Many explainable AI methods for generating medical image saliency maps exist, but most are devoted to working on trained neural network-based models. At the same time, many medical image classification neural networks use as input radiomic features derived from images. The mathematics for radiomic feature computations are not always represented by a differentiable function, which makes it impossible to apply existing saliency map methods to obtain input image pixel attributions because they heavily rely on gradient calculation possibility. For this reason, a novel method (SRFAMap) is introduced to map the statistical radiomic feature attributions derived by applying the Integrated Gradients methods, often used in explainable AI, to image saliency maps. In detail, integrated gradients are used to compute radiomic feature attributions of chest X-ray scans for a ResNet-50 convolutional network model trained to distinguish healthy lungs from tuberculosis lesions. These are subsequently mapped to saliency maps over the original scans to facilitate their interpretation for diagnostics. Findings show that, in most cases, SRFAMap can generate saliency maps with an acceptable level of faithfulness. The increase-in-confidence metric reached at least 34%, while the Average Drop reached 38% at most. Finally, the percentage of the statistically significant target class score increases reached at least 71% on 20 random folds for the grey-level co-occurrence matrix and higher for other methods when only pixels under positive saliency map values were revealed on the blurred image. The main contribution is a method of mapping integrated gradients to saliency maps to facilitate the visual interpretation of the relevant statistical radiomic features of X-ray scans of lungs responsible for discriminating types of lesions.

The work that lies behind this Working Paper began, for Professor Mati, long before she began working with IWMI.But for IWMI, it also has a history of several years.During the years 2000-2001, as we built up our program and staff in Africa, we spent quite some time trying to identify what should be our main priorities in sub-Saharan Africa.By 2002 we realized that one very important focus for IWMI's work ought to be on water and land management in "rain-fed" as well as "irrigated" agriculture; we called our incipient effort "intensifying rain-fed agriculture," and focused on such topics as rainwater harvesting, small individualized technologies such as bucket and drum drip kits and pedal pumps, and low-cost water storage.As part of our effort to understand what is already happening in this arena, we asked Professor Mati to work with us on this topic and among other things to provide an overview of experiences in East Africa based on her ongoing and past experiences.The first draft of this paper was produced in 2003.She continued to work on it, adding examples, illustrations, data from more recent experiences, and for the final version, added a section on participatory approaches to implementation and scaling up.She has patiently answered questions from nonspecialists like me, and worked hard to finalize this document for wider dissemination.I am sure that many people will find this overview a useful guide and menu.It is aimed at government officials, NGOs, and donors who are supporting the implementation of improved lowcost water and land management practices by and with poor farmers.I want, therefore, to thank Professor Mati for her work.I share Professor Mati's belief that helping farmers adapt and make good use of better but low-cost water and land management practices and technologies can have a high impact on poverty and food security in Africa.As will be seen from her acknowledgments, many people have participated in various ways to make this Working Paper a reality.

1 Оптическая диагностика многолуночных планшетов с реагирующей смесью при производстве 1 Институт теплофизики им.С. С. Кутателадзе СО РАН, г.Новосибирск, Российская Федерация * e-mail:

топливно-воздушной струи1 Институт теплофизики им.С. С. Кутателадзе СО РАН,

This study introduces an optimization of the voting function’s structure and parameters within a self-organizing forest of decision trees, based on an enhanced stepwise logistic regression algorithm and a positional voting method. The stepwise regression algorithm is improved through GMDH-based optimization for selecting threshold values for significance levels, both for the inclusion and exclusion of voting function arguments. The voting function’s arguments are formed using the positional method from branches of self-organizing trees and are supplemented by the forest’s input arguments. Classification results obtained from the standard version of Random Forest, the self-organizing forest, and the proposed enhanced version of the decision tree forest are compared.

Global texture characteristics are powerful tools for solving medical image classification tasks. There are many such characteristics like Grey-Level Co-occurrence Matrices, Grey-Level Run-Length Matrices, Grey-Level Size Zone Matrices, texture matrices and others. However, not all are important when solving particular image classification tasks, while their calculation requires many computational resources. The current work aims to evaluate the importance of each characteristic, taking into account a large dimensionality of the texture characteristics matrices. To achieve this aim, it is proposed to use neural networks and a novel mean integrated gradient eXplainable Artificial Intelligence method to achieve the stated aim. The experiment showed that texture matrices with higher mean integrated gradient values are more important than others while solving pneumonia lesions classification tasks on X-Ray lung images. The result also indicates that classification quality does not degrade and even improves after shrinking the feature set with the proposed method. These facts prove that the mean integrated gradients can be used for solving feature selection tasks for classification purposes.

The authors propose an approach to the construction of classifiers in the class of Random Forest algorithms. A genetic algorithm is used to determine the optimal combination and composition of ensembles of features in the construction of forest trees. The principles of the group method of data handling are used to optimize the structure of the trees. Optimization of the tree voting procedure in the forest is implemented by the analytic hierarchy process. Examples of the use of the proposed algorithm for the detection of pathologies on medical images are provided, as well as the classification results in comparison with other known analogs.

The current study considers the development of a 5-layer pipeline for identifying and classifying COVID-19-induced lung lesions. Such system is multilayer, built upon convolutional and fully connected neural networks and logistic self-organised forest built using the group method of data handling (GMDH) principles. This pipeline includes a mechanism for finding lesions regions in lungs computer tomography images and for calculating related lung damage volume. The layer for finding images with lesions reached a Matthews Correlation Coefficient of 0.98. The layer for lesions segmentation reached a Dice similarity coefficient of 0.74, while the layer for lesions classification reached Fl-scores of 1, 0.95, 0.93 respectively for the ground-glass, opacity, crazy-paving and consolidation lesion type. Results demonstrate the effectiveness of the implemented multi-layer system in solving tasks of lesions identification and classification while being composed into a single pipeline.

Стаття присвячена аналізу лексичного рівня офіційно-ділового стилю юридичної документації. Офіційно-ділова документація характеризується своєрідністю лексики, зокрема термінологічної, яка потребує постійного дослідження. Актуальність обраної теми аргументована постійним створенням нових комунікацій як у міжнародних відносинах, так і в економічній сфері. У статті виокремлено такі властивості офіційно-ділових документів, як: аргументативність, інтертекстуальність, діалогічність та логічність. Дослідження лексики офіційно-ділового стилю ґрунтується саме цих характеристиках. Для аналізу було обрано контракт як тип документа. У статті проаналізовано лексеми та лексичні звороти іншомовного походження, канцеляризми та власне терміни. Такий аналіз дає змогу простежити діахронний розвиток юридичної лексики через призму документів міжнародного права.

The paper proposes a new voting technology for random forest trees – the Positional Approach to the Voting Function Formation (PAVFF). In contrast to existing forms of organizing the voting of random forest trees, the paper proposes to change the subjects of voting and to use as such individual finite elements of the tree, with weights determined in accordance with the competences of new voting units. Each forest tree in the voting function is represented by its individual branch (voting unit) with the corresponding competence level assigned at the stage of tree finite element verification. Furthermore, we propose different mechanisms for organizing the received units in the voting process. The effectiveness of the new mechanism is shown by the example of the differentiation problem of drug-sensitive and drug-resistant forms of tuberculosis. The task feature space is formed by the ROI (regions of interest) textural characteristics on the patient’s lungs CT scan. The initial feature space was composed of the elements of a few textural characteristic matrices. From over half a million input features, a few optimal ensembles were selected to form random forest trees. We used intra- and inter-class variance selection techniques for this purpose, with the final selection made by a genetic algorithm using a combined correlation criterion. After verification of voting units (finite elements of trees) 3 variants of voting by competence were formed: by the most competent unit, weighted average participation of all participants, and group voting with coefficient revaluation by the Group Method of Data Handling. The results were compared with similar voting by random forest trees. A 5% improvement in classification quality is shown.