Image recognition neural network: IRNN

Machine learning in analytical spectroscopy for nuclear diagnostics [Invited]

Analytical spectroscopy methods have shown many possible uses for nuclear material diagnostics and measurements in recent studies. In particular, the application potential for various atomic spectroscopy techniques is uniquely diverse and generates interest across a wide range of nuclear science areas. Over the last decade, techniques such as laser-induced breakdown spectroscopy, Raman spectroscopy, and x-ray fluorescence spectroscopy have yielded considerable improvements in the diagnostic analysis of nuclear materials, especially with machine learning implementations. These techniques have been applied for analytical solutions to problems concerning nuclear forensics, nuclear fuel manufacturing, nuclear fuel quality control, and general diagnostic analysis of nuclear materials. The data yielded from atomic spectroscopy methods provide innovative solutions to problems surrounding the characterization of nuclear materials, particularly for compounds with complex chemistry. Implementing these optical spectroscopy techniques can provide comprehensive new insights into the chemical analysis of nuclear materials. In particular, recent advances coupling machine learning methods to the processing of atomic emission spectra have yielded novel, robust solutions for nuclear material characterization. This review paper will provide a summation of several of these recent advances and will discuss key experimental studies that have advanced the use of analytical atomic spectroscopy techniques as active tools for nuclear diagnostic measurements.

Development of advanced machine learning models for analysis of plutonium surrogate optical emission spectra

This work investigates and applies machine learning paradigms seldom seen in analytical spectroscopy for quantification of gallium in cerium matrices via processing of laser-plasma spectra. Ensemble regressions, support vector machine regressions, Gaussian kernel regressions, and artificial neural network techniques are trained and tested on cerium-gallium pellet spectra. A thorough hyperparameter optimization experiment is conducted initially to determine the best design features for each model. The optimized models are evaluated for sensitivity and precision using the limit of detection (LoD) and root mean-squared error of prediction (RMSEP) metrics, respectively. Gaussian kernel regression yields the superlative predictive model with an RMSEP of 0.33% and an LoD of 0.015% for quantification of Ga in a Ce matrix. This study concludes that these machine learning methods could yield robust prediction models for rapid quality control analysis of plutonium alloys.

Effects of dataset size and interactions on the prediction performance of logistic regression and deep learning models

BACKGROUND AND OBJECTIVE

Machine learning and deep learning models are very powerful in predicting the presence of a disease. To achieve good predictions, those models require a certain amount of data to train on, whereas this amount i) is generally limited and difficult to obtain; and, ii) increases with the complexity of the interactions between the outcome (disease presence) and the model variables. This study compares the ways training dataset size and interactions affect the performance of those prediction models.

METHODS

To compare the two influences, several datasets were simulated that differed in the number of observations and the complexity of the interactions between the variables and the outcome. A few logistic regressions and neural networks were trained on the simulated datasets and their performance evaluated by cross-validation and compared using accuracy, F1 score, and AUC metrics.

RESULTS

Models trained on simulated datasets without interactions provided good results: AUC close to 0.80 with either logistic regression or neural networks. Models trained on simulated dataset with order 2 interactions led also to AUCs close to 0.80 with either logistic regression or neural networks. Models trained on simulated datasets with order 4 interactions led to AUC close to 0.80 with neural networks and 0.85 with penalized logistic regressions. Whatever the interaction order, increasing the dataset size did not significantly affect model performance, especially that of machine learning models.

CONCLUSION

Machine learning models were the less influenced by the dataset size but needed interaction terms to achieve good performance, whereas deep learning models could achieve good performance without interaction terms. Conclusively, with the considered scenarios, well-specified machine learning models outperformed deep learning models.

Recognition of partially occluded and rotated images with a network of spiking neurons

In this paper, we introduce a novel system for recognition of partially occluded and rotated images. The system is based on a hierarchical network of integrate-and-fire spiking neurons with random synaptic connections and a novel organization process. The network generates integrated output sequences that are used for image classification. The proposed network is shown to provide satisfactory predictive performance given that the number of the recognition neurons and synaptic connections are adjusted to the size of the input image. Comparison of synaptic plasticity activity rule (SAPR) and spike timing dependant plasticity rules, which are used to learn connections between the spiking neurons, indicates that the former gives better results and thus the SAPR rule is used. Test results show that the proposed network performs better than a recognition system based on support vector machines.

Identification of complex shapes using a self organizing neural system

We present a multilayer hierarchical neural system for automatic classification of complex contour patterns. The system consists of a neocognitron-like network structure combined with self-organizing maps to automatically determine feature classes. We present results showing that multilayer hierarchical networks are able to tolerate pattern distortion considerably better than standard neural network implementations.

A knowledge-guided active model method of cortical structure segmentation on pediatric MR images

PURPOSE

To develop an automated method for quantification of cortical structures on pediatric MR images.

MATERIALS AND METHODS

A knowledge-guided active model (KAM) approach was proposed with a novel object function similar to the Gibbs free energy function. Triangular mesh models were transformed to images of a given subject by maximizing entropy, and then actively slithered to boundaries of structures by minimizing enthalpy. Volumetric results and image similarities of 10 different cortical structures segmented by KAM were compared with those traced manually. Furthermore, the segmentation performances of KAM and SPM2, (statistical parametric mapping, a MATLAB software package) were compared.

RESULTS

The averaged volumetric agreements between KAM- and manually-defined structures (both 0.95 for structures in healthy children and children with medulloblastoma) were higher than the volumetric agreement for SPM2 (0.90 and 0.80, respectively). The similarity measurements (kappa) between KAM- and manually-defined structures (0.95 and 0.93, respectively) were higher than those for SPM2 (both 0.86).

CONCLUSION

We have developed a novel automatic algorithm, KAM, for segmentation of cortical structures on MR images of pediatric patients. Our preliminary results indicated that when segmenting cortical structures, KAM was in better agreement with manually-delineated structures than SPM2. KAM can potentially be used to segment cortical structures for conformal radiation therapy planning and for quantitative evaluation of changes in disease or abnormality.

Automated human frontal lobe identification in MR images based on fuzzy-logic encoded expert anatomic knowledge

Identification of human brain structures in MR images comprises an area of increasing interest, which also presents numerous methodological challenges. Here we describe a new knowledge-based automated method designed to identify several major brain sulci and then to define the frontal lobes by using the identified sulci as landmarks. To identify brain sulci, sulcal images were generated by morphologic operations and then separated into different components based on connectivity analysis. Subsequently, the individual anatomic features were evaluated by using fuzzy membership functions. The crisp decisions, i.e., the identification of sulci, were made by taking the maximum of the summation of all the membership functions. The identification was designed in a hierarchical order. The longitudinal fissure was extracted first. The left and right central sulci were then identified based on the left and right hemispheres. Next, the lateral sulci were identified based on the central sulci and hemispheres. Finally, the left and right frontal lobes were defined from the two hemispheres. The method was evaluated by visual inspection, comparison with manual segmentation, and comparison with manually volumetric results in references. The average Jaccard similarities of left and right frontal lobes between the automated and manual segmentation were 0.89 and 0.91, respectively. The average Kappa indices of left and right frontal lobes between the automated and manual segmentation were 0.94 and 0.95, respectively. These results show relatively high accuracy of using this novel method for human frontal lobe identification and segmentation.

A knowledge-guided active contour method of segmentation of cerebella on MR images of pediatric patients with medulloblastoma

PURPOSE

To develop an automated method for identification of the cerebella on magnetic resonance (MR) images of patients with medulloblastoma.

MATERIALS AND METHODS

The method used a template constructed from 10 patients' aligned MR head images, and the contour of this template was superimposed on the aligned data set of a given patient as the starting contour. The starting contour was then actively adjusted to locate the boundary of the cerebellum of the given patient. Morphologic operations were applied to the outlined volume to generate cerebellum images. The method was then applied to data sets of 20 other patients to generate cerebellum images and volumetric results.

RESULTS

Comparison of the automatically generated cerebellum images with two sets of manually traced images showed a strong correlation between the automatically and manually generated volumetric results (correlation coefficient, 0.97). The average Jaccard similarities were 0.89 and 0.88 in comparison to each of two manually traced images, respectively. The same comparisons yielded average kappa indexes of 0.94 and 0.93, respectively.

CONCLUSION

The method was robust and accurate for cerebellum segmentation on MR images of patients with medulloblastoma. The method may be applied to investigations that require segmentation and quantitative measurement of MR images of the cerebellum.

Classification of reflectance spectra from pigmented skin lesions, a comparison of multivariate discriminant analysis and artificial neural networks

Successful treatment of skin cancer, especially melanoma, depends on early detection, but diagnostic accuracy, even by experts, can be as low as 56% so there is an urgent need for a simple, accurate, non-invasive diagnostic tool. In this paper we have compared the performance of an artificial neural network (ANN) and multivariate discriminant analysis (MDA) for the classification of optical reflectance spectra (320 to 1100 nm) from malignant melanoma and benign naevi. The ANN was significantly better than MDA, especially when a larger data set was used, where the classification accuracy was 86.7% for ANN and 72.0% for MDA (p < 0.001). ANN was better at learning new cases than MDA for this particular classification task. This study has confirmed that the convenience of ANNs could lead to the medical community and patients benefiting from the improved diagnostic performance which can be achieved by objective measurement of pigmented skin lesions using spectrophotometry.