A Light, 3D UNet-based Architecture for Fully Automatic Segmentation of Prostate Lesions from T2-MRI Images.

INTRODUCTION: Prostate magnetic resonance imaging (MRI) has been recently integrated into the pathway of diagnosis of prostate cancer (PCa). However, the lack of an optimal contrast-to-noise ratio hinders automatic recognition of suspicious lesions, thus developing a solution for proper delimitation of the tumour and its separation from the healthy parenchyma, which is of primordial importance. METHOD: As a solution to this unmet medical need, we aimed to develop a decision support system based on artificial intelligence, which automatically segments the prostate and any suspect area from the 3D MRI images. We assessed retrospective data from all patients diagnosed with PCa by MRI-US fusion prostate biopsy, who underwent prostate MRI in our department due to a clinical or biochemical suspicion of PCa (n=33). All examinations were performed using a 1.5 Tesla MRI scanner. All images were reviewed by two radiologists, who performed manual segmentation of the prostate and all lesions. A total of 145 augmented datasets were generated. The performance of our fully automated end-to-end segmentation model based on a 3D UNet architecture and trained in two learning scenarios (on 14 or 28 patient datasets) was evaluated by two loss functions. RESULTS: Our model had an accuracy of over 90% for automatic segmentation of prostate and PCa nodules, as compared to manual segmentation. We have shown low complexity networks, UNet architecture with less than five layers, as feasible and to show good performance for automatic 3D MRI image segmentation. A larger training dataset could further improve the results. CONCLUSION: Therefore, herein, we propose a less complex network, a slim 3D UNet with superior performance, being faster than the original five-layer UNet architecture.

Fully automated bladder tumor segmentation from T2 MRI images using 3D U-Net algorithm.

Introduction: Bladder magnetic resonance imaging (MRI) has been recently integrated in the diagnosis pathway of bladder cancer. However, automatic recognition of suspicious lesions is still challenging. Thus, development of a solution for proper delimitation of the tumor and its separation from the healthy tissue is of primordial importance. As a solution to this unmet medical need, we aimed to develop an artificial intelligence-based decision support system, which automatically segments the bladder wall and the tumor as well as any suspect area from the 3D MRI images. Materials: We retrospectively assessed all patients diagnosed with bladder cancer, who underwent MRI at our department (n=33). All examinations were performed using a 1.5 Tesla MRI scanner. All images were reviewed by two radiologists, who performed manual segmentation of the bladder wall and all lesions. First, the performance of our fully automated end-to-end segmentation model based on a 3D U-Net architecture (by considering various depths of 4, 5 or 6 blocks) trained in two data augmentation scenarios (on 5 and 10 augmentation datasets per original data, respectively) was tested. Second, two learning setups were analyzed by training the segmentation algorithm with 7 and 14 MRI original volumes, respectively. Results: We obtained a Dice-based performance over 0.878 for automatic segmentation of bladder wall and tumors, as compared to manual segmentation. A larger training dataset using 10 augmentations for 7 patients could further improve the results of the U-Net-5 model (0.902 Dice coefficient at image level). This model performed best in terms of automated segmentation of bladder, as compared to U-Net-4 and U-Net-6. However, in this case increased time for learning was needed as compared to U-Net-4. We observed that an extended dataset for training led to significantly improved segmentation of the bladder wall, but not of the tumor. Conclusion: We developed an intelligent system for bladder tumors automated diagnostic, that uses a deep learning model to segment both the bladder wall and the tumor. As a conclusion, low complexity networks, with less than five-layers U-Net architecture are feasible and show good performance for automatic 3D MRI image segmentation in patients with bladder tumors.

On the use of multi-objective evolutionary classifiers for breast cancer detection.

PURPOSE: Breast cancer is one of the most common tumours in women, nevertheless, it is also one of the cancers that is most usually treated. As a result, early detection is critical, which can be accomplished by routine mammograms. This paper aims to describe, analyze, compare and evaluate three image descriptors involved in classifying breast cancer images from four databases. APPROACH: Multi-Objective Evolutionary Algorithms (MOEAs) prove themselves as being efficient methods for selection and classification problems. This paper aims to study combinations of well-known classification objectives in order to compare the results of their application in solving very specific learning problems. The experimental results undergo empirical analysis which is supported by a statistical approach. The results are illustrated on a collection of medical image databases, but with a focus on the MOEAs' performance in terms of several well-known measures. The databases were chosen specifically to feature reliable human annotations, so as to measure the correlation between the gold standard classifications and the various MOEA classifications. RESULTS: We have seen how different statistical tests rank one algorithm over the others in our set as being better. These findings are unsurprising, revealing that there is no single gold standard for comparing diverse techniques or evolutionary algorithms. Furthermore, building meta-classifiers and evaluating them using a single, favorable metric is both extremely unwise and unsatisfactory, as the impact is to skew the results. CONCLUSIONS: The best method to address these flaws is to select the right set of objectives and criteria. Using accuracy-related objectives, for example, is directly linked to maximizing the number of true positives. If, on the other hand, accuracy is chosen as the generic metric, the primary classification goal is shifted to increasing the positively categorized data points.

SERS liquid biopsy in breast cancer. What can we learn from SERS on serum and urine?

SERS analysis of biofluids, coupled with classification algorithms, has recently emerged as a candidate for point-of-care medical diagnosis. Nonetheless, despite the impressive results reported in the literature, there are still gaps in our knowledge of the biochemical information provided by the SERS analysis of biofluids. Therefore, by a critical assignment of the SERS bands, our work aims to provide a systematic analysis of the molecular information that can be achieved from the SERS analysis of serum and urine obtained from breast cancer patients and controls. Further, we compared the relative performance of five different machine learning algorithms for breast cancer and control samples classification based on the serum and urine SERS datasets, and found comparable classification accuracies in the range of 61-89%. This result is not surprising since both biofluids show striking similarities in their SERS spectra providing similar metabolic information, related to purine metabolites. Lastly, by carefully comparing the two datasets (i.e., serum and urine) we show that it is possible to link the misclassified samples to specific metabolic imbalances, such as carotenoid levels, or variations in the creatinine concentration.

A Transfer Learning Approach on the Optimization of Edge Detectors for Medical Images Using Particle Swarm Optimization.

Edge detection is a fundamental image analysis task, as it provides insight on the content of an image. There are weaknesses in some of the edge detectors developed until now, such as disconnected edges, the impossibility to detect branching edges, or the need for a ground truth that is not always accessible. Therefore, a specialized detector that is optimized for the image particularities can help improve edge detection performance. In this paper, we apply transfer learning to optimize cellular automata (CA) rules for edge detection using particle swarm optimization (PSO). Cellular automata provide fast computation, while rule optimization provides adaptability to the properties of the target images. We use transfer learning from synthetic to medical images because expert-annotated medical data is typically difficult to obtain. We show that our method is tunable for medical images with different properties, and we show that, for more difficult edge detection tasks, batch optimization can be used to boost the quality of the edges. Our method is suitable for the identification of structures, such as cardiac cavities on medical images, and could be used as a component of an automatic radiology decision support tool.

Butterfly Effect in Chaotic Image Segmentation.

The exploitation of the important features exhibited by the complex systems found in the surrounding natural and artificial space will improve computational model performance. Therefore, the purpose of the current paper is to use cellular automata as a tool simulating complexity, able to bring forth an interesting global behaviour based only on simple, local interactions. We show that, in the context of image segmentation, a butterfly effect arises when we perturb the neighbourhood system of a cellular automaton. Specifically, we enhance a classical GrowCut cellular automaton with chaotic features, which are also able to improve its performance (e.g., a Dice coefficient of 71% in case of 2D images). This enhanced GrowCut flavor (referred to as Band-Based GrowCut) uses an extended, stochastic neighbourhood, in which randomly-selected remote neighbours reinforce the standard local ones. We demonstrate the presence of the butterfly effect and an increase in segmentation performance by numerical experiments performed on synthetic and natural images. Thus, our results suggest that, by having small changes in the initial conditions of the performed task, we can induce major changes in the final outcome of the segmentation.

Predictability in cellular automata.

Modelled as finite homogeneous Markov chains, probabilistic cellular automata with local transition probabilities in (0, 1) always posses a stationary distribution. This result alone is not very helpful when it comes to predicting the final configuration; one needs also a formula connecting the probabilities in the stationary distribution to some intrinsic feature of the lattice configuration. Previous results on the asynchronous cellular automata have showed that such feature really exists. It is the number of zero-one borders within the automaton's binary configuration. An exponential formula in the number of zero-one borders has been proved for the 1-D, 2-D and 3-D asynchronous automata with neighborhood three, five and seven, respectively. We perform computer experiments on a synchronous cellular automaton to check whether the empirical distribution obeys also that theoretical formula. The numerical results indicate a perfect fit for neighbourhood three and five, which opens the way for a rigorous proof of the formula in this new, synchronous case.

Probabilistic cellular automata.

Cellular automata are binary lattices used for modeling complex dynamical systems. The automaton evolves iteratively from one configuration to another, using some local transition rule based on the number of ones in the neighborhood of each cell. With respect to the number of cells allowed to change per iteration, we speak of either synchronous or asynchronous automata. If randomness is involved to some degree in the transition rule, we speak of probabilistic automata, otherwise they are called deterministic. With either type of cellular automaton we are dealing with, the main theoretical challenge stays the same: starting from an arbitrary initial configuration, predict (with highest accuracy) the end configuration. If the automaton is deterministic, the outcome simplifies to one of two configurations, all zeros or all ones. If the automaton is probabilistic, the whole process is modeled by a finite homogeneous Markov chain, and the outcome is the corresponding stationary distribution. Based on our previous results for the asynchronous case-connecting the probability of a configuration in the stationary distribution to its number of zero-one borders-the article offers both numerical and theoretical insight into the long-term behavior of synchronous cellular automata.

Game theory and extremal optimization for community detection in complex dynamic networks.

The detection of evolving communities in dynamic complex networks is a challenging problem that recently received attention from the research community. Dynamics clearly add another complexity dimension to the difficult task of community detection. Methods should be able to detect changes in the network structure and produce a set of community structures corresponding to different timestamps and reflecting the evolution in time of network data. We propose a novel approach based on game theory elements and extremal optimization to address dynamic communities detection. Thus, the problem is formulated as a mathematical game in which nodes take the role of players that seek to choose a community that maximizes their profit viewed as a fitness function. Numerical results obtained for both synthetic and real-world networks illustrate the competitive performance of this game theoretical approach.