Easily created prediction model using deep learning software (Prediction One, Sony Network Communications Inc.) for subarachnoid hemorrhage outcomes from small dataset at admission

A radiomics model combining machine learning and neural networks for high-accuracy prediction of cervical lymph node metastasis on ultrasound of head and neck squamous cell carcinoma

OBJECTIVE

This study aimed to develop an ultrasound image-based radiomics model for diagnosing cervical lymph node (LN) metastasis in patients with head and neck squamous cell carcinoma (HNSCC) that shows higher accuracy than previous models.

STUDY DESIGN

A total of 537 LN (260 metastatic and 277 nonmetastatic) from 126 patients (78 men, 48 women, average age 63 years) were enrolled. The multivariate analysis software Prediction One (Sony Network Communications Corporation) was used to create the diagnostic models. Furthermore, three machine learning methods were adopted as comparison approaches. Based on a combination of texture analysis results, clinical information, and ultrasound findings interpretated by specialists, a total of 12 models were created, three for each machine learning method, and their diagnostic performance was compared.

RESULTS

The three best models had area under the curve of 0.98. Parameters related to ultrasound findings, such as presence of a hilum, echogenicity, and granular parenchymal echoes, showed particularly high contributions. Other significant contributors were those from texture analysis that indicated the minimum pixel value, number of contiguous pixels with the same echogenicity, and uniformity of gray levels.

CONCLUSIONS

The radiomics model developed was able to accurately diagnose cervical LN metastasis in HNSCC.

Deep learning from head CT scans to predict elevated intracranial pressure

BACKGROUND AND PURPOSE

Elevated intracranial pressure (ICP) resulting from severe head injury or stroke poses a risk of secondary brain injury that requires neurosurgical intervention. However, currently available noninvasive monitoring techniques for predicting ICP are not sufficiently advanced. We aimed to develop a minimally invasive ICP prediction model using simple CT images to prevent secondary brain injury caused by elevated ICP.

METHODS

We used the following three methods to determine the presence or absence of elevated ICP using midbrain-level CT images: (1) a deep learning model created using the Python (PY) programming language; (2) a model based on cistern narrowing and scaling of brainstem deformities and presence of hydrocephalus, analyzed using the statistical tool Prediction One (PO); and (3) identification of ICP by senior residents (SRs). We compared the accuracy of the validation and test data using fivefold cross-validation and visualized or quantified the areas of interest in the models.

RESULTS

The accuracy of the validation data for the PY, PO, and SR methods was 83.68% (83.42%-85.13%), 85.71% (73.81%-88.10%), and 66.67% (55.96%-72.62%), respectively. Significant differences in accuracy were observed between the PY and SR methods. Test data accuracy was 77.27% (70.45%-77.2%), 84.09% (75.00%-85.23%), and 61.36% (56.82%-68.18%), respectively.

CONCLUSIONS

Overall, the outcomes suggest that these newly developed models may be valuable tools for the rapid and accurate detection of elevated ICP in clinical practice. These models can easily be applied to other sites, as a single CT image at the midbrain level can provide a highly accurate diagnosis.

Anodal Electrical Taste Stimulation to the Chin Enhances the Salt Taste Perception in Subarachnoid Hemorrhage Patients

Aneurysmal subarachnoid hemorrhage (SAH) is a critical condition associated with high mortality rates. Hypertension is a significant risk factor for SAH development and recurrence following coil embolization for a ruptured aneurysm. While reduction of salt consumption is crucial for managing hypertension, it often compromises food taste. Anodal electrical taste stimulation (ETS) has been proposed to enhance taste perception without altering salt content. We present the case of a 69-year-old female SAH patient with a ruptured aneurysm at the anterior communicating artery who underwent coil embolization and in whom we tested ETS's efficacy in enhancing the salt taste perception on day 42 after the procedure. ETS effectively enhanced the salt taste perception threshold and perceived concentration; the threshold for salt taste without electrical stimulation was 0.8% of salt-impregnated filter paper, whereas that with electrical stimulation was 0.6%. The perception of salt taste was enhanced: 0.8% and 1.0% of filter papers were perceived as 0.6% and 0.8% without electrical stimulation and 1.0% and 1.2% with electrical stimulation, respectively. This is the first report describing the salt perception-enhancing effect of ETS in an actual patient. Further studies involving actual patients are required to determine how ETS affects habitual salt intake and blood pressure trends.

Clinical prediction models for aneurysmal subarachnoid hemorrhage: a systematic review update

BACKGROUND

A systematic review of clinical prediction models for aneurysmal subarachnoid hemorrhage (aSAH) reported in 2011 noted that clinical prediction models for aSAH were developed using poor methods and were not externally validated. This study aimed to update the above review to guide the future development of predictive models in aSAH.

METHODS

We systematically searched Embase and MEDLINE databases (January 2010 to February 2022) for articles that reported the development of a clinical prediction model to predict functional outcomes in aSAH. Our reviews are based on the items included in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement (PRISMA) checklist, and on data abstracted from each study in accord with the Checklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies (CHARMS) 2014 checklist. Bias and applicability were assessed using the Prediction model Risk Of Bias Assessment Tool (PROBAST).

RESULTS

We reviewed data on 30 466 patients contributing to 29 prediction models abstracted from 22 studies identified from an initial search of 7858 studies. Most models were developed using logistic regression (n=20) or machine learning (n=9) with prognostic variables selected through a range of methods. Age (n=13), World Federation of Neurological Surgeons (WFNS) grade (n=11), hypertension (n=6), aneurysm size (n=5), Fisher grade (n=12), Hunt and Hess score (n=5), and Glasgow Coma Scale (n=8) were the variables most frequently included in the reported models. External validation was performed in only four studies. All but one model had a high or unclear risk of bias due to poor performance or lack of validation.

CONCLUSION

Externally validated models for the prediction of functional outcome in aSAH patients have now become available. However, most of them still have a high risk of bias.

A preliminary prediction model using a deep learning software program for prolonged hospitalization after cardiovascular surgery

A prolonged length of hospital stay (LOS) has become an important issue among patients undergoing cardiovascular surgery in our aging society. However, there are no established prediction models for a prolonged LOS. We therefore created a prediction model of a prolonged LOS using a deep learning software program (Prediction One; Sony Network Communications Inc., Tokyo, Japan) using preoperative data. Subjects were 157 patients (121 for training data, 36 for validation data). A prolonged LOS was defined as a more than 30-day postoperative stay due to physical inactivity. The area under the receiver operating characteristic curve and the accuracy of the model in the validation data were 0.806 and 67%, respectively. In conclusion, the preliminary model demonstrated acceptable performance for the prediction of a prolonged LOS after cardiovascular surgery.

Strokes and Predictors of Outcomes

Stroke is a leading cause of long-term disability and fifth leading cause of death. Acute ischemic stroke, intracerebral hemorrhage, and subarachnoid hemorrhage, the 3 subtypes of strokes, have varying treatment modalities. Common themes in management advocate for early interventions to reduce morbidity and mortality but not all perception is supported through randomized controlled trials. Each stroke subtype has varying premorbid-related and ictus-related outcome predictive models that have differing sensitivities and specificities.

A comparison of low- versus standard-dose bridging alteplase in acute ischemic stroke mechanical thrombectomy using indirect methods

Background

Whether low-dose alteplase is similar to standard-dose bridging alteplase prior to endovascular mechanical thrombectomy in patients with acute ischemic stroke (AIS) remains uncertain.

Aims

The aim of this study was to compare the efficacy and safety outcomes of low- versus standard-dose bridging alteplase therapy (BT) in patients with acute ischemic stroke (AIS) who are eligible for intravenous thrombolysis (IVT) within 4.5 h after onset.

Methods

We conducted an indirect comparison of low- versus standard-dose bridging alteplase before mechanical thrombectomy in AIS of current available clinical randomized controlled trials (RCTs) that compared direct mechanical thrombectomy treatment (dMT) to BT. Primary efficacy outcomes were functional independence and excellent recovery defined as a dichotomized modified Rankin Scale (mRS) 0-2 and 0-1 at 90 days. Safety outcomes included symptomatic intracranial hemorrhage (sICH) and any intracranial hemorrhage (ICH).

Results

We included six RCTs of 2334 AIS patients in this analysis, including one trial using low-dose bridging alteplase (n = 103) and five trials using standard-dose bridging alteplase (n = 1067) against a common comparator (dMT). Indirect comparisons of low- to standard-dose bridging alteplase yielded an odds ratio (OR) of 0.84 (95% CI 0.47-1.50) for 90-day mRS 0-2, 1.18 (95% CI 0.65-2.12) for 90-day mRS 0-1, 1.21 (95% CI 0.44-3.36) for mortality, and 1.11 (95% CI 0.39-3.14) for successful recanalization. There were no significant differences in the odds for sICH (OR 1.05, 95% CI 0.32-3.41) or any ICH (OR 1.71, 95% CI 0.94-3.10) between low- and standard-dose bridging alteplase.

Conclusion

Indirect evidence shows that the effects of low- and standard-dose bridging alteplase are similar for key efficacy and safety outcomes. Due to the wide confidence intervals, larger randomized trials comparing low- and standard-dose alteplase bridging therapy are required.

Carotid tortuosity is associated with extracranial carotid artery aneurysms

BACKGROUND

Tortuous arteries may be associated with carotid dissection. The intima disruption caused by a carotid dissection is a possible cause of extracranial carotid artery aneurysms (ECAAs). The aim was to investigate if carotid tortuosity is also associated with ECAA in patients without presence or history of a carotid artery dissection.

METHODS

A retrospective case-control study was performed including 35 unilateral ECAA patients (cases) and 105 age- and sex-matched controls. Tortuosity was expressed as tortuosity-index (TI), curvature, and torsion measured on computed tomography angiography (CTA) data in 3Mensio Vascular and MATLAB by two independent investigators. Primary comparison was tortuosity in ipsi- versus contralateral carotid artery within the cohort of ECAA patients. Secondary comparison was tortuosity with ipsilateral carotid arteries in control patients. All observations were assessed on inter- and intra-operator reproducibility.

RESULTS

Carotid tortuosity was comparable within the cohort of ECAA patients (Spearman correlation 0.76, P<0.001), yet distinctively higher in comparison with unilateral controls. After adjustment for patient characteristics, presence of ECAA was associated with TI (β 0.146, 95% CI: 0.100-0.192). All tortuosity observations showed excellent inter- and intra-operator reproducibility.

CONCLUSIONS

Carotid tortuosity seems to be a risk factor for development of ECAA. Surveillance of individuals with increased carotid tortuosity therefore potentially ensures prompt diagnosis and treatment of ECAA. However, future research should investigate if persons with an increased tortuosity do indeed develop ECAA.

Machine learning prediction of symptomatic intracerebral hemorrhage after stroke thrombolysis: a cross-cultural validation in Caucasian and Han Chinese cohort

Background

Previous studies found that Asians seemed to have higher risk of HT after thrombolysis than Caucasians due to its race differences in genetic polymorphism. Whether the model developed by Caucasians could predict risk of symptomatic intracerebral hemorrhage (sICH) in Asians was unknown.

Objectives

To develop a machine learning-based model for predicting sICH after stroke thrombolysis in Caucasians and externally validate it in an independent Han Chinese cohort.

Design

The derivation Caucasian sample included 1738 ischemic stroke (IS) patients from the Virtual International Stroke Trials Archive (VISTA) data set, and the external validation Han Chinese cohort included 296 IS patients who were treated with intravenous thrombolysis.

Methods

Twenty-eight variables were collected across both samples. According to their properties, we classified them into six distinct clusters (ie, demographic variables, medical history, previous medication, baseline blood biomarkers, neuroimaging markers on initial CT scan and clinical characteristics). A support vector machine (SVM) model, which consisted of data processing, model training, testing and a 10-fold cross-validation, was developed to predict the risk of sICH after stroke thrombolysis. The receiving operating characteristic (ROC) was used to assess the prediction performance of the SVM model. A domain contribution analysis was then performed to test which cluster had the highest influence on the performance of the model.

Results

In total, 85 (4.9%) patients developed sICH in the Caucasians, and 29 (9.8%) patients developed sICH in the Han Chinese cohort. Eight features including age, NIHSS score, SBP (systolic blood pressure), DBP (diastolic blood pressure), ALP (alkaline phosphatase), ALT (alanine transaminase), glucose, and creatine level were included in the final model, all of which were from demographic, clinical characteristics, and blood biomarkers clusters, respectively. The SVM model showed a good predictive performance in both Caucasians (AUC = 0.87) and Han Chinese patients (AUC = 0.74). Domain contribution analysis showed that inclusion/exclusion of clinical characteristic cluster (NIHSS score, SBP, and DBP), had the highest influence on the performance of predicting sICH in both Caucasian and Han Chinese cohorts.

Conclusion

The established SVM model is feasible for predicting the risk of sICH after thrombolysis quickly and efficiently in both Caucasian and Han Chinese cohort.

Supporting adolescents' mental health during COVID-19 by utilising lessons from the aftermath of the Great East Japan Earthquake

Historical data can determine how adolescents recover from difficult situations such as the Coronavirus disease 2019 (COVID-19) pandemic. This study analysed 3 years of data obtained from high-school students who had been affected by the 2011 Great East Japan Earthquake and consequently evidenced the importance of increasing resilience among affected adolescents. This involved identifying factors contributing to resilience through a model that assessed for each tsunami disaster. This model was determined by assessing the correlation between survivors' resilience scores and their measured psychological and lifestyle scores. This approach showed that, in all tsunami damage models, resilience was most affected by the depressed emotions. Thus, our approach suggests that interventions for improving the depressed mood may improve resilience in adolescents during the COVID-19 pandemic.

Development of an artificial intelligence-based diagnostic model for Alzheimer's disease

Introduction

The diagnosis of Alzheimer's disease (AD) is sometimes difficult for nonspecialists, resulting in misdiagnosis. A missed diagnosis can lead to improper management and poor outcomes. Moreover, nonspecialists lack a simple diagnostic model with high accuracy for AD diagnosis.

Methods

Randomly assigned data, including training data, of 6000 patients and test data of 1932 from 7932 patients who visited our memory clinic between 2009 and 2021 were introduced into the artificial intelligence (AI)-based AD diagnostic model, which we had developed.

Results

The AI-based AD diagnostic model used age, sex, Hasegawa's Dementia Scale-Revised, the Mini-Mental State Examination, the educational level, and the voxel-based specific regional analysis system for Alzheimer's disease (VSRAD) score. It had a sensitivity, specificity, and c-static value of 0.954, 0.453, and 0.819, respectively. The other AI-based model that did not use the VSRAD had a sensitivity, specificity, and c-static value of 0.940, 0.504, and 0.817, respectively.

Discussion

We created an AD diagnostic model with high sensitivity for AD diagnosis using only data acquired in daily clinical practice. By using these AI-based models, nonspecialists could reduce missed diagnoses and contribute to the appropriate use of medical resources.

Venous Flow Profiles on Perfusion CT are Associated with Futile Recanalization After Thrombectomy

Background and Purpose

Robust venous outflow (VO) reflects favourable tissue reperfusion in acute ischaemic stroke (AIS) patients with large vessel occlusion (LVO). We aimed to investigate the association of the venous outflow profile on computed tomographic perfusion (CTP) and futile recanalization in anterior circulation AIS patients with LVO after thrombectomy.

Methods

This was a retrospective study of consecutive AIS patients due to anterior circulation LVO who underwent CTP before thrombectomy. Patients who achieved successful recanalization defined as a modified Thrombolysis in Cerebral Infarction (mTICI) score of 2b or 3 after thrombectomy were included. Based on the venous time-intensity curve of CTP, the peak time of venous outflow (PTV), total venous outflow time (TVT), and difference value of arteriovenous peak time (D-value) were recorded. A modified mRS score of 3-6 at 3 months was regarded as futile recanalization (FR). Logistic regression analysis was applied to assess risk factors for FR. We used receiver operating characteristic curves (ROCs) to evaluate the predictive value of venous outflow time parameters based on VO for FR.

Results

Eighty patients were included; 35 (43.8%) achieved good functional outcomes, and 45 (56.3%) had unfavourable functional outcomes, that is, FR. Adjusting confounding factors, binary stepwise logistic regression analysis showed that delayed PTV was independently associated with FR (odds ratio, 1.374 [95% CI, 1.093-1.726], P = 0.007). ROCs indicated that PTV effectively predicted unfavourable outcomes at 3 months (area under the curve (AUC) = 0.729, p< 0.001). The combined model was a powerful predictor of FR with an AUC of 0.824 and a cut-off value of 0.631 (p< 0.001).

Conclusion

Delayed PTV is independently related to FR in anterior circulation AIS patients with LVO achieving successful recanalization after thrombectomy. Our results highlight that the time of VO may be a potential marker for FR.

A Magnetic Resonance Angiography-Based Study Comparing Machine Learning and Clinical Evaluation: Screening Intracranial Regions Associated with the Hemorrhagic Stroke of Adult Moyamoya Disease

OBJECTIVES

Moyamoya disease patients with hemorrhagic stroke usually have a poor prognosis. This study aimed to determine whether hemorrhagic moyamoya disease could be distinguished from MRA images using transfer deep learning and to screen potential regions that contain rich distinguishing information from MRA images in moyamoya disease.

MATERIALS AND METHODS

A total of 116 adult patients with bilateral moyamoya diseases suffering from hemorrhagic or ischemia complications were retrospectively screened. Based on original MRA images at the level of the basal cistern, basal ganglia, and centrum semiovale, we adopted the pretrained ResNet18 to build three models for differentiating hemorrhagic moyamoya disease. Grad-CAM was applied to visualize the regions of interest.

RESULTS

For the test set, the accuracies of model differentiation in the basal cistern, basal ganglia, and centrum semiovale were 93.3%, 91.5%, and 86.4%, respectively. Visualization of the regions of interest demonstrated that the models focused on the deep and periventricular white matter and abnormal collateral vessels in hemorrhagic moyamoya disease.

CONCLUSION

A transfer learning model based on MRA images of the basal cistern and basal ganglia showed a good ability to differentiate between patients with hemorrhagic moyamoya disease and those with ischemic moyamoya disease. The deep and periventricular white matter and collateral vessels at the level of the basal cistern and basal ganglia may contain rich distinguishing information.

External Validation of a Neural Network Model in Aneurysmal Subarachnoid Hemorrhage: A Comparison With Conventional Logistic Regression Models

BACKGROUND

Interest in machine learning (ML)-based predictive modeling has led to the development of models predicting outcomes after aneurysmal subarachnoid hemorrhage (aSAH), including the Nijmegen acute subarachnoid hemorrhage calculator (Nutshell). Generalizability of such models to external data remains unclear.

OBJECTIVE

To externally validate the performance of the Nutshell tool while comparing it with the conventional Subarachnoid Hemorrhage International Trialists (SAHIT) models and to review the ML literature on outcome prediction after aSAH and aneurysm treatment.

METHODS

A prospectively maintained database of patients with aSAH presenting consecutively to our institution in the 2013 to 2018 period was used. The web-based Nutshell and SAHIT calculators were used to derive the risks of poor long-term (12-18 months) outcomes and 30-day mortality. Discrimination was evaluated using the area under the curve (AUC), and calibration was investigated using calibration plots. The literature on relevant ML models was surveyed for a synopsis.

RESULTS

In 269 patients with aSAH, the SAHIT models outperformed the Nutshell tool (AUC: 0.786 vs 0.689, P = .025) in predicting long-term functional outcomes. A logistic regression model of the Nutshell variables derived from our data achieved adequate discrimination (AUC = 0.759) of poor outcomes. The SAHIT models outperformed the Nutshell tool in predicting 30-day mortality (AUC: 0.810 vs 0.636, P < .001). Calibration properties were more favorable for the SAHIT models. Most published aneurysm-related ML-based outcome models lack external validation and usable testing platforms.

CONCLUSION

The Nutshell tool demonstrated limited performance on external validation in comparison with the SAHIT models. External validation and the dissemination of testing platforms for ML models must be emphasized.

The Effectiveness of Machine Learning in Predicting Lateral Lymph Node Metastasis From Lower Rectal Cancer: A Single Center Development and Validation Study

AIM

Accurate preoperative diagnosis of lateral lymph node metastasis (LLNM) from lower rectal cancer is important to identify patients who require lateral lymph node dissection (LLND). We aimed to create an effective prediction model for LLNM using machine learning by combining preoperative information.

METHODS

We retrospectively examined patients who underwent primary rectal cancer surgery with unilateral or bilateral LLND between April 2010 and March 2020 at a single institution. Using the machine learning software "Prediction One" (Sony Network Communications), we developed a prediction model in the training cohort that included 267 consecutive patients (500 sides) from April 2010. Clinicopathological data obtained from the preoperative examinations were used as the learning items. In the validation cohort that included subsequent patients until March 2020, we compared the discriminating powers of the prediction model and the conventional method using the short-axis diameter of the largest lateral lymph node, as detected on magnetic resonance imaging.

RESULTS

The area under the receiver operating characteristic curve (AUC) of the prediction model was 0.903 in the validation cohort comprising 56 patients (107 sides). This indicated significantly higher predictive power than that of the conventional method (AUC = 0.754; P = .022). Using the cutoff values defined in the training cohort, the accuracy, sensitivity, and specificity of the prediction model were 80.4%, 90.0%, and 79.4%, respectively. The model was able to correctly predict four of five sides comprising LLNM with the short-axis diameters ≤4 mm.

CONCLUSION

Machine learning contributed to the creation of an effective prediction model for LLNM.

Deep Learning for Outcome Prediction in Neurosurgery: A Systematic Review of Design, Reporting, and Reproducibility

Deep learning (DL) is a powerful machine learning technique that has increasingly been used to predict surgical outcomes. However, the large quantity of data required and lack of model interpretability represent substantial barriers to the validity and reproducibility of DL models. The objective of this study was to systematically review the characteristics of DL studies involving neurosurgical outcome prediction and to assess their bias and reporting quality. Literature search using the PubMed, Scopus, and Embase databases identified 1949 records of which 35 studies were included. Of these, 32 (91%) developed and validated a DL model while 3 (9%) validated a pre-existing model. The most commonly represented subspecialty areas were oncology (16 of 35, 46%), spine (8 of 35, 23%), and vascular (6 of 35, 17%). Risk of bias was low in 18 studies (51%), unclear in 5 (14%), and high in 12 (34%), most commonly because of data quality deficiencies. Adherence to transparent reporting of a multivariable prediction model for individual prognosis or diagnosis reporting standards was low, with a median of 12 transparent reporting of a multivariable prediction model for individual prognosis or diagnosis items (39%) per study not reported. Model transparency was severely limited because code was provided in only 3 studies (9%) and final models in 2 (6%). With the exception of public databases, no study data sets were readily available. No studies described DL models as ready for clinical use. The use of DL for neurosurgical outcome prediction remains nascent. Lack of appropriate data sets poses a major concern for bias. Although studies have demonstrated promising results, greater transparency in model development and reporting is needed to facilitate reproducibility and validation.

Deep Learning-Based Functional Independence Measure Score Prediction After Stroke in Kaifukuki (Convalescent) Rehabilitation Ward Annexed to Acute Care Hospital

Introduction

Prediction models of functional independent measure (FIM) score after kaifukuki (convalescent) rehabilitation ward (KRW) are needed to decide the treatment strategies and save medical resources. Statistical models were reported, but their accuracies were not satisfactory. We made such prediction models using the deep learning (DL) framework, Prediction One (Sony Network Communications Inc., Tokyo, Japan).

Methods

Of the 559 consecutive stroke patients, 122 patients were transferred to our KRW. We divided our 122 patients’ data randomly into halves of training and validation datasets. Prediction One made three prediction models from the training dataset using (1) variables at the acute care ward admission, (2) those at the KRW admission, and (3) those combined (1) and (2). The models’ determination coefficients (R²), correlation coefficients (rs), and residuals were calculated using the validation dataset.

Results

Of the 122 patients, the median age was 71, length of stay (LOS) in acute care ward 23 (17-30) days, LOS in KRW 53 days, total FIM scores at the admission of KRW 85, those at discharge 108. The mean FIM gain and FIM efficiency were 19 and 0.417. All patients were discharged home. Model (1), (2), and (3)’s R² were 0.794, 0.970, and 0.972. Their mean residuals between the predicted and actual total FIM scores were -1.56±24.6, -4.49±17.1, and -2.69±15.7.

Conclusion

Our FIM gain and efficiency were better than national averages of FIM gain 17.1 and FIM efficiency 0.187. We made DL-based total FIM score prediction models, and their accuracies were superior to those of previous statistically calculated ones. The DL-based FIM score prediction models would save medical costs and perform efficient stroke and rehabilitation medicine.

Relationship Between Medical Questionnaire and Influenza Rapid Test Positivity: Subjective Pretest Probability, "I Think I Have Influenza," Contributes to the Positivity Rate

Introduction Rapid influenza diagnostic tests (RIDTs) are considered essential for determining when to start influenza treatment using anti-influenza drugs, but their accuracy is about 70%. Under the COVID-19 pandemic, we hope to refrain from performing unnecessary RIDTs considering droplet infection of COVID-19 and influenza. We re-examined the medical questionnaire's importance and its relationship to the positivity of RIDTs. Then we built a positivity prediction model for RIDTs using automated artificial intelligence (AI). Methods We retrospectively investigated 96 patients who underwent RIDTs at the outpatient department from December 2019 to March 2020. We used a questionnaire sheet with 24 items before conducting RIDTs. The factors associated with the positivity of RIDTs were statistically analyzed. We then used an automated AI framework to produce the positivity prediction model using the 24 items, sex, and age, with five-fold cross-validation. Results Of the 47 women and 49 men (median age was 39 years), 56 patients were RIDT positive with influenza A. The AI-based model using 26 variables had an area under the curve (AUC) of 0.980. The stronger variables are subjective pretest probability, which is a numerically described score ranging from 0% to 100% of "I think I have influenza," cough, past hours after the onset, muscle pain, and maximum body temperature in order. Conclusion We easily built the RIDT positivity prediction model using automated AI. Its AUC was satisfactory, and it suggested the importance of a detailed medical interview. Both the univariate analysis and AI-based model suggested that subjective pretest probability, "I think I have influenza," might be useful.

Easily Created Prediction Model Using Automated Artificial Intelligence Framework (Prediction One, Sony Network Communications Inc., Tokyo, Japan) for Subarachnoid Hemorrhage Outcomes Treated by Coiling and Delayed Cerebral Ischemia

Introduction Reliable prediction models of subarachnoid hemorrhage (SAH) outcomes and delayed cerebral ischemia (DCI) are needed to decide the treatment strategy. Automated artificial intelligence (AutoAI) is attractive, but there are few reports on AutoAI-based models for SAH functional outcomes and DCI. We herein made models using an AutoAI framework, Prediction One (Sony Network Communications Inc., Tokyo, Japan), and compared it to other previous statistical prediction scores. Methods We used an open dataset of 298 SAH patients, who were with non-severe neurological grade and treated by coiling. Modified Rankin Scale 0-3 at six months was defined as a favorable functional outcome and DCI occurrence as another outcome. We randomly divided them into a 248-patient training dataset and a 50-patient test dataset. Prediction One made the model using training dataset with 5-fold cross-validation. We evaluated the model using the test dataset and compared the area under the curves (AUCs) of the created models. Those of the modified SAFIRE score and the Fisher computed tomography (CT) scale to predict the outcomes. Results The AUCs of the AutoAI-based models for functional outcome in the training and test dataset were 0.994 and 0.801, and those for the DCI occurrence were 0.969 and 0.650. AUCs for functional outcome calculated using modified SAFIRE score were 0.844 and 0.892. Those for the DCI occurrence calculated using the Fisher CT scale were 0.577 and 0.544. Conclusions We easily and quickly made AutoAI-based prediction models. The models' AUCs were not inferior to the previous prediction models despite the easiness.

Postsurgical functional outcome prediction model using deep learning framework (Prediction One, Sony Network Communications Inc.) for hypertensive intracerebral hemorrhage

BACKGROUND

Reliable prediction models of intracerebral hemorrhage (ICH) outcomes are needed for decision-making of the treatment. Statistically making such prediction models needs a large number of samples and time-consuming statistical analysis. Deep learning (DL), one of the artificial intelligence, is attractive, but there were no reports on DL-based functional outcome prediction models for ICH outcomes after surgery. We herein made a functional outcome prediction model using DLframework, Prediction One (Sony Network Communications Inc., Tokyo, Japan), and compared it to original ICH score, ICH Grading Scale, and FUNC score.

METHODS

We used 140 consecutive hypertensive ICH patients' data in our hospital between 2012 and 2019. All patients were surgically treated. Modified Rankin Scale 0-3 at 6 months was defined as a favorable outcome. We randomly divided them into 100 patients training dataset and 40 patients validation dataset. Prediction One made the prediction model using the training dataset with 5-fold cross-validation. We calculated area under the curves (AUCs) regarding the outcome using the DL-based model, ICH score, ICH Grading Scale, and FUNC score. The AUCs were compared.

RESULTS

The model made by Prediction One using 64 variables had AUC of 0.997 in the training dataset and that of 0.884 in the validation dataset. These AUCs were superior to those derived from ICH score, ICH Grading Scale, and FUNC score.

CONCLUSION

We easily and quickly made prediction models using Prediction One, even with a small single-center dataset. The accuracy of the DL-based model was superior to those of previous statistically calculated models.

Prediction Model of Deep Learning for Ambulance Transports in Kesennuma City by Meteorological Data

PURPOSE

With the aging population in Japan, the prediction of ambulance transports is needed to save the limited medical resources. Some meteorological factors were risks of ambulance transports, but it is difficult to predict in a classically statistical way because Japan has 4 seasons. We tried to make prediction models for ambulance transports using the deep learning (DL) framework, Prediction One (Sony Network Communications Inc., Tokyo, Japan), with the meteorological and calendarial variables.

MATERIALS AND METHODS

We retrospectively investigated the daily ambulance transports and meteorological data between 2017 and 2019. First, to confirm their association, we performed classically statistical analysis. Second, to test the DL framework's utility for ambulance transports prediction, we made 3 prediction models for daily ambulance transports (total daily ambulance transports more than 5 or not, cardiopulmonary arrest (CPA), and trauma) using meteorological and calendarial factors and evaluated their accuracies by internal cross-validation.

RESULTS

During the 1095 days of 3 years, the total ambulance transports were 5948, including 240 CPAs and 337 traumas. Cardiogenic CPA accounted for 72.3%, according to the Utstein classification. The relation between ambulance transports and meteorological parameters by polynomial curves were statistically obtained, but their r2s were small. On the other hand, all DL-based prediction models obtained satisfactory accuracies in the internal cross-validation. The areas under the curves obtained from each model were all over 0.947.

CONCLUSION

We could statistically make polynomial curves between the meteorological variables and the number of ambulance transport. We also preliminarily made DL-based prediction models. The DL-based prediction for daily ambulance transports would be used in the future, leading to solving the lack of medical resources in Japan.

Preliminary development of a prediction model for daily stroke occurrences based on meteorological and calendar information using deep learning framework (Prediction One; Sony Network Communications Inc., Japan)

BACKGROUND

Chronologically meteorological and calendar factors were risks of stroke occurrence. However, the prediction of stroke occurrences is difficult depending on only meteorological and calendar factors. We tried to make prediction models for stroke occurrences using deep learning (DL) software, Prediction One (Sony Network Communications Inc., Tokyo, Japan), with those variables.

METHODS

We retrospectively investigated the daily stroke occurrences between 2017 and 2019. We used Prediction One software to make the prediction models for daily stroke occurrences (present or absent) using 221 chronologically meteorological and calendar factors. We made a prediction models from the 3-year dataset and evaluated their accuracies using the internal cross-validation. Areas under the curves (AUCs) of receiver operating characteristic curves were used as accuracies.

RESULTS

The 371 cerebral infarction (CI), 184 intracerebral hemorrhage (ICH), and 53 subarachnoid hemorrhage patients were included in the study. The AUCs of the several DL-based prediction models for all stroke occurrences were 0.532-0.757. Those for CI were 0.600-0.782. Those for ICH were 0.714-0.988.

CONCLUSION

Our preliminary results suggested a probability of the DL-based prediction models for stroke occurrence only by meteorological and calendar factors. In the future, by synchronizing a variety of medical information among the electronic medical records and personal smartphones as well as integrating the physical activities or meteorological conditions in real time, the prediction of stroke occurrence could be performed with high accuracy, to save medical resources, to have patients care for themselves, and to perform efficient medicine.

Preliminary development of a deep learning-based automated primary headache diagnosis model using Japanese natural language processing of medical questionnaire

BACKGROUND

Primary headaches are widespread and costly public health problems. However, there are insufficient medical resources for their treatment in Japan due to two reasons. First, the numbers of headache specialists and clinics remain insufficient. Second, neurologists and neurosurgeons mainly treat headaches in Japan. However, they mainly work as general stroke neurologists, so they cannot focus on primary headache treatment. To solve these problems, we preliminarily developed a deep learning (DL)-based automated diagnosis model from patients' Japanese unstructured sentences in the medical questionnaire using a DL framework. We hypothesized that the model would reduce the time and burden on both doctors and patients and improve their quality of life.

METHODS

We retrospectively investigated our primary headache database and developed a diagnosis model using the DL framework (Prediction One, Sony Network Communications Inc., Japan). We used age, sex, date, and embedding layer made by the medical questionnaire's natural language processing (NLP).

RESULTS

Eight hundred and forty-eight primary headache patients (495 women and 353 men) are included. The median (interquartile range) age was 59 (40-74). Migraine accounted for 46%, tension-type headache for 47%, trigeminal autonomic cephalalgias for 5%, and other primary headache disorders for 2%. The accuracy, mean precision, mean recall, and mean F value of the developed diagnosis model were 0.7759, 0.8537, 0.6086, and 0.6353, which were satisfactory.

CONCLUSION

The DL-based diagnosis model for primary headaches using the raw medical questionnaire's Japanese NLP would be useful in performing efficient medical practice after ruling out the secondary headaches.