Artificial Intelligence Decision Support Systems Research Articles

Constructing and implementing a performance evaluation indicator set for artificial intelligence decision support systems in pediatric outpatient clinics: an observational study

Artificial intelligence (AI) decision support systems in pediatric healthcare have a complex application background. As an AI decision support system (AI-DSS) can be costly, once applied, it is crucial to focus on its performance, interpret its success, and then monitor and update it to ensure ongoing success consistently. Therefore, a set of evaluation indicators was explicitly developed for AI-DSS in pediatric healthcare, enabling continuous and systematic performance monitoring. The study unfolded in two stages. The first stage encompassed establishing the evaluation indicator set through a literature review, a focus group interview, and expert consultation using the Delphi method. In the second stage, weight analysis was conducted. Subjective weights were calculated based on expert opinions through analytic hierarchy process, while objective weights were determined using the entropy weight method. Subsequently, subject and object weights were synthesized to form the combined weight. In the two rounds of expert consultation, the authority coefficients were 0.834 and 0.846, Kendall's coordination coefficient was 0.135 in Round 1 and 0.312 in Round 2. The final evaluation indicator set has three first-class indicators, fifteen second-class indicators, and forty-seven third-class indicators. Indicator I-1(Organizational performance) carries the highest weight, followed by Indicator I-2(Societal performance) and Indicator I-3(User experience performance) in the objective and combined weights. Conversely, 'Societal performance' holds the most weight among the subjective weights, followed by 'Organizational performance' and 'User experience performance'. In this study, a comprehensive and specialized set of evaluation indicators for the AI-DSS in the pediatric outpatient clinic was established, and then implemented. Continuous evaluation still requires long-term data collection to optimize the weight proportions of the established indicators.

Cross-Check QA: A Quality Assurance Workflow to Prevent Missed Diagnoses by Alerting Inadvertent Discordance Between the Radiologist and Artificial Intelligence in the Interpretation of High-Acuity CT Scans

PurposeThe aim of this study was to implement and evaluate a quality assurance (QA) workflow that leverages natural language processing to rapidly resolve inadvertent discordance between radiologists and an artificial intelligence (AI) decision support system (DSS) in the interpretation of high-acuity CT studies when the radiologist does not engage with AI DSS output. MethodsAll consecutive high-acuity adult CT examinations performed in a health system between March 1, 2020, and September 20, 2022, were interpreted alongside an AI DSS (Aidoc) for intracranial hemorrhage, cervical spine fracture, and pulmonary embolus. CT studies were flagged for this QA workflow if they met three criteria: (1) negative results by radiologist report, (2) a high probability of positive results by the AI DSS, and (3) unviewed AI DSS output. In these cases, an automated e-mail notification was sent to our quality team. If discordance was confirmed on secondary review—an initially missed diagnosis—addendum and communication documentation was performed. ResultsOf 111,674 high-acuity CT examinations interpreted alongside the AI DSS over this 2.5-year time period, the frequency of missed diagnoses (intracranial hemorrhage, pulmonary embolus, and cervical spine fracture) uncovered by this workflow was 0.02% (n = 26). Of 12,412 CT studies prioritized as depicting positive findings by the AI DSS, 0.4% (n = 46) were discordant, unengaged, and flagged for QA. Among these discordant cases, 57% (26 of 46) were determined to be true positives. Addendum and communication documentation was performed within 24 hours of the initial report signing in 85% of these cases. ConclusionsInadvertent discordance between radiologists and the AI DSS occurred in a small number of cases. This QA workflow leveraged natural language processing to rapidly detect, notify, and resolve these discrepancies and prevent potential missed diagnoses.

Investigating feasibility of 2021 WHO protocol for cervical cancer screening in underscreened populations: PREvention and SCReening Innovation Project Toward Elimination of Cervical Cancer (PRESCRIP-TEC)

BackgroundHigh-risk human papillomavirus (hrHPV) testing has been recommended by the World Health Organization as the primary screening test in cervical screening programs. The option of self-sampling for this screening method can potentially increase women’s participation. Designing screening programs to implement this method among underscreened populations will require contextualized evidence.MethodsPREvention and SCReening Innovation Project Toward Elimination of Cervical Cancer (PRESCRIP-TEC) will use a multi-method approach to investigate the feasibility of implementing a cervical cancer screening strategy with hrHPV self-testing as the primary screening test in Bangladesh, India, Slovak Republic and Uganda. The primary outcomes of study include uptake and coverage of the screening program and adherence to follow-up. These outcomes will be evaluated through a pre-post quasi-experimental study design. Secondary objectives of the study include the analysis of client-related factors and health system factors related to cervical cancer screening, a validation study of an artificial intelligence decision support system and an economic evaluation of the screening strategy.DiscussionPRESCRIP-TEC aims to provide evidence regarding hrHPV self-testing and the World Health Organization’s recommendations for cervical cancer screening in a variety of settings, targeting vulnerable groups. The main quantitative findings of the project related to the impact on uptake and coverage of screening will be complemented by qualitative analyses of various determinants of successful implementation of screening. The study will also provide decision-makers with insights into economic aspects of implementing hrHPV self-testing, as well as evaluate the feasibility of using artificial intelligence for task-shifting in visual inspection with acetic acid.Trial registrationClinicalTrials.gov, NCT05234112. Registered 10 February 2022

Protocol for a qualitative study to explore acceptability, barriers and facilitators of the implementation of new teleophthalmology technologies between community optometry practices and hospital eye services

IntroductionNovel teleophthalmology technologies have the potential to reduce unnecessary and inaccurate referrals between community optometry practices and hospital eye services and as a result improve patients’ access to appropriate and...

Innovative Development of Intangible Culture of Arts and Crafts in Artificial Intelligence Decision Support System

The Chinese nation has accumulated a lot of precious, rich, and wonderful material and intangible culture in its historical evolution, but these cultures are facing the problem of inheritance difficulties in the long term, and it is becoming more and more difficult to innovate. In order to solve these problems, this paper puts forward the innovative development of the intangible culture of arts and crafts based on artificial intelligence decision support system, with the purpose of studying cultural inheritance, regional brand hematopoiesis promotion, and organizational goal’s support for intangible culture. The method of this paper is to study the basic principles of artificial bee colony algorithms and decision support systems, and then evaluate the value of intangible culture. The purpose of this paper is to analyze the development of local intangible culture and innovation models based on local knowledge systems and confirm the influence of technological innovation on intangible culture so as to promote the sustainable development of intangible culture. This paper describes the present situation of intangible cultural innovation and then explores the innovative ways of intangible culture based on the experiment of cultural development of artificial intelligence. The experimental results show that after using the decision support method mentioned in this paper, the technical progress index of Jingdezhen porcelain culture has increased by 2.4%.

Teleophthalmology-enabled and artificial intelligence-ready referral pathway for community optometry referrals of retinal disease (HERMES): a Cluster Randomised Superiority Trial with a linked Diagnostic Accuracy Study—HERMES study report 1—study protocol

IntroductionRecent years have witnessed an upsurge of demand in eye care services in the UK. With a large proportion of patients referred to Hospital Eye Services (HES) for diagnostics and...

Learning Aided System for Agriculture Monitoring Designed Using Image Processing and IoT-CNN

The Internet of Things (IoT) and artificial intelligence (AI) based methods for monitoring, control, and decision support are combined to design of a smart agriculture assistance system. The proposed system has a sensor pack that provides continuous data capture of temperature records, air and soil moisture and a camera for obtaining near-infrared (NIR) images of the plant leaves for use with an AI decision support system. We identify twelve types of vegetation for the study, out of which five disease classes of the tomato leaves are categorized using a Convolutional Neural Network (CNN). The work also includes experiments conducted with multiple clustering-based segmentation methods and some features namely Gray level co-occurrence matrix (GLCM), Local binary pattern (LBP), Local Binary Gray Level Co-occurrence Matrix (LBGLCM), Gray Level Run Length Matrix (GLRLM), and Segmentation-based Fractal Texture Analysis (SFTA). Out of several AI tools, CNN proves to be effective in providing automated decision support for classifying the plant leaf disease types through a cloud server that can be accessed using an app. Extensive on-field trials show that the system (VGG16 CNN, GLCM and a fuzzy based clustering) is effective in hot and humid conditions and proves to be reliable in identifying disease classes of certain vegetable types, certain usable vegetation cover of farmland and regulation of watering mechanism of crops.

Application of Artificial Intelligence and Collaborative Knowledge for Manufacturing Design

With the rapid development of science and technology, digital technology has brought the world economy and management into a new stage. Collaborative design can realize product design process by cross-regional and cross-industry designers and share and exchange product information through network. With the rapid development of big data and artificial intelligence, knowledge services have gradually developed into multirole collaborative design activities based on artificial intelligence decision support. Traditional manufacturing industry has gradually transformed into modern manufacturing service industry after integrating information technology means such as Internet, communication, computer, and modern management methods. This article focuses on artificial intelligence decision support systems and the complex product manufacturing industry. We present a detailed analysis of how to integrate the knowledge generated by the product life cycle in the era of big data. We calculate the influence coefficient and sensitivity index of four different industries and propose a metadata architecture to improve the value of products as a whole. The findings of the research study imply that a knowledge-based collaborative platform should be designed by the enterprises and industries and a platform-based construction approach for economical, practical, and reliable production. We also present a detailed discussion about other factors such as the network cost of symmetric services, raw data and forecast data, and the number of nodes and the processing complexity.

Artificial Intelligence Decision Support System Based on Artificial Neural Networks to Predict the Commercialization Time by the Evolution of Peach Quality

Climacteric fruit such as peaches are stored in cold chambers after harvest and usually are maintained there until the desired ripening is reached to direct these fruit to market. Producers, food industries and or traders have difficulties in defining the period when fruit are at the highest level of quality desired by consumers in terms of the physical-chemical parameters (hardness –H–, soluble solids content –SSC–, and acidity –Ac–). The evolution of peach quality in terms of these parameters depends directly on storage temperature –T– and relative humidity –RH–, as well on the storage duration –t–. This paper describes an Artificial Intelligence (AI) Decision Support System (DSS) designed to predict the evolution of the quality of peaches, namely the storage time required before commercialization as well as the late commercialization time. The peaches quality is stated in terms of the values of SSC, H and Ac that consumers most like for the storage T and RH. An Artificial neuronal network (ANN) is proposed to provide this prediction. The training and validation of the ANN were conducted with experimental data acquired in three different farmers’ cold storage facilities. A user interface was developed to provide an expedited and simple prediction of the marketable time of peaches, considering the storage temperature, relative humidity, and initial physical and chemical parameters. This AI DSS may help the vegetable sector (logistics and retailers), especially smaller neighborhood grocery stores, define the marketable period of fruit. It will contribute with advantages and benefits for all parties—producers, traders, retailers, and consumers—by being able to provide fruit at the highest quality and reducing waste in the process. In this sense, the ANN DSS proposed in this study contributes to new AI-based solutions for smart cities.

Beyond the relationship between the healthcare provider and the patient with an emerging next generation of electronic health records

IntroductionAn emerging next generation of electronic health records (EHRs) is expected to see greater realisation of benefits through empowering healthcare stakeholders including patients, providers, and researchers to do things differently. EHR technology is evolving because of the widespread availability of key enabling technologies, such as the Internet of Things (IoT), artificial intelligence (AI), genomic sequencing, and cloud storage. With the increasing adoption of EHRs and their ever-evolving functionality and utility, comes a wide spectrum of associated ethical and social concerns. Emerging technology, such as ePortals, AI decision support systems, voice recognition software and wearable devices, are expected to increase the prominence of EHR technology within the health domain.Policy context and objectiveThis research is timely because EHR technology is still at an early stage of adoption in many countries. For example, large improvements are projected in the next 10 years in the Irish health IT infrastructure. Several Irish government policies aim to transform healthcare through eHealth, e.g. Sláintecare eHealth pillar extends to 2030, the National Development Plan 2018 – 2027, the Department of Health Statement of Strategy 2016 – 2019, and Project Ireland 2040.We have conducted a scoping literature review, to explore which EHR related ethical concerns were being debated in the literature and why these ethical concerns were debated. This review will be useful to a range of stakeholders involved in a safe, legal and ethical development, implementation and use of EHRs including healthcare providers, policymakers, and IT developHighlights The literature identified a large range of ethical values associated with EHR technology. There was a strong consensus in the academic literature on the ethical values associated with EHR technology. Privacy has been identified as the most frequently mentioned value in the academic literature. The literature on the ethical and social effects of EHRs focuses predominantly on the relation between the healthcare professional and the patient.Comments on transferabilityThis abstract describes the academic discussion on the ethics of EHRs in the English language. The findings are therefore not limited to one jurisdiction.ConclusionsBecause of the growing prominence of EHR technology, the focus on the relationship between care provider and patient needs to be expanded. More emphasis is needed on relationships between the technology and healthcare stakeholders including patients, providers, and technologists.The growing awareness of the value of personal data is being experienced across all sectors incorporating digital technology. Recent scandals around technology in other sectors, for example around disinformation and news on social networks, can inform ethical inquiries into EHR technology. These inquiries should elucidate issues around the importance of relationships between stakeholders, transparency and accountability or lack thereof. The role of health technology assessment (HTA) to inform stakeholders about the social aspects, and ethical and legal issues is becoming more urgent. "

Diagnostic Accuracy and Failure Mode Analysis of a Deep Learning Algorithm for the Detection of Cervical Spine Fractures.

Artificial intelligence decision support systems are a rapidly growing class of tools to help manage ever-increasing imaging volumes. The aim of this study was to evaluate the performance of an artificial intelligence decision support system, Aidoc, for the detection of cervical spinal fractures on noncontrast cervical spine CT scans and to conduct a failure mode analysis to identify areas of poor performance. This retrospective study included 1904 emergent noncontrast cervical spine CT scans of adult patients (60 [SD, 22] years, 50.3% men). The presence of cervical spinal fracture was determined by Aidoc and an attending neuroradiologist; discrepancies were independently adjudicated. Algorithm performance was assessed by calculation of the diagnostic accuracy, and a failure mode analysis was performed. Aidoc and the neuroradiologist's interpretation were concordant in 91.5% of cases. Aidoc correctly identified 67 of 122 fractures (54.9%) with 106 false-positive flagged studies. Diagnostic performance was calculated as the following: sensitivity, 54.9% (95% CI, 45.7%-63.9%); specificity, 94.1% (95% CI, 92.9%-95.1%); positive predictive value, 38.7% (95% CI, 33.1%-44.7%); and negative predictive value, 96.8% (95% CI, 96.2%-97.4%). Worsened performance was observed in the detection of chronic fractures; differences in diagnostic performance were not altered by study indication or patient characteristics. We observed poor diagnostic accuracy of an artificial intelligence decision support system for the detection of cervical spine fractures. Many similar algorithms have also received little or no external validation, and this study raises concerns about their generalizability, utility, and rapid pace of deployment. Further rigorous evaluations are needed to understand the weaknesses of these tools before widespread implementation.

Diagnostic Accuracy and Failure Mode Analysis of a Deep Learning Algorithm for the Detection of Intracranial Hemorrhage.

To determine the institutional diagnostic accuracy of an artificial intelligence (AI) decision support systems (DSS), Aidoc, in diagnosing intracranial hemorrhage (ICH) on noncontrast head CTs and to assess the potential generalizability of an AI DSS. This retrospective study included 3,605 consecutive, emergent, adult noncontrast head CT scans performed between July 1, 2019, and December 30, 2019, at our institution (51% female subjects, mean age of 61 ± 21 years). Each scan was evaluated for ICH by both a certificate of added qualification certified neuroradiologist and Aidoc. We determined the diagnostic accuracy of the AI model and performed a failure mode analysis with quantitative CT radiomic image characterization. Of the 3,605 scans, 349 cases of ICH (9.7% of studies) were identified. The neuroradiologist and Aidoc interpretations were concordant in 96.9% of cases and the overall sensitivity, specificity, positive predictive value, and negative predictive value were 92.3%, 97.7%, 81.3%, and 99.2%, respectively, with positive predictive values unexpectedly lower than in previously reported studies. Prior neurosurgery, type of ICH, and number of ICHs were significantly associated with decreased model performance. Quantitative image characterization with CT radiomics failed to reveal significant differences between concordant and discordant studies. This study revealed decreased diagnostic accuracy of an AI DSS at our institution. Despite extensive evaluation, we were unable to identify the source of this discrepancy, raising concerns about the generalizability of these tools with indeterminate failure modes. These results further highlight the need for standardized study design to allow for rigorous and reproducible site-to-site comparison of emerging deep learning technologies.

A deep learning model (ALNet) for the diagnosis of acute leukaemia lineage using peripheral blood cell images

Background and objectivesMorphological differentiation among blasts circulating in blood in acute leukaemia is challenging. Artificial intelligence decision support systems hold substantial promise as part of clinical practise in detecting haematological malignancy. This study aims to develop a deep learning-based system to predict the diagnosis of acute leukaemia using blood cell images. MethodsA set of 731 blood smears containing 16,450 single-cell images was analysed from 100 healthy controls, 191 patients with viral infections and 148 with acute leukaemia. Training and testing sets were arranged with 85% and 15% of these smears, respectively. To find the best architecture for acute leukaemia classification VGG16, ResNet101, DenseNet121 and SENet154 were evaluated. Fine-tuning was implemented to these pre-trained CNNs to adapt their layers to our data. Once the best architecture was chosen, a system with two modules working sequentially was configured (ALNet). The first module recognised abnormal promyelocytes among other mononuclear blood cell images, such as lymphocytes, monocytes, reactive lymphocytes and blasts. The second distinguished if blasts were myeloid or lymphoid lineage. The final strategy was to predict patients’ initial diagnosis of acute leukaemia lineage using the blood smear review. ALNet was assessed with smears of the testing set. ResultsALNet provided the correct diagnostic prediction of all patients with promyelocytic and myeloid leukaemia. Sensitivity, specificity and precision values of 100%, 92.3% and 93.7%, respectively, were obtained for myeloid leukaemia. Regarding lymphoid leukaemia, a sensitivity of 89% and specificity and precision values of 100% were obtained. ConclusionsALNet is a predictive model designed with two serially connected convolutional networks. It is proposed to assist clinical pathologists in the diagnosis of acute leukaemia during the blood smear review. It has been proved to distinguish neoplastic (leukaemia) and non-neoplastic (infections) diseases, as well as recognise the leukaemia lineage.

Prognostic value of a normal Artificial Intelligence applicative response in subjects with stable chest pain. From the ARTICA co-operative registry

Abstract Background An innovative artificial intelligence (AI) Decision Support System (DSS) ESC guidelines based has already been used at point of care with efficacy for evaluating subjects with stable chest pain (SCP) and it has been proved to correctly identify absence of significant coronary artery disease (CAD) versus standard care approach (SD) without known prognostic implication yet. Purpose The aim is to determine the prognostic value of “no testing/Follow up” AI DSS response in a referral population evaluated for SCP. Methods From 2016 to 2019, an AI DSS ESC guidelines based applicative was used on 1.291 subjects with SCP to determine testing appropriateness compared with human specialist standard evaluation. 590 of them, 332 males, mean age 62±11 years deemed to be completely negative – by “no testing/follow up” response - were evaluated. The negative response was defined and confirmed by a normal Coronary Tomography Angiography scan executed in all these subjects. Mean follow-up was 3.46±1.76 years. Two groups based on pre-test likelihood of having CAD were analyzed – low and intermediate. No subjects with high pre-test likelihood were present. The primary endpoint was cumulative incidence of cardiovascular death, hospitalizations for acute coronary syndrome and coronary revascularizations. Results The primary endpoints classification is displayed in the Table. The unadjusted hazard ratio for primary endpoint was 3.84 (95% CI 0.32–8.68, p=0.009) in patients with intermediate compared to those with low pre-test likelihood of CAD. Moreover, the “no testing and Follow up” response showed an incremental prognostic value over conventional risk factors (χ2=7, P=0.022) and over a combination of conventional factors and ST-T changes (χ2=9, P=0.014). Conclusions In an outpatient population without known CAD evaluated for chest pain, after the administration of AI DSS, a “No tests/Follow up” confers an excellent prognosis regardless of the ESC SCORE Charts and rest ECG abnormalities. These preliminary data confirms the safety of ARTICA AI DSS use in subjects with stable chest pain. Funding Acknowledgement Type of funding source: None

An artificial intelligence decision support system for the management of type 1 diabetes.

Type 1 diabetes (T1D) is characterized by pancreatic beta cell dysfunction and insulin depletion. Over 40% of people with T1D manage their glucose through multiple injections of long-acting basal and short-acting bolus insulin, so called multiple daily injections (MDI).1,2 Errors in dosing can lead to life-threatening hypoglycaemic events (< 70 mg dL^−1) and hyperglycaemia (> 180 mg dL^−1), increasing the risk of retinopathy, neuropathy, and nephropathy. Machine learning (artificial intelligence) approaches are being harnessed to incorporate decision support into many medical specialties. Here we report an algorithm that provides weekly insulin dosage recommendations to adults with T1D using MDI therapy. We employ a unique virtual platform3 to generate over 50,000 glucose observations to train a K-nearest-neighbours4 decision support system (KNN-DSS) to identify causes of hyperglycaemia or hypoglycaemia and determine necessary insulin adjustments from a set of 12 potential recommendations. The KNN-DSS algorithm achieves an overall agreement with board-certified endocrinologists of 67.9% when validated on real-world human data, and delivers safe recommendations per endocrinologist review. A comparison of physician-recommended adjustments to insulin pump therapy indicates full agreement of 41.2% among endocrinologists, which is consistent with previous measures of inter-physician agreement (41–45%)5. In silico3,6 benchmarking using a platform accepted by the U.S. Food and Drug Administration for evaluation of artificial pancreas technologies, indicates substantial improvement in glycaemic outcomes after 12-weeks of KNN-DSS use. Our data indicate that the KNN-DSS allows for early identification of dangerous insulin regimens and may be used to improve glycaemic outcomes and prevent life-threatening complications in people with T1D.

P5245Cost-saving diagnosis approach by artificial intelligence tool in patients with suspected coronary artery disease. The co-operative ARTICA registry database

Abstract Background Although coronary tomographic angiography (CTA) has shown promise as a “gatekeeper” to invasive coronary angiography (ICA) in longitudinal cohort studies, it remains unknown whether the strategy of direct initial performance of CTA is cost-effective when compared with selected exercise treadmill testing (ETT) +/− functional cardiac imaging strategies in patients with suspected coronary artery disease (CAD). An innovative artificial intelligence (AI) Decision Support System (DSS) ESC guidelines based has been used at point of care for evaluating subjects with stable chest pain (SCP). Purpose The objective was to verify the cost-saving effect of the robotic AI DSS vs direct CTA by human standard care (SD) for diagnosing CAD in subjects presenting with SCP. Methods From October 2016 over three hospitals, 1017 subjects, 620 males, age 62±11 years, with clinically SCP being referred for CTA by SD received also a same day pre-scan AI DSS administration. All patients did not demonstrate significant CAD at CTA. CTA/ICA, or exercise treadmill test (ETT)/ stress echocardiography (SE), gated myocardial perfusion scintigraphy (gMPS) or Follow up/No tests (FNT) strategies by AI DSS were analyzed and compared to direct CTA SD. Pre-test likelihood (pt-lk) of CAD consider clinical risk factors into the model. Sensitivity and specificity of non-invasive diagnostic tests within our model were based upon a bivariate analysis of data from published multicenter trials. Costs of procedures were calculated by the sum of technical and professional components. Probabilistic sensitivity analysis was conducted to assess the impact of uncertainty in model parameters. Results The direct approach used performing direct CTA strategy by SD in all subjects costed 406.800 €. Costs of each procedure and distribution of AI DSS outputs are shown in the Table. Across the range of pt-lk of CAD, total costs of AI DSS strategy resulted 146.030€ with −65% vs SD approach. AI DSS tests distribution and costs pt-lk (pt/%) FNT (0€) ETT (90€) SE (350€) Stress gated MPS (750€) CCTA (400€) ICA (3.000€) High (29/2.8) 0 0 1 2 0 26 Int (371/36.5) 259 5 51 48 7 1 Low (612/60.7) 595 2 2 0 13 0 Total costs (€) 0 630 18,900.00 37,500.00 8,000.00 81,000.00 Conclusion These results from ARTICA registry seem to demonstrate that AI DSS is extremely cost-saving in subjects with stable chest pain across the whole range of pt-lk of CAD.

Analysis of Chinese acral and mucosal melanoma patient genomic and neoantigen profiles in cancer vaccine development: A pilot study.

e14300 Background: Acral and mucosal melanoma are rare subtypes accounting for about 3% of all melanoma cases. The cutaneous melanoma genomic landscape is well defined; however, little is known about the acral and mucosal melanoma mutational spectrum. In this pilot study, we evaluated the genomic and neo-antigen profiles and tumor mutational burden (TMB) from acral and mucosal melanoma patients with the aim of designing personalized vaccines and longitudinally tracking patients’ clinical courses. Methods: Tumor whole exome sequencing and neo-antigen profiling of 5 acral and 3 mucosal melanoma patients at Shanghai Tenth Peoples Hospital, Tongji University, China between April 2018 and January 2019 was performed using YuceBio’s proprietary analytics platform. Watsonä for Genomics, an artificial intelligence decision-support system, was used for variant interpretation and annotation. A comparative analysis was performed on Chinese acral melanoma data with the published Caucasian acral cohort from the Translational Genomics Research Institute (TGen) and The Cancer Genome Atlas (TCGA) predominantly Caucasian cutaneous melanoma data set. Results: TMB in our acral/mucosal melanoma cohort was 2.26/Megabase (Mb) compared to over 20/Mb in published cutaneous melanoma studies. Tumor neo-antigen burden (TNAB) in our group was 1.03 neo-epitopes/Mb. Low TNAB levels were associated with low TMB levels in all tumors. Incidence of BRAF and NRAS mutant cases in our cohort was 0% (0/8) and 13% (1/8) respectively compared to 19% (5/27) and 7% (2/27) of the Caucasian acral population in the TGen dataset. Incidence of BRAF and NRAS mutations in the TCGA cutaneous melanoma dataset was 54% (237/440) and 28% (125/440), respectively. Conclusions: TMB was significantly lower in acral/mucosal than in cutaneous melanoma and may be a surrogate for TNAB. Detection of BRAF and NRAS mutations, the two most prevalent driver mutations in cutaneous melanoma, were significantly lower frequencies in both Chinese and Caucasian acral melanoma patients in this study, suggesting alternate cancer drivers may exist in this subtype. Strategies to address challenges of low TNAB in vaccine development are being explored.

Clinical insights for hematological malignancies from an artificial intelligence decision-support tool.

e13023 Background: Next generation sequencing (NGS) in hematological tumors is increasingly shaping clinical treatment decisions at the point of care. While the impact of NGS panels in solid tumors is largely therapeutic, targeted sequencing in hematological tumors can additionally provide diagnostic and prognostic insights. Additional data generated in hematological tumor sequencing makes manual interpretation and annotation of variants tedious and non-scalable. In this study we compared hematological tumor variant interpretation using an artificial intelligence decision-support system, Watsonä for Genomics (WfG), with expert guided manual curation. Methods: Patients with hematological tumors at Hallym University, College of Medicine between December 2017 and December 2018, were sequenced using the 54 gene Illumina TruSight Myeloid Panel. WfG interpreted and annotated all patients’ sequencing results, a subset of which were assessed manually to ascertain concordance. Results: 54 South Korean patients with hematological malignancies were analyzed (23 Acute Myeloid Leukemia, 12 myeloproliferative neoplasm, 5 myelodysplastic syndrome, 5 multiple myeloma and 9 others). Comparison of manual and WfG interpretation of 10 randomly selected cases yielded 90% (9/10) concordance and identification of 9 clinically actionable variants (33%) not found in manual interpretation. In total, WfG identified that 71% (38/54) of all cases had at least one clinically actionable therapeutic alteration (a variant targeted by a US FDA approved drug, off-label drug, or clinical trial). 33% (18/54) of cases had genes that were targeted by a US FDA approved therapy including JAK2, IDH1, IDH2, and FLT3. In cases without therapeutic alterations, WfG identified diagnostic or prognostic insights in an additional 20% (11/54) of patients. 9% (5/54) had no clinically actionable information. Conclusions: WfG variant interpretation correlated well with manually curated expert opinion and identified clinically actionable insights missed by manual interpretation. WfG has obviated the need for labor-intensive manual curation of clinical trials and therapy, enabling our center to exponentially scale our NGS operations.

Comparative analysis of prognostic molecular signatures in Asian and Caucasian AML populations.

e13025 Background: Next generation sequencing (NGS) studies on various cancer types have revealed clinical insights on the molecular etiology of tumors and defined new treatment paradigms. However, the genomic landscape of a tumor type may be predicated on race. Application of broad-brush findings from one population to another may lead to erroneous treatment decisions. While the majority of published NGS cancer studies focus on Caucasian populations, few studies evaluate the molecular profiles of a tumor type between different demographies. In this study, acute myeloid leukemia (AML) patients of East Asian descent were sequenced at a South Korean hospital, interpreted, and compared with the Caucasian population from The Cancer Genome Atlas (TCGA). Methods: 23 South Korean AML patients were sequenced in 2018 using the 54 gene Illumina TruSight Myeloid Panel at Hallym University, College of Medicine. Orthogonal testing for FLT3-internal tandem duplication (ITD) was done by Sanger sequencing. Watsonä for Genomics (WfG), an artificial intelligence decision-support system, was used for variant interpretation and annotation. Additionally, 181 AML patients of Caucasian descent from the TCGA dataset were analyzed for comparison. Results: WfG identified at least 1 clinically actionable therapeutic alteration in 70% (16/23) of all Asian cohort cases. FLT3-ITD or tyrosine kinase domain (TKD) mutations were reported in 27% (49/181) of cases in the TCGA cohort but only 9% (2/23) of the Asian cohort. DNMT3A mutations were detected in 25% (45/181) and 74% (17/23) of the TCGA and Asian cohorts, respectively. Other oncogenic mutations in AML including NRAS, IDH1, IDH2, CEBPA, TET2, and NPM1 were represented at comparable frequencies between the two populations. Conclusions: Tumor biomarkers that affect prognosis can be informative in the clinical setting. In our study, FLT3 and DNMT3A, predictors of poor prognosis in AML, demonstrated decreased and increased respective frequencies in South Koreans compared to Caucasians, suggesting that some mutational signatures that predict cancer outcome may vary by race.

An international observational study suggests that artificial intelligence for clinical decision support optimizes anemia management in hemodialysis patients

Managing anemia in hemodialysis patients can be challenging because of competing therapeutic targets and individual variability. Because therapy recommendations provided by a decision support system can benefit both patients and doctors, we evaluated the impact of an artificial intelligence decision support system, the Anemia Control Model (ACM), on anemia outcomes. Based on patient profiles, the ACM was built to recommend suitable erythropoietic-stimulating agent doses. Our retrospective study consisted of a 12-month control phase (standard anemia care), followed by a 12-month observation phase (ACM-guided care) encompassing 752 patients undergoing hemodialysis therapy in 3 NephroCare clinics located in separate countries. The percentage of hemoglobin values on target, the median darbepoetin dose, and individual hemoglobin fluctuation (estimated from the intrapatient hemoglobin standard deviation) were deemed primary outcomes. In the observation phase, median darbepoetin consumption significantly decreased from 0.63 to 0.46 μg/kg/month, whereas on-target hemoglobin values significantly increased from 70.6% to 76.6%, reaching 83.2% when the ACM suggestions were implemented. Moreover, ACM introduction led to a significant decrease in hemoglobin fluctuation (intrapatient standard deviation decreased from 0.95 g/dl to 0.83 g/dl). Thus, ACM support helped improve anemia outcomes of hemodialysis patients, minimizing erythropoietic-stimulating agent use with the potential to reduce the cost of treatment.