Florin Gorunescu

Neural network analysis of dynamic sequences of EUS elastography used for the differential diagnosis of chronic pancreatitis and pancreatic cancer

BACKGROUND EUS elastography is a newly developed imaging procedure that characterizes the differences of hardness and strain between diseased and normal tissue. OBJECTIVE To assess the accuracy of real-time EUS elastography in pancreatic lesions. DESIGN Cross-sectional feasibility study. PATIENTS The study group included, in total, 68 patients with normal pancreas (N = 22), chronic pancreatitis (N = 11), pancreatic adenocarcinoma (N = 32), and pancreatic neuroendocrine tumors (N = 3). A subgroup analysis of 43 cases with focal pancreatic masses was also performed. INTERVENTIONS A postprocessing software analysis was used to examine the EUS elastography movies by calculating hue histograms of each individual image, data that were further subjected to an extended neural network analysis to differentiate benign from malignant patterns. MAIN OUTCOME MEASUREMENTS To differentiate normal pancreas, chronic pancreatitis, pancreatic cancer, and neuroendocrine tumors. RESULTS Based on a cutoff of 175 for the mean hue histogram values recorded on the region of interest, the sensitivity, specificity, and accuracy of differentiation of benign and malignant masses were 91.4%, 87.9%, and 89.7%, respectively. The positive and negative predictive values were 88.9% and 90.6%, respectively. Multilayer perceptron neural networks with both one and two hidden layers of neurons (3-layer perceptron and 4-layer perceptron) were trained to learn how to classify cases as benign or malignant, and yielded an excellent testing performance of 95% on average, together with a high training performance that equaled 97% on average. LIMITATION A lack of the surgical standard in all cases. CONCLUSIONS EUS elastography is a promising method that allows characterization and differentiation of normal pancreas, chronic pancreatitis, and pancreatic cancer. The currently developed methodology, based on artificial neural network processing of EUS elastography digitalized movies, enabled an optimal prediction of the types of pancreatic lesions. Future multicentric, randomized studies with adequate power will have to establish the clinical impact of this procedure for the differential diagnosis of focal pancreatic masses.

A queueing model for bed-occupancy management and planning of hospitals

The aim of this paper is, on the one hand, to describe the movement of patients through a hospital department by using classical queueing theory and, on the other hand, to present a way of optimising the use of hospital resources in order to improve hospital care. A queueing model is used to determine the main characteristics of the access of patients to hospital, such as mean bed occupancy and the probability that a demand for hospital care is lost because all beds are occupied. Moreover, we present a technique for optimising the number of beds in order to maintain an acceptable delay probability at a sufficiently low level and, finally, a way of optimising the average cost per day by balancing costs of empty beds against costs of delayed patients.

Randomized controlled trial of endoscopic ultrasound-guided fine-needle sampling with or without suction for better cytological diagnosis.

Objective. Endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA) is a highly accurate method to obtain specific diagnosis in various diseases. The optimal method of EUS-guided sampling of material for pathologic diagnosis has not been clearly established. The aim of our study was to compare two different techniques of EUS-guided sampling of solid masses, using either non-suction or suction with a 10-ml syringe. Material and methods. Patients assessed during a 6-month period were randomized to three passes of EUS-guided sampling with suction (26 patients) or non-suction (26 patients). The samples were characterized for cellularity and bloodiness, with a final cytology diagnosis established blindly. The final diagnosis was reached either by EUS-FNA if malignancy was definite, or by surgery and/or clinical follow-up of a minimum of 6 months in the cases of non-specific benign lesions. Results. EUS-guided fine-needle sampling with suction of solid masses increased the number of pathology slides (17.8±7.1 slides for suction as compared with 10.2±5.5 for non-suction, p=0.0001), without increasing the overall bloodiness of each sample. Sensitivity and the negative predictive values were higher when suction was applied, as compared to the non-suction group (85.7% as compared with 66.7%, p=0.05). Conclusions. This prospective randomized study showed that EUS-guided fine-needle sampling of solid masses using suction yields a higher number of slides without increasing bloodiness. Although, the proportion of target cells was relatively similar between the suction and non-suction sampling techniques, the sensitivity and negative predictive values of the procedure were significantly higher when suction was added.

Using a queueing model to help plan bed allocation in a department of geriatric medicine.

By integrating queuing theory and compartmental models of flow we demonstrate how changing admission rates, length of stay and bed allocation influence bed occupancy, emptiness and rejection in departments of geriatric medicine. By extending the model to include waiting beds, we show how the provision of extra, emergency use, unstaffed, back up beds could improve performance while controlling costs. The model is applicable to all lengths of stay, admission rates and bed allocations. The results show why 10–15% bed emptiness is necessary to maintain service efficiency and demonstrate how unstaffed beds can serve to provide a more responsive and cost effective service. Further work is needed to test the validity and applicability of the model.

Combined contrast-enhanced power Doppler and real-time sonoelastography performed during EUS, used in the differential diagnosis of focal pancreatic masses (with videos).

BACKGROUND Contrast-enhanced power Doppler (CEPD) and real-time sonoelastography (RTSE) performed during EUS were previously described to be useful for the differential diagnosis between chronic pseudotumoral pancreatitis and pancreatic cancer. OBJECTIVE To prospectively assess the accuracy of the combination of CEPD and RTSE to differentiate pancreatic focal masses. DESIGN Cross-sectional feasibility study. SETTING A tertiary-care academic referral center. PATIENTS The study group included 54 patients with chronic pancreatitis (n = 21) and pancreatic adenocarcinoma (n = 33). INTERVENTIONS Both imaging methods (CEPD and RTSE) were performed sequentially during the same EUS examination. Power Doppler mode examination was performed after intravenous injection of a second-generation contrast agent (2.4 mL of SonoVue), and the data were digitally recorded, comprising both the early arterial phase and venous/late phase. Three 10-second sonoelastographic videos were also digitally recorded that included the focal mass and the surrounding pancreatic parenchyma. Postprocessing analyses based on specially designed software were used to analyze the CEPD and RTSE videos. A power Doppler vascularity index was used to characterize CEPD videos, the values being averaged during a 10-second video in the venous phase. Hue histogram analysis was used to characterize RTSE videos, with the mean hue histogram values being also averaged during a 10-second video. MAIN OUTCOME MEASUREMENTS To differentiate chronic pancreatitis and pancreatic cancer. RESULTS The sensitivity, specificity, and accuracy of combined information provided by CEPD and RTSE to differentiate hypovascular hard masses suggestive of pancreatic carcinoma were 75.8%, 95.2%, and 83.3%, respectively, with a positive predictive value and negative predictive value of 96.2% and 71.4%, respectively. LIMITATION A single-center, average size of study population. CONCLUSIONS A combination of CEPD and RTSE performed during EUS seems to be a promising method that allows characterization and differentiation of focal pancreatic masses.

Power Doppler endoscopic ultrasonography for the differential diagnosis between pancreatic cancer and pseudotumoral chronic pancreatitis.

Objective. The accuracy of endoscopic ultrasonography (EUS) and EUS‐guided fine‐needle aspiration for the differential diagnosis of pancreatic masses is variable in the literature, being as low as 75% in some studies. The aim of the study was to assess the accuracy of power Doppler EUS for the differential diagnosis between pancreatic cancer and pseudotumoral chronic pancreatitis. Methods. We included 42 consecutive patients with pancreatic tumor masses (27 men and 15 women) examined by EUS between January 2002 and August 2004. Endoscopic ultrasonographic procedures included power Doppler EUS as well as EUS‐guided fine‐needle aspiration in all patients. Final diagnosis of pancreatic cancer was confirmed in 29 patients on the basis of a combination of information provided by imaging tests, follow‐up of at least 6 months, and laparotomy in 18 patients for diagnostic or palliative reasons. Results. Sensitivity and specificity of the absence of power Doppler signals inside the suggestive pancreatic mass were 93% and 77%, respectively, with accuracy of 88%. Moreover, the addition of the information provided by the presence of peripancreatic collaterals improved the sensitivity and specificity to 97% and 92%, with accuracy of 95%. Conclusions. Power Doppler EUS provides useful information for the differential diagnosis of pancreatic masses. The results were in concordance with previous studies that showed a hypovascular pattern of pancreatic carcinoma, as well as the formation of collaterals in advanced cases due to the invasion of the splenic or portal veins. Further studies of dynamic EUS with contrast agents are necessary to better characterize pancreatic masses.

A machine learning approach to keystroke dynamics based user authentication

The majority of computer systems employ a login ID and password as the principal method for access security. In stand-alone situations, this level of security may be adequate, but when computers are connected to the internet, the vulnerability to a security breach is increased. In order to reduce vulnerability to attack, biometric solutions have been employed. In this paper, we investigate the use of a behavioural biometric based on keystroke dynamics. Although there are several implementations of keystroke dynamics available, their effectiveness is variable and dependent on the data sample and its acquisition methodology. The results from this study indicate that the Equal Error Rate (EER) is significantly influenced by the attribute selection process and to a lesser extent on the authentication algorithm employed. Our results also provide evidence that a Probabilistic Neural Network (PNN) can be superior in terms of reduced training time and classification accuracy when compared with a typical MLFN back-propagation trained neural network.

A Breast Cancer Diagnosis System: A Combined Approach Using Rough Sets and Probabilistic Neural Networks

In this paper, we present a medical decision support system based on a hybrid approach utilizing rough sets and a probabilistic neural network. We utilized the ability of rough sets to perform dimensionality reduction to eliminate redundant attributes from a biomedical dataset. We then utilized a probabilistic neural network to perform supervised classification. Our results indicate that rough sets were able to reduce the number of attributes in the dataset by 67% without sacrificing classification accuracy. Our classification accuracy results yielded results on the order of 93%

Multicenter randomized controlled trial comparing the performance of 22 gauge versus 25 gauge EUS–FNA needles in solid masses

Abstract Background and aims. Few randomized studies have assessed the clinical performance of 25-gauge (25G) needles compared with 22-gauge (22G) needles during endoscopic ultrasound-guided fine needle aspiration (EUS–FNA) biopsy of intra-abdominal lesions. We aimed to compare the diagnostic yield, as well as performance characteristics of 22G versus 25G EUS biopsy needles by determining their diagnostic capabilities, the number of needle passes as well as cellularity of aspirated tissue specimen. Methods. The study is a prospective, randomized, multicenter study. Patients were referred between January 2009 and January 2010 for diagnostic EUS including EUS-guided FNA of different lesions adjacent to the upper GI tract. All patients were randomized to EUS–FNA performed with either a 22G or 25G aspiration needle. Results. EUS–FNA was performed in 135 patients (62 patients with a 22G needle). Sensitivity and specificity of the 22G needle was 94.1% and 95.8%, respectively, and for the 25G needle 94.1% and 100%, respectively. Investigators reported better visualization and performance for the 22G needle compared to the 25G (p < 0.0001). The number of tissue slides obtained was higher for the 22G needle during the second and third needle passes (p < 0.05). We did not observe significant differences between the number and preservation status of obtained cells (p > 0.05). Conclusions. A significant difference was found between the two types of needles in terms of reduced visualization of the 25G needle and suboptimal performance rating. However, this did not impact on overall results since both needles were equally successful in terms of a high diagnostic yield and overall accuracy.

A hybrid neural network/genetic algorithm applied to breast cancer detection and recurrence

Genetic algorithms (GAs) and neural networks (NNs) are both inspired by computation in biological systems and many attempts have been made to combine the two methodologies to boost the NNs performance. This paper deals with the evolutionary training of a feedforward NN for both breast cancer detection and recurrence. A multi-layer perceptron (MLP) has been designed for this purpose, using a GA routine to set weights, and a Java implementation of this hybrid model has been made. Four databases concerning cancer detection and recurrence have been used, two databases containing numerical attributes only, one database containing ordinal (categorical) attributes solely and one database with mixed attributes. In comparison to some standard NNs, the performance of this approach using the same databases is shown to be superior. Moreover, this hybrid MLP/GA model is very flexible in terms of providing accurate classification, even with different types of attributes, which is usually found in medical studies.