min) was from a single thyroid lobe with a tissue diagnosed parathyroid adenoma also seen on scanning as retained marker. The two total thyroid ROIs without substantial pathology had MRT values of 278 and 350 min overlapping a published 99mTc-MIBI thyroid MRT value. One combined value and four unrelated washout-only models were tested and exhibited R-squared values for MRT with the GDC, i.e., a more complete concentration model, ranging from 0.0183 to 0.9395. CONCLUSIONS: The GDC models had a small enough TAC noise-image misregistration (0.8%) that they have a plausible use as simulations of thyroid activity for querying performance of other models such as washout models, for altered ROI size, noise, administered dose, and image framing rates. Indeed, of the four washout-only models tested, no single model approached the apparent accuracy of the GDC model using only 90 min of data. Ninety minutes is a long gamma-camera acquisition time for a patient, but a short a time for most kinetic models. Consequently, the results should be regarded as preliminary.

Time Varying Apparent Volume of Distribution and Drug Half-Lives Following Intravenous Bolus Injections.

We present a model that generalizes the apparent volume of distribution and half-life as functions of time following intravenous bolus injection. This generalized model defines a time varying apparent volume of drug distribution. The half-lives of drug remaining in the body vary in time and become longer as time elapses, eventually converging to the terminal half-life. Two example fit models were substituted into the general model: biexponential models from the least relative concentration error, and gamma variate models using adaptive regularization for least relative error of clearance. Using adult population parameters from 41 studies of the renal glomerular filtration marker 169Yb-DTPA, simulations of extracellular fluid volumes of 5, 10, 15 and 20 litres and plasma clearances of 40 and 100 ml/min were obtained. Of these models, the adaptively obtained gamma variate models had longer times to 95% of terminal volume and longer half-lives.

Demonstration of synchrotron x-ray phase contrast imaging computed tomography of infiltrative transitional cell carcinoma of the prostatic urethra in a dog.

Prostatic urethral transitional cell carcinoma with prostatic invasion in a dog was imaged with abdominal radiography and abdominal ultrasonography antemortem. Synchrotron in-line x-ray phase contrast imaging computed tomography (XPCI-CT) was performed on the prostate ex vivo at the Canadian Light Source Synchrotron and compared to histology. XPCI-CT imaging provides greater soft tissue contrast than conventional absorption-based x-ray imaging modalities, permitting visualization of regions of inflammatory cell infiltration, differentiation of invasive versus noninvasive tumor regions, and areas of necrosis and mineralization. This represents the first report of XPCI-CT images of an invasive prostatic urothelial neoplasm in a dog.

Accurate and precise plasma clearance measurement using four 99mTc-DTPA plasma samples over 4 h.

OBJECTIVES: Glomerular filtration rate can be measured as the plasma clearance (CL) of a glomerular filtration rate marker despite body fluid disturbances using numerous, prolonged time samples. We desire a simplified technique without compromised accuracy and precision. MATERIALS AND METHODS: We compared CL values derived from two plasma concentration curve area methods - (a) biexponential fitting [CL (E2)] and (b) Tikhonov adaptively regularized gamma variate fitting [CL (Tk-GV)] - for 4 versus 8 h time samplings from 412 Tc-DTPA studies in 142 patients, mostly paediatric patients, with suspected fluid disturbances. RESULTS: CL (Tk-GV) from four samples/4 h and from nine samples/8 h, both accurately and precisely agreed with the standard, which was taken to be nine samples/8 h CL from (noncompartmental) numerical integration [CL (NI)]. The E2 method, four samples/4 h, and nine samples/8 h median CL values significantly overestimated the CL (NI) values by 4.9 and 3.8%, respectively. CONCLUSION: Compared with the standard, CL (E2) from four samples/4 h and from nine samples/8 h proved to be the most inaccurate and imprecise method examined, and can be replaced by better methods for calculating CL. The CL (Tk-GV) can be used to reduce sampling time in half from 8 to 4 h and from nine to four samples for a precise and accurate, yet more easily tolerated and simplified test.

A simple method for determining split renal function from dynamic (99m)Tc-MAG3 scintigraphic data.

PURPOSE: Commonly used methods for determining split renal function (SRF) from dynamic scintigraphic data require extrarenal background subtraction and additional correction for intrarenal vascular activity. The use of these additional regions of interest (ROIs) can produce inaccurate results and be challenging, e.g. if the heart is out of the camera field of view. The purpose of this study was to evaluate a new method for determining SRF called the blood pool compensation (BPC) technique, which is simple to implement, does not require extrarenal background correction and intrinsically corrects for intrarenal vascular activity. METHODS: In the BPC method SRF is derived from a parametric plot of the curves generated by one blood-pool and two renal ROIs. Data from 107 patients who underwent (99m)Tc-MAG3 scintigraphy were used to determine SRF values. Values calculated using the BPC method were compared to those obtained with the integral (IN) and Patlak-Rutland (PR) techniques using Bland-Altman plotting and Passing-Bablok regression. The interobserver variability of the BPC technique was also assessed for two observers. RESULTS: The SRF values obtained with the BPC method did not differ significantly from those obtained with the PR method and showed no consistent bias, while SRF values obtained with the IN method showed significant differences with some bias in comparison to those obtained with either the PR or BPC method. No significant interobserver variability was found between two observers calculating SRF using the BPC method. CONCLUSION: The BPC method requires only three ROIs to produce reliable estimates of SRF, was simple to implement, and in this study yielded statistically equivalent results to the PR method with appreciable interobserver agreement. As such, it adds a new reliable method for quality control of monitoring relative kidney function.

The early plasma concentration of 51Cr-EDTA in patients with cirrhosis and ascites: a comparison of three models.

OBJECTIVES: The aim of the study was to determine which of three two-parameter fitting functions (exponential, linear-log, and negative-power function of time) most accurately models early chromium-51-EDTA (51Cr-EDTA) plasma concentration data prior to 120 min in patients with cirrhosis and ascites and understand how these fitting functions affect the calculation of the area under the plasma concentration curve (AUC). METHODS: A bolus, antecubital intravenous injection of 2.6 MBq of 51Cr-EDTA was given to 13 patients with cirrhosis and ascites. Up to 16 blood samples were drawn at time points ranging from 5 to 1440 min following injection. The concentration data prior to 120 min were used as reference data. Early time concentration values, estimated by fitting exponential, linear-log, and negative-power functions of time to the time samples at 120, 180, and 240 min, were then compared with reference data. The AUC was calculated for each patient using the exponential, Bröchner-Mortensen-corrected exponential, and linear-log functions, and these values were compared. RESULTS: The withheld, observed plasma concentrations were (a) most accurately estimated by linear-log functions (Wilcoxon P=0.4548), (b) significantly underestimated by exponential functions (Wilcoxon P=0.0002), and (c) significantly overestimated by negative-power functions (Wilcoxon P=0.0034). The relative errors when ranked from best to worst were those for the linear-log (12.0%, 9.0%), exponential (22.9%, 14.2%), and negative-power (31.9%, 48.4%) functions of time, respectively (median, interquartile range). For each patient, the values for AUC calculated by the exponential function differed significantly (range=3.4-15.3%, median=8.3%) from those calculated by the corrected Bröchner-Mortensen exponential, as to a lesser extent did those values calculated using linear-log functions (range=0.4-8.0%, median=3.0%). CONCLUSION: In patients with cirrhosis, linear-log functions were significantly more accurate than exponential or power functions in estimating early time plasma concentrations (<120 min). However, the improved linear-log early time plasma concentration model does not provide as much correction to the total AUC as does the corrected Bröchner-Mortensen exponential method. This is likely because of the large contribution of late time data to the AUC, and future work is suggested to explore the late time fit problem.

Nonlinear optical and chemical effects in the irradiation of liquid benzene with femtosecond pulses.

We show that a spectral resonance between the π→π* absorption band in liquid benzene and the third harmonic (TH) of a propagating 800 nm femtosecond laser beam causes large positive changes in the real refractive index at the TH wavelength. This produces an increase in the third-order optical susceptibility and leads to the enhancement of nonlinear optical effects including TH generation and self-focusing. Enhanced filamentation is observed in liquid benzene, but this effect is not seen in perdeuterated liquid benzene under similar irradiation conditions. Filamentation is associated with the decomposition of benzene molecules, plasma emission from the focal region, and the appearance of carbon nanoparticles. This indicates that a complex chemistry accompanies the onset of filamentation. Chemical products formed under these conditions have been characterized using combined gas chromatography mass spectroscopy techniques. We also find that the presence of a TH filament is indicated by the appearance of a photocurrent and increased electrical conductivity in the solution. This photocurrent is found to be 50-60 times smaller in C6D6 where the π→π* resonance with the TH is much weaker. The intensity dependence of this photocurrent confirms the role played by TH generation in the overall interaction.