ADHD-KG: a knowledge graph of attention deficit hyperactivity disorder

Purpose

Attention Deficit Hyperactivity Disorder (ADHD) is a widespread condition that affects human behaviour and can interfere with daily activities and relationships. Medication or medical information about ADHD can be found in several data sources on the Web. Such distribution of knowledge raises notable obstacles since researchers and clinicians must manually combine various sources to deeply explore aspects of ADHD. Knowledge graphs have been widely used in medical applications due to their data integration capabilities, offering rich data stores of information built from heterogeneous sources; however, general purpose knowledge graphs cannot represent knowledge in sufficient detail, thus there is an increasing interest in domain-specific knowledge graphs.

Methods

In this work we propose a Knowledge Graph of ADHD. In particular, we introduce an automated procedure enabling the construction of a knowledge graph that covers knowledge from a wide range of data sources primarily focusing on adult ADHD. These include relevant literature and clinical trials, prescribed medication and their known side-effects. Data integration between these data sources is accomplished by employing a suite of information linking procedures, which aim to connect resources by relating them to common concepts found in medical thesauri.

Results

The usability and appropriateness of the developed knowledge graph is evaluated through a series of use cases that illustrate its ability to enhance and accelerate information retrieval.

Conclusion

The Knowledge Graph of ADHD can provide valuable assistance to researchers and clinicians in the research, training, diagnostic and treatment processes for ADHD.

Diagnosing attention-deficit hyperactivity disorder (ADHD) using artificial intelligence: a clinical study in the UK

Attention-deficit hyperactivity disorder (ADHD) is a neurodevelopmental disorder affecting a large percentage of the adult population. A series of ongoing efforts has led to the development of a hybrid AI algorithm (a combination of a machine learning model and a knowledge-based model) for assisting adult ADHD diagnosis, and its clinical trial currently operating in the largest National Health Service (NHS) for adults with ADHD in the UK. Most recently, more data was made available that has lead to a total collection of 501 anonymized records as of 2022 July. This prompted the ongoing research to carefully examine the model by retraining and optimizing the machine learning algorithm in order to update the model with better generalization capability. Based on the large data collection so far, this paper also pilots a study to examine the effectiveness of variables other than the Diagnostic Interview for ADHD in adults (DIVA) assessment, which adds considerable cost in the screenining process as it relies on specially trained senior clinicians. Results reported in this paper demonstrate that the newly trained machine learning model reaches an accuracy of 75.03% when all features are used; the hybrid model obtains an accuracy of 93.61%. Exceeding what clinical experts expected in the absence of DIVA, achieving an accuracy of 65.27% using a rule-based machine learning model alone encourages the development of a cost effective model in the future.

A dominant set-informed interpretable fuzzy system for automated diagnosis of dementia

Dementia is an incurable neurodegenerative disease primarily affecting the older population, for which the World Health Organisation has set to promoting early diagnosis and timely management as one of the primary goals for dementia care. While a range of popular machine learning algorithms and their variants have been applied for dementia diagnosis, fuzzy systems, which have been known effective in dealing with uncertainty and offer to explicitly reason how a diagnosis can be inferred, sporadically appear in recent literature. Given the advantages of a fuzzy rule-based model, which could potentially result in a clinical decision support system that offers understandable rules and a transparent inference process to support dementia diagnosis, this paper proposes a novel fuzzy inference system by adapting the concept of dominant sets that arise from the study of graph theory. A peeling-off strategy is used to iteratively extract from the constructed edge-weighted graph a collection of dominant sets. Each dominant set is further converted into a parameterized fuzzy rule, which is finally optimized in a supervised adaptive network-based fuzzy inference framework. An illustrative example is provided that demonstrates the interpretable rules and the transparent reasoning process of reaching a decision. Further systematic experiments conducted on data from the Open Access Series of Imaging Studies (OASIS) repository, also validate its superior performance over alternative methods.

Mental Health Diagnosis: A Case for Explainable Artificial Intelligence

Mental illnesses are becoming increasingly prevalent, in turn leading to an increased interest in exploring artificial intelligence (AI) solutions to facilitate and enhance healthcare processes ranging from diagnosis to monitoring and treatment. In contrast to application areas where black box systems may be acceptable, explainability in healthcare applications is essential, especially in the case of diagnosing complex and sensitive mental health issues. In this paper, we first summarize recent developments in AI research for mental health, followed by an overview of approaches to explainable AI and their potential benefits in healthcare settings. We then present a recent case study of applying explainable AI for ADHD diagnosis which is used as a basis to identify challenges in realizing explainable AI solutions for mental health diagnosis and potential future research directions to address these challenges.

A hybrid AI approach for supporting clinical diagnosis of attention deficit hyperactivity disorder (ADHD) in adults

Attention deficit hyperactivity disorder (ADHD) is a neurodevelopmental disorder that includes symptoms such as inattentiveness, hyperactivity and impulsiveness. It is considered as an important public health issue and prevalence of, as well as demand for diagnosis, has increased as awareness of the disease grew over the past years. Supply of specialist medical experts has not kept pace with the increasing demand for assessment, both due to financial pressures on health systems and the difficulty to train new experts, resulting in growing waiting lists. Patients are not being treated quickly enough causing problems in other areas of health systems (e.g. increased GP visits, increased risk of self-harm and accidents) and more broadly (e.g. time off work, relationship problems). Advances in AI make it possible to support the clinical diagnosis of ADHD based on the analysis of relevant data. This paper reports on findings related to the mental health services of a specialist Trust within the UK’s National Health Service (NHS). The analysis studied data of adult patients who underwent diagnosis over the past few years, and developed a hybrid approach, consisting of two different models: a machine learning model obtained by training on data of past cases; and a knowledge model capturing the expertise of medical experts through knowledge engineering. The resulting algorithm has an accuracy of 95% on data currently available, and is currently being tested in a clinical environment.

A Decision Tree-Initialised Neuro-fuzzy Approach for Clinical Decision Support

Apart from the need for superior accuracy, healthcare applications of intelligent systems also demand the deployment of interpretable machine learning models which allow clinicians to interrogate and validate extracted medical knowledge. Fuzzy rule-based models are generally considered interpretable that are able to reflect the associations between medical conditions and associated symptoms, through the use of linguistic if-then statements. Systems built on top of fuzzy sets are of particular appealing to medical applications since they enable the tolerance of vague and imprecise concepts that are often embedded in medical entities such as symptom description and test results. They facilitate an approximate reasoning framework which mimics human reasoning and supports the linguistic delivery of medical expertise often expressed in statements such as 'weight low' or 'glucose level high' while describing symptoms. This paper proposes an approach by performing data-driven learning of accurate and interpretable fuzzy rule bases for clinical decision support. The approach starts with the generation of a crisp rule base through a decision tree learning mechanism, capable of capturing simple rule structures. The crisp rule base is then transformed into a fuzzy rule base, which forms the input to the framework of adaptive network-based fuzzy inference system (ANFIS), thereby further optimising the parameters of both rule antecedents and consequents. Experimental studies on popular medical data benchmarks demonstrate that the proposed work is able to learn compact rule bases involving simple rule antecedents, with statistically better or comparable performance to those achieved by state-of-the-art fuzzy classifiers.

Toward Automatic Risk Assessment to Support Suicide Prevention

Background: Suicide has been considered an important public health issue for years and is one of the main causes of death worldwide. Despite prevention strategies being applied, the rate of suicide has not changed substantially over the past decades. Suicide risk has proven extremely difficult to assess for medical specialists, and traditional methodologies deployed have been ineffective. Advances in machine learning make it possible to attempt to predict suicide with the analysis of relevant data aiming to inform clinical practice. Aims: We aimed to (a) test our artificial intelligence based, referral-centric methodology in the context of the National Health Service (NHS), (b) determine whether statistically relevant results can be derived from data related to previous suicides, and (c) develop ideas for various exploitation strategies. Method: The analysis used data of patients who died by suicide in the period 2013-2016 including both structured data and free-text medical notes, necessitating the deployment of state-of-the-art machine learning and text mining methods. Limitations: Sample size is a limiting factor for this study, along with the absence of non-suicide cases. Specific analytical solutions were adopted for addressing both issues. Results and Conclusion: The results of this pilot study indicate that machine learning shows promise for predicting within a specified period which people are most at risk of taking their own life at the time of referral to a mental health service.

Evasion of No-Hair Theorems and Novel Black-Hole Solutions in Gauss-Bonnet Theories

We consider a general Einstein-scalar-Gauss-Bonnet theory with a coupling function f(ϕ). We demonstrate that black-hole solutions appear as a generic feature of this theory since a regular horizon and an asymptotically flat solution may be easily constructed under mild assumptions for f(ϕ). We show that the existing no-hair theorems are easily evaded, and a large number of regular black-hole solutions with scalar hair are then presented for a plethora of coupling functions f(ϕ).

Representing Time and Space for the Semantic Web

Representation of temporal and spatial information for the Semantic Web often involves qualitative defined information (i.e., information described using natural language terms such as “before” or “overlaps”) since precise dates or coordinates are not always available. This work proposes several temporal representations for time points and intervals and spatial topological representations in ontologies by means of OWL properties and reasoning rules in SWRL. All representations are fully compliant with existing Semantic Web standards and W3C recommendations. Although qualitative representations for temporal interval and point relations and spatial topological relations exist, this is the first work proposing representations combining qualitative and quantitative information for the Semantic Web. In addition to this, several existing and proposed approaches are compared using different reasoners and experimental results are presented in detail. The proposed approach is applied to topological relations (RCC5 and RCC8) supporting both qualitative and quantitative (i.e., using coordinates) spatial relations. Experimental results illustrate that reasoning performance differs greatly between different representations and reasoners. To the best of our knowledge, this is the first such experimental evaluation of both qualitative and quantitative Semantic Web temporal and spatial representations. In addition to the above, querying performance using SPARQL is evaluated. Evaluation results demonstrate that extracting qualitative relations from quantitative representations using reasoning rules and querying qualitative relations instead of directly querying quantitative representations increases performance at query time.

Selective fusion in adolescent idiopathic scoliosis: a radiographic evaluation of risk factors for imbalance

STUDY DESIGN

Retrospective database, chart and medical imaging review.

OBJECTIVES

To report on the outcome and evaluate possible risk factors for postoperative complications following selective spinal fusion in patients with adolescent idiopathic scoliosis (AIS).

MATERIALS AND METHODS

All patients with AIS who underwent either a selective thoracic or selective thoracolumbar/lumbar spinal fusion at our institution from January 2001 to December 2011 inclusive were included in this study. The minimum postoperative follow-up period of all patients was 2 years.

RESULTS

During the 11-year study period, 157 patients with AIS underwent surgery for their progressive spinal deformity. Thirty patients (19 %) had a selective spinal fusion, with 16 patients (group A) having a selective thoracic, and 14 patients (group B) having a selective thoracolumbar/lumbar spinal arthrodesis. In both groups the main postoperative complications were adding-on (25 % group A, 36 % group B) and coronal decompensation (25 % group A, 29 % group B). In group A, no statistically significant risk factors for postoperative complications were identified. In group B, global coronal balance was identified as a significant risk factor for adding-on. Patients with adding-on had significantly higher coronal balance scores (mean 3.6) than those who did not experience adding-on (mean 1.9) (p = 0.03). In addition, those with adding-on had a significantly smaller bending lumbar Cobb angle (mean 15) than those without adding-on (mean 31.6) (p = 0.015). None of the patients who underwent selective spinal fusion required revision surgery.

CONCLUSION

Although the complication rate after performing a selective spinal fusion is high, the revision rate remains low and the debate whether or not to perform a selective spinal fusion will continue.

Routine versus no drain placement after elective laparoscopic cholecystectomy: meta-analysis of randomized controlled trials

Routine drainage of the subhepatic space has been a surgical trend of open cholecystectomy, carried on to the era of laparoscopic surgery without substantial evidence. Avoiding the potentially devastating sequelae of an undetected bile leakage is the main rationale behind this practice. Aim of this meta-analysis was to compare evidence on routine drain placement after laparoscopic cholecystectomy versus no drainage. A meta-analysis of randomized controlled trials was conducted; outcome variables included postoperative pain, subhepatic collection, 30-day morbidity, wound-related complications, and drainage interventions. The fixed- and random effects models were used in order to calculate combined overall effect sizes of pooled data. Data are presented as the odds ratio (OR) or difference in means with 95% confidence interval (CI). Six randomized trials including 1167 patients were identified. Pain scores were significantly higher in the drainage group both at 6-12h (mean difference 1.12, 95% CI 1.01-1.24, P<0.0001) and at 12-24h after surgery (mean difference 1.12, 95% CI 0.86-1.39, P<0.0001). No difference was found with regard to the incidence of subhepatic collection and drainage procedures. A trend in favor of the no drain approach with regard to 30-day morbidity and wound infection was registered, although this was less pronounced after sensitivity analysis. The possible clinical benefit of routine use of abdominal drainage in uncomplicated laparoscopic cholecystectomies requires larger study populations. The approach is however not encouraged on the basis of the present analysis, as it results in increased postoperative pain and overall morbidity.

Hypnosis antenatal training for childbirth: a randomised controlled trial

OBJECTIVE

To determine the use of pharmacologic analgesia during childbirth when antenatal hypnosis is added to standard care.

DESIGN

Randomised controlled clinical trial, conducted from December 2005 to December 2010.

SETTING

The largest tertiary referral centre for maternity care in South Australia.

POPULATION

A cohort of 448 women at >34 weeks of gestation, with a singleton pregnancy and cephalic presentation, planning a vaginal birth. Exclusions were: the need for an interpreter; pre-existing pain; psychiatric illness; younger than 18 years; and previous experience of hypnosis for childbirth.

METHODS

All participants received usual care. The group of women termed Hypnosis + CD (hypnotherapist guided) were offered three antenatal live hypnosis sessions plus each session's corresponding audio CD for further practise, as well as a final fourth CD to listen to during labour. The group of women termed CD only (nurse administered) were played the same antenatal hypnosis CDs as group 1, but did not receive live hypnosis training. The control group participants were given no additional intervention or CDs.

MAIN OUTCOME MEASURE

Use of pharmacological analgesia during labour and childbirth.

RESULTS

No difference in the use of pharmacological analgesia during labour and childbirth was found comparing hypnosis + CD with control (81.2 versus 76.2%; relative risk, RR 1.07; 95% confidence interval, 95% CI 0.95-1.20), or comparing CD only with control (76.9 versus 76.2%, RR 1.01, 95% CI 0.89-1.15).

CONCLUSIONS

Antenatal group hypnosis using the Hypnosis Antenatal Training for Childbirth (HATCh) intervention in late pregnancy does not reduce the use of pharmacological analgesia during labour and childbirth.

Phlebosclerosis in lower extremities veins

Objectives: To determine the prevalence, distribution and extent of phlebosclerosis in an outpatient population referred to vascular clinic.

Design: Prospective cohort study.

Patients and Methods: Seven hundred and forty consecutive patients, 302 (41%) men with a cohort median age of 49 (range, 18–84), were referred to the vascular clinic for detection of lower extremity acute or chronic venous disease. A total of 1000 limbs were assessed using duplex ultrasonography. Deep veins were not assessed. Characteristics of venous wall and lumen were studied. Phlebosclerosis was diagnosed based on the increased echogenicity and thickness or calcification of the venous wall. Patients with known trauma or any surgery in the lower extremities were excluded. Two control groups were used for comparison, one of consecutive patients undergoing vein mapping ( n = 100) and the other of younger volunteers free of any vein disease ( n = 25).

Results: Of the 1000 limbs (right: 458, left: 542) studied by duplex ultrasound, 21 limbs had intense brightness of the venous wall. Fifteen of them had also calcifications and 61 wall thickening. The prevalence of phlebosclerosis was significantly higher in the patient group when compared with control groups 1 (P = 0.019) and 2 (P = 0.011). The mean age of patients with phlebosclerosis was higher compared with the whole patient group (57 versus 49 years, P < 0.0001). Phlebosclerosis affected all superficial veins with greater prevalence in the small saphenous vein (SSV). The mean length of the phlebosclerotic lesion was 2 cm (range 1–7 cm).

Conclusion: The prevalence of phlebosclerosis in the lower extremities appears to be low with no significant sex differences. Age and chronic venous disease are important factors for its development. It may be present in the absence of thrombosis or reflux. A greater prevalence of phlebosclerotic lesions was found in the SSV.

Slipped capital femoral epiphysis: rising rates with obesity and aboriginality in South Australia

We analysed the incidence of slipped capital femoral epiphysis (SCFE) in South Australia, investigating possible associations between an increased incidence of SCFE, the local indigenous population and the Australian obesity epidemic during the last 20 years. Data including race, age and gender were collected to obtain a profile of the South Australian SCFE patient, and were then compared with epidemiological data for South Australian adolescents. We concluded that the incidence of both obesity and SCFE is increasing. We also noted that the median weight of SCFE patients has increased and the mean age at diagnosis has decreased. Despite weight profiles comparable with those of the general population, we noted that an indigenous child was three times more likely to develop SCFE than a non-indigenous child. As far as we know there is no published literature on the predisposition of Aboriginal Australians to SCFE.

Endovascular stent graft repair of renal artery aneurysms

AIM

To evaluate the published outcomes of endovascular stent graft repair of renal artery aneurysms.

METHODS

The English literature was systematically searched using the MEDLINE electronic database up to April 2010. All reports of endovascular stent graft repair of renal artery aneurysms were considered. The primary outcome measures were technical success and perioperative, 30-day and overall morbidity and mortality.

RESULTS

Data for the final analysis were extracted from 20 papers reporting on endovascular stent graft repair of 22 aneurysms in 21 patients. Elective repair was performed in 20 cases (91%), whereas 2 patients (9%) underwent an endovascular repair of a ruptured aneurysm. The technical success rate was 91% (20/22). Neither perioperative/30-day mortality nor significant perioperative morbidity was noticed in this study cohort. Minor perioperative complications developed 6 patients (27%), and were most commonly associated with occlusion of a side branch vessel resulting in segmental infarction of the kidney. None of the patients developed renal impairment. During a mean follow up of 8 months, no patients developed aneurysm-related complications requiring revision.

CONCLUSION

Endovascular stent graft repair of renal artery aneurysms is reported to be a safe and effective treatment option, with good immediate and short-term results. However, no long-term data exists and larger series are required to draw solid conclusions regarding the outcomes of this method.

A Modal Defeasible Reasoner of Deontic Logic for the Semantic Web

Defeasible logic is a non-monotonic formalism that deals with incomplete and conflicting information, whereas modal logic deals with the concepts of necessity and possibility. These types of logics play a significant role in the emerging Semantic Web, which enriches the available Web information with meaning, leading to better cooperation between end-users and applications. Defeasible and modal logics, in general, and, particularly, deontic logic provide means for modeling agent communities, where each agent is characterized by its cognitive profile and normative system, as well as policies, which define privacy requirements, access permissions, and individual rights. Toward this direction, this article discusses the extension of DR-DEVICE, a Semantic Web-aware defeasible reasoner, with a mechanism for expressing modal logic operators, while testing the implementation via deontic logic operators, concerned with obligations, permissions, and related concepts. The motivation behind this work is to develop a practical defeasible reasoner for the Semantic Web that takes advantage of the expressive power offered by modal logics, accompanied by the flexibility to define diverse agent behaviours. A further incentive is to study the various motivational notions of deontic logic and discuss the cognitive state of agents, as well as the interactions among them.

Ethical and legal aspects on the use of images and photographs in medical teaching and publication

The aim of the study was to investigate the legal and ethical concerns raised from the use of photographs and images in medical publication. A search in the pertinent literature was performed. It is of paramount importance that the patient's autonomy, privacy and confidentiality is respected. In all cases in which photographs and images contain identifiable information patient's consent for any potential use of this material is mandatory. Patients should be aware that with the evolution of electronic publication, once an image is published, there is no efficient control of its future misuse. Physicians and hospitals have a duty to use with confidentiality any material kept in the patient's medical records. Efforts should be made to anonymised images and photographs used in teaching and publication so that such information does not raise ethical and legal concerns. The procedures for using photographs and images in medical publication and teaching should respect the ethical principles and contain only anonymous information to avoid legal consequences. Continuous scrutiny and reform is required in order to adapt to the changing social and scientific environment.

Extended RDF as a Semantic Foundation of Rule Markup Languages

Ontologies and automated reasoning are the building blocks of the Semantic Web initiative. Derivation rules can be included in an ontology to define derived concepts, based on base concepts. For example, rules allow to define the extension of a class or property, based on a complex relation between the extensions of the same or other classes and properties. On the other hand, the inclusion of negative information both in the form of negation-as-failure and explicit negative information is also needed to enable various forms of reasoning. In this paper, we extend RDF graphs with weak and strong negation, as well as derivation rules. The ERDF stable model semantics of the extended framework (Extended RDF) is defined, extending RDF(S) semantics. A distinctive feature of our theory, which is based on Partial Logic, is that both truth and falsity extensions of properties and classes are considered, allowing for truth value gaps. Our framework supports both closed-world and open-world reasoning through the explicit representation of the particular closed-world assumptions and the ERDF ontological categories of total properties and total classes.

Popliteal artery pseudoaneurysm after total knee replacement

We report the case of a popliteal pseudoaneurysm following total knee replacement. A 70-year-old woman underwent total left knee replacement because of severe osteoarthritis. Eight days later, she presented with oedema and pain in her left calf. She had palpable foot pulses on the left leg and the ankle-brachial index was 0.98. The patient was treated for deep vein thrombosis. Two days later her calf pain and oedema deteriorated and her distal pulses were no longer palpable, while she developed limb coldness and paraesthesia, and the ankle-brachial index dropped to 0.4. Sonography was urgently performed indicating a large popliteal artery aneurysm (5.8 × 6.9 × 7.2 cm), confirmed by angiography. The patient was managed with removal of a 3.5 cm long segment of the popliteal artery and reconstruction with synthetic graft (PTFE 6 mm). Her condition soon improved and the patient is capable of walking approximately 1 km per day at 18-month follow-up.

Sudden infant death syndrome (SIDS) in South Australia 1968-97. Part 2: the epidemiology of non-prone and non-covered SIDS infants

OBJECTIVE

To identify the risk factors for infants who die suddenly and unexpectedly, but whose deaths are not related to prone position, or having the head covered.

METHODOLOGY

A case-control study was designed in which the cases were infants who had died of sudden infant death syndrome (SIDS) in South Australia between January 1974 and December 1997, who were found not prone, not bed sharing and with the head not covered. The controls were two infants for each case, born in the same year and found in the prone position (again not bed sharing and with the head not covered).

RESULTS

Sudden unexpected death infancy is rare in non-prone infants with the head not covered. occurring on average twice a year in South Australia, where there are 18,000-21,000 births per year. In this group there was a higher percentage of infants with features associated with low socio-economic groups (teenage pregnancies and maternal smoking), sibling SIDS, suspicion of non-accidental injury and the presence of minor congenital anomalies, especially cardiac anomalies.

CONCLUSIONS

The majority of unexpected deaths in infancy can be prevented by not allowing infants to be unobserved in prone position, and by preventing them from getting their faces covered. For the few infants not found in these positions, a careful investigation should be made for malformations or non-accidental injury.