Artificial Intelligence in Medicine

Distant Supervision for Treatment Relation Extraction by Leveraging MeSH Subheadings Publication date: Available online 7 June 2019 Source: Artificial Intelligence in Medicine Author(s): Tung Tran, Ramakanth Kavuluru Abstract The growing body of knowledge in biomedicine is too vast for human consumption. Hence there is a need for automated systems able to navigate and distill the emerging wealth of information. One fundamental task to that end is relation extraction, whereby linguistic expressions of semantic relationships between biomedical entities are recognized and extracted. In this study, we propose a novel distant supervision approach for relation extraction of binary treatment relationships such that high quality positive/negative training examples are generated from PubMed abstracts by leveraging associated MeSH subheadings. The quality of generated examples is assessed based on the quality of supervised models they induce; that is, the mean performance of trained models (derived via bootstrapped ensembling) on a gold standard test set is used as a proxy for data quality. We show that our approach is preferable to traditional distant supervision for treatment relations and is closer to human crowd annotations in terms of annotation quality. For treatment relations, our generated training data performs at 81.38%, compared to traditional distant supervision at 64.33% and crowd-sourced annotations at 90.57% on the model-wide PR-AUC metric. We also demonstrate that examples generated using our method can be used to augment crowd-sourced datasets. Augmented models improve over non-augmented models by more than two absolute points on the more established F1 metric. We lastly demonstrate that performance can be further improved by implementing a classification loss that is resistant to label noise.

A Deep Survival Analysis Method Based on Ranking Publication date: Available online 6 June 2019 Source: Artificial Intelligence in Medicine Author(s): Bingzhong Jing, Tao Zhang, Zixian Wang, Ying Jin, Kuiyuan Liu, Wenze Qiu, Liangru Ke, Ying Sun, Caisheng He, Dan Hou, Linquan Tang, Xing Lv, Chaofeng Li Abstract Survival analyses of populations and the establishment of prognoses for individual patients are important activities in the practice of medicine. Standard survival models, such as the Cox proportional hazards model, require extensive feature engineering or prior knowledge to model at an individual level. Some survival analysis models can avoid these problems by using machine learning extended the CPH model, and higher performance has been reported. In this paper, we propose an innovative loss function that is defined as the sum of an extended mean squared error loss and a pairwise ranking loss based on ranking information on survival data. We apply this loss function to optimize a deep feed-forward neural network (RankDeepSurv), which can be used to model survival data. We demonstrate that the performance of our model, RankDeepSurv, is superior to that of other state-of-the-art survival models based on an analysis of 4 public medical clinical datasets. When modelling the prognosis of nasopharyngeal carcinoma (NPC), RankDeepSurv achieved better prognostic accuracy than the CPH established by clinical experts. The difference between high and low risk groups in the RankDeepSurv model is greater than the difference in the CPH. The results show that our method has considerable potential to model survival data in medical settings.

Combining clustering and classification ensembles: A novel pipeline to identify breast cancer profiles Publication date: June 2019 Source: Artificial Intelligence in Medicine, Volume 97 Author(s): Utkarsh Agrawal, Daniele Soria, Christian Wagner, Jonathan Garibaldi, Ian O. Ellis, John M.S. Bartlett, David Cameron, Emad A. Rakha, Andrew R. Green Abstract Breast Cancer is one of the most common causes of cancer death in women, representing a very complex disease with varied molecular alterations. To assist breast cancer prognosis, the classification of patients into biological groups is of great significance for treatment strategies. Recent studies have used an ensemble of multiple clustering algorithms to elucidate the most characteristic biological groups of breast cancer. However, the combination of various clustering methods resulted in a number of patients remaining unclustered. Therefore, a framework still needs to be developed which can assign as many unclustered (i.e. biologically diverse) patients to one of the identified groups in order to improve classification. Therefore, in this paper we develop a novel classification framework which introduces a new ensemble classification stage after the ensemble clustering stage to target the unclustered patients. Thus, a step-by-step pipeline is introduced which couples ensemble clustering with ensemble classification for the identification of core groups, data distribution in them and improvement in final classification results by targeting the unclustered data. The proposed pipeline is employed on a novel real world breast cancer dataset and subsequently its robustness and stability are examined by testing it on standard datasets. The results show that by using the presented framework, an improved classification is obtained. Finally, the results have been verified using statistical tests, visualisation techniques, cluster quality assessment and interpretation from clinical experts.

Recurrent neural networks with segment attention and entity description for relation extraction from clinical texts Publication date: June 2019 Source: Artificial Intelligence in Medicine, Volume 97 Author(s): Zhi Li, Jinshan Yang, Xu Gou, Xiaorong Qi Abstract At present, great progress has been achieved on the relation extraction for clinical texts, but we have noticed that the current models have great drawbacks when dealing with long sentences and multiple entities in a sentence. In this paper, we propose a novel neural network architecture based on Bidirectional Long Short-Term Memory Networks for relation classification. Firstly, we utilize a concat-attention mechanism for capturing the most important context words for relation extraction in a sentence. In addition, a segment attention mechanism is proposed to improve the performance of the model processing long sentences. Finally, a tensor-based entity description is used to overcome the performance degradation of the model when there are multiple entities in a sentence. The performance of the proposed model is evaluated on a part of the i2b2-2010 shared task clinical relation extraction dataset. The result indicates that our model can effectively overcome the above two problems and improve the F1-score by approximately 3% compared with baseline model.

Incorporated region detection and classification using deep convolutional networks for bone age assessment Publication date: June 2019 Source: Artificial Intelligence in Medicine, Volume 97 Author(s): Toan Duc Bui, Jae-Joon Lee, Jitae Shin Abstract Bone age assessment plays an important role in the endocrinology and genetic investigation of patients. In this paper, we proposed a deep learning-based approach for bone age assessment by integration of the Tanner-Whitehouse (TW3) methods and deep convolution networks based on extracted regions of interest (ROI)-detection and classification using Faster-RCNN and Inception-v4 networks, respectively. The proposed method allows exploration of expert knowledge from TW3 and features engineering from deep convolution networks to enhance the accuracy of bone age assessment. The experimental results showed a mean absolute error of about 0.59 years between expert radiologists and the proposed method, which is the best performance among state-of-the-art methods.

Texture descriptors and voxels for the early diagnosis of Alzheimer's disease Publication date: June 2019 Source: Artificial Intelligence in Medicine, Volume 97 Author(s): Loris Nanni, Sheryl Brahnam, Christian Salvatore, Isabella Castiglioni, the Alzheimer's Disease Neuroimaging Initiative Abstract Background and objective Early and accurate diagnosis of Alzheimer's Disease (AD) is critical since early treatment effectively slows the progression of the disease thereby adding productive years to those afflicted by this disease. A major problem encountered in the classification of MRI for the automatic diagnosis of AD is the so-called curse-of-dimensionality, which is a consequence of the high dimensionality of MRI feature vectors and the low number of training patterns available in most MRI datasets relevant to AD. Methods A method for performing early diagnosis of AD is proposed that combines a set of SVMs trained on different texture descriptors (which reduce dimensionality) extracted from slices of Magnetic Resonance Image (MRI) with a set of SVMs trained on markers built from the voxels of MRIs. The dimension of the voxel-based features is reduced by using different feature selection algorithms, each of which trains a separate SVM. These two sets of SVMs are then combined by weighted-sum rule for a final decision. Results Experimental results show that 2D texture descriptors improve the performance of state-of-the-art voxel-based methods. The evaluation of our system on the four ADNI datasets demonstrates the efficacy of the proposed ensemble and demonstrates a contribution to the accurate prediction of AD. Conclusions Ensembles of texture descriptors combine partially uncorrelated information with respect to standard approaches based on voxels, feature selection, and classification by SVM. In other words, the fusion of a system based on voxels and an ensemble of texture descriptors enhances the performance of voxel-based approaches.

Supporting the Distributed Execution of Clinical Guidelines by Multiple Agents Publication date: Available online 18 May 2019 Source: Artificial Intelligence in Medicine Author(s): Alessio Bottrighi, Luca Piovesan, Paolo Terenziani Abstract Clinical guidelines (GLs) are widely adopted in order to improve the quality of patient care, and to optimize it. To achieve such goals, their application on a specific patient usually requires the interventions of different agents, with different roles (e.g., physician, nurse), abilities (e.g., specialist in the treatment of alcohol-related problems) and contexts (e.g., many chronic patients may be treated at home). Additionally, the responsibility of the application of a guideline to a patient is usually retained by a physician, but delegation of responsibility (of the whole guideline, or of a part of it) is often used\required (e.g., delegation to a specialist), as well as the possibility, for a responsible, to select the executor of an action (e.g., a physician may retain the responsibility of an action, but delegate to a nurse its execution). To manage such phenomena, proper support to agent interaction and communication must be provided, providing agents with facilities for (1) treatment continuity (2) contextualization, (3) responsibility assignment and delegation (4) check of agent "appropriateness". In this paper we extend GLARE, a computerized GL management system, to support such needs. We illustrate our approach by means of a practical case study.

Personalized conciliation of clinical guidelines for comorbid patients through multi-agent planning Publication date: May 2019 Source: Artificial Intelligence in Medicine, Volume 96 Author(s): Juan Fdez-Olivares, Eva Onaindia, Luis Castillo, Jaume Jordán, Juan Cózar Abstract The conciliation of multiple single-disease guidelines for comorbid patients entails solving potential clinical interactions, discovering synergies in the diagnosis and the recommendations, and managing clinical equipoise situations. Personalized conciliation of multiple guidelines considering additionally patient preferences brings some further difficulties. Recently, several works have explored distinct techniques to come up with an automated process for the conciliation of clinical guidelines for comorbid patients but very little attention has been put in integrating the patient preferences into this process. In this work, a Multi-Agent Planning (MAP) framework that extends previous work on single-disease temporal Hierarchical Task Networks (HTN) is proposed for the automated conciliation of clinical guidelines with patient-centered preferences. Each agent encapsulates a single-disease Computer Interpretable Guideline (CIG) formalized as an HTN domain and conciliates the decision procedures that encode the clinical recommendations of its CIG with the decision procedures of the other agents' CIGs. During conciliation, drug-related interactions, scheduling constraints as well as redundant actions and multiple support interactions are solved by an automated planning process. Moreover, the simultaneous application of the patient preferences in multiple diseases may potentially bring about contradictory clinical decisions and more interactions. As a final step, the most adequate personalized treatment plan according to the patient preferences is selected by a Multi-Criteria Decision Making (MCDM) process. The MAP approach is tested on a case study that builds upon a simplified representation of two real clinical guidelines for Diabetes Mellitus and Arterial Hypertension.

Indexing the Event Calculus: Towards practical human-readable Personal Health Systems Publication date: May 2019 Source: Artificial Intelligence in Medicine, Volume 96 Author(s): Nicola Falcionelli, Paolo Sernani, Albert Brugués, Dagmawi Neway Mekuria, Davide Calvaresi, Michael Schumacher, Aldo Franco Dragoni, Stefano Bromuri Abstract Personal Health Systems (PHS) are mobile solutions tailored to monitoring patients affected by chronic non communicable diseases. In general, a patient affected by a chronic disease can generate large amounts of events: for example, in Type 1 Diabetic patients generate several glucose events per day, ranging from at least 6 events per day (under normal monitoring) to 288 per day when wearing a continuous glucose monitor (CGM) that samples the blood every 5 minutes for several days. Just by itself, without considering other physiological parameters, it would be impossible for medical doctors to individually and accurately follow every patient, highlighting the need of simple approaches towards querying physiological time series. Achieving this with current technology is not an easy task, as on one hand it cannot be expected that medical doctors have the technical knowledge to query databases and on the other hand these time series include thousands of events, which requires to re-think the way data is indexed. Anyhow, handling data streams efficiently is not enough. Domain experts' knowledge must be explicitly included into PHSs in a way that it can be easily readed and modified by medical staffs. Logic programming represents the perfect programming paradygm to accomplish this task. In this work, an Event Calculus-based reasoning framework to standardize and express domain-knowledge in the form of monitoring rules is suggested, and applied to three different use cases. However, if online monitoring has to be achieved, the reasoning performance must improve dramatically. For this reason, three promising mechanisms to index the Event Calculus Knowledge Base are proposed. All of them are based on different types of tree indexing structures: k-d trees, interval trees and red-black trees. The paper then compares and analyzes the performance of the three indexing techniques, by computing the time needed to check different type of rules (and eventually generating alerts), when the number of recorded events (e.g. values of physiological parameters) increases. The results show that customized jREC performs much better when the event average inter-arrival time is little compared to the checked rule time-window. Instead, where the events are more sparse, the use of k-d trees with standard EC is advisable. Finally, the Multi-Agent paradigm helps to wrap the various components of the system: the reasoning engines represent the agent minds, and the sensors are its body. The said agents have been developed in MAGPIE, a mobile event based Java agent platform.

Towards a modular decision support system for radiomics: A case study on rectal cancer Publication date: May 2019 Source: Artificial Intelligence in Medicine, Volume 96 Author(s): Roberto Gatta, Mauro Vallati, Nicola Dinapoli, Carlotta Masciocchi, Jacopo Lenkowicz, Davide Cusumano, Calogero Casá, Alessandra Farchione, Andrea Damiani, Johan van Soest, Andre Dekker, Vincenzo Valentini Abstract Following the personalized medicine paradigm, there is a growing interest in medical agents capable of predicting the effect of therapies on patients, by exploiting the amount of data that is now available for each patient. In disciplines like oncology, where images and scans are available, the exploitation of medical images can provide an additional source of potentially useful information. The study and analysis of features extracted by medical images, exploited for predictive purposes, is termed radiomics. A number of tools are available for supporting some of the steps of the radiomics process, but there is a lack of approaches which are able to deal with all the steps of the process. In this paper, we introduce a medical agent-based decision support system capable of handling the whole radiomics process. The proposed system is tested on two independent data sets of patients treated for rectal cancer. Experimental results indicate that the system is able to generate highly performant centre-specific predictive model, and show the issues related to differences in data sets collected by different centres, and how such issues can affect the performance of the generated predictive models.

Alexandros Sfakianakis
Anapafseos 5 . Agios Nikolaos
Crete.Greece.72100
2841026182

6948891480

alsfakia@gmail.com