Fully Automated Tumor Bud Assessment in Hematoxylin and Eosin-Stained Whole Slide Images of Colorectal Cancer.

Tumor budding (TB), the presence of single cells or small clusters of up to 4 tumor cells at the invasive front of colorectal cancer (CRC), is a proven risk factor for adverse outcomes. International definitions are necessary to reduce interobserver variability. According to the current international guidelines, hotspots at the invasive front should be counted in hematoxylin and eosin (H&E)-stained slides. This is time-consuming and prone to interobserver variability; therefore, there is a need for computer-aided diagnosis solutions. In this study, we report an artificial intelligence-based method for detecting TB in H&E-stained whole slide images. We propose a fully automated pipeline to identify the tumor border, detect tumor buds, characterize them based on the number of tumor cells, and produce a TB density map to identify the TB hotspot. The method outputs the TB count in the hotspot as a computational biomarker. We show that the proposed automated TB detection workflow performs on par with a panel of 5 pathologists at detecting tumor buds and that the hotspot-based TB count is an independent prognosticator in both the univariate and the multivariate analysis, validated on a cohort of n = 981 patients with CRC. Computer-aided detection of tumor buds based on deep learning can perform on par with expert pathologists for the detection and quantification of tumor buds in H&E-stained CRC histopathology slides, strongly facilitating the introduction of budding as an independent prognosticator in clinical routine and clinical trials.

Automated approach for quality assessment of RDF resources.

INTRODUCTION: The Semantic Web community provides a common Resource Description Framework (RDF) that allows representation of resources such that they can be linked. To maximize the potential of linked data - machine-actionable interlinked resources on the Web - a certain level of quality of RDF resources should be established, particularly in the biomedical domain in which concepts are complex and high-quality biomedical ontologies are in high demand. However, it is unclear which quality metrics for RDF resources exist that can be automated, which is required given the multitude of RDF resources. Therefore, we aim to determine these metrics and demonstrate an automated approach to assess such metrics of RDF resources. METHODS: An initial set of metrics are identified through literature, standards, and existing tooling. Of these, metrics are selected that fulfil these criteria: (1) objective; (2) automatable; and (3) foundational. Selected metrics are represented in RDF and semantically aligned to existing standards. These metrics are then implemented in an open-source tool. To demonstrate the tool, eight commonly used RDF resources were assessed, including data models in the healthcare domain (HL7 RIM, HL7 FHIR, CDISC CDASH), ontologies (DCT, SIO, FOAF, ORDO), and a metadata profile (GRDDL). RESULTS: Six objective metrics are identified in 3 categories: Resolvability (1), Parsability (1), and Consistency (4), and represented in RDF. The tool demonstrates that these metrics can be automated, and application in the healthcare domain shows non-resolvable URIs (ranging from 0.3% to 97%) among all eight resources and undefined URIs in HL7 RIM, and FHIR. In the tested resources no errors were found for parsability and the other three consistency metrics for correct usage of classes and properties. CONCLUSION: We extracted six objective and automatable metrics from literature, as the foundational quality requirements of RDF resources to maximize the potential of linked data. Automated tooling to assess resources has shown to be effective to identify quality issues that must be avoided. This approach can be expanded to incorporate more automatable metrics so as to reflect additional quality dimensions with the assessment tool implementing more metrics.

Automatic Post-Stroke Severity Assessment Using Novel Unsupervised Consensus Learning for Wearable and Camera-Based Sensor Datasets.

Stroke survivors often suffer from movement impairments that significantly affect their daily activities. The advancements in sensor technology and IoT have provided opportunities to automate the assessment and rehabilitation process for stroke survivors. This paper aims to provide a smart post-stroke severity assessment using AI-driven models. With the absence of labelled data and expert assessment, there is a research gap in providing virtual assessment, especially for unlabeled data. Inspired by the advances in consensus learning, in this paper, we propose a consensus clustering algorithm, PSA-NMF, that combines various clusterings into one united clustering, i.e., cluster consensus, to produce more stable and robust results compared to individual clustering. This paper is the first to investigate severity level using unsupervised learning and trunk displacement features in the frequency domain for post-stroke smart assessment. Two different methods of data collection from the U-limb datasets-the camera-based method (Vicon) and wearable sensor-based technology (Xsens)-were used. The trunk displacement method labelled each cluster based on the compensatory movements that stroke survivors employed for their daily activities. The proposed method uses the position and acceleration data in the frequency domain. Experimental results have demonstrated that the proposed clustering method that uses the post-stroke assessment approach increased the evaluation metrics such as accuracy and F-score. These findings can lead to a more effective and automated stroke rehabilitation process that is suitable for clinical settings, thus improving the quality of life for stroke survivors.

Machine Learning-Enabled Fully Automated Assessment of Left Ventricular Volume, Ejection Fraction and Strain: Experience in Pediatric and Young Adult Echocardiography.

BACKGROUND: Left ventricular (LV) volumes, ejection fraction (EF), and myocardial strain have been shown to be predictive of clinical and subclinical heart disease. Automation of LV functional assessment overcomes difficult technical challenges and complexities. We sought to assess whether a fully automated assessment of LV function could be reliably used in children and young adults. METHODS: Fifty normal volunteers (22/28, female/male) were prospectively recruited for research echocardiography. LV volumes, EF, and strain were measured both manually and automatically. An experienced sonographer performed all the manual analysis and recorded the analysis timing. The fully automated analyses were accomplished by 5 groups of observers with different knowledge and medical background. AutoLV and AutoSTRAIN (TomTec) were employed for the fully automated LV analysis. The LV volumes, EF, strain, and analysis time were compared between manual and automated methods, and among the 5 groups of observers. RESULTS: Software-determined endocardial border detection was achievable in all subjects. The analysis times of the experienced sonographer were significantly shorter for AutoLV and AutoSTRAIN than manual analyses (both p < 0.001). Strong correlations were seen between conventional EF and AutoLV (r = 0.8373), and between conventional three view global longitudinal strain (GLS) and AutoSTRAIN (r = 0.9766). The volumes from AutoLV and three view GLS from AutoSTRAIN had strong correlations among different observers regardless of level of expertise. EF from AutoLV analysis had moderately strong correlations among different observers. CONCLUSION: Automated pediatric LV analysis is feasible in normal hearts. Machine learning-enabled image analysis saves time and produces results that are comparable to traditional methods.

Automated phonological analysis and treatment target selection using AutoPATT.

Automated analyses of speech samples can offer improved accuracy and timesaving advantages that streamline clinical assessment for children with a suspected speech sound disorder. In this paper, we introduce AutoPATT, an automated tool for clinical analysis of speech samples. This free, open-source tool was developed as a plug-in for Phon and follows the procedures of the Phonological Analysis and Treatment Target Selection protocol, including extraction of a phonetic inventory, phonemic inventory with corresponding minimal pairs, and initial consonant cluster inventory. AutoPATT also provides suggestions for complex treatment targets using evidence-based guidelines. Automated analyses and target suggestions were compared to manual analyses of 25 speech samples from children with phonological disorder. Results indicate that AutoPATT inventory analyses are more accurate than manual analyses. However, treatment targets generated by AutoPATT should be viewed as suggestions and not used to substitute necessary clinical judgement in the target selection process.

Using Automatic Assessment of Speech Production to Predict Current and Future Cognitive Function in Older Adults.

Neurodegenerative conditions like Alzheimer disease affect millions and have no known cure, making early detection important. In addition to memory impairments, dementia causes substantial changes in speech production, particularly lexical-semantic characteristics. Existing clinical tools for detecting change often require considerable expertise or time, and efficient methods for identifying persons at risk are needed. This study examined whether early stages of cognitive decline can be identified using an automated calculation of lexical-semantic features of participants' spontaneous speech. Unimpaired or mildly impaired older adults (N = 39, mean 81 years old) produced several monologues (picture descriptions and expository descriptions) and completed a neuropsychological battery, including the Modified Mini-Mental State Exam. Most participants (N = 30) returned one year later for follow-up. Lexical-semantic features of participants' speech (particularly lexical frequency) were significantly correlated with cognitive status at the same visit and also with cognitive status one year in the future. Thus, automated analysis of speech production is closely associated with current and future cognitive test performance and could provide a novel, scalable method for longitudinal tracking of cognitive health.

Monitoring of Gait Parameters in Post-Stroke Individuals: A Feasibility Study Using RGB-D Sensors.

Stroke is one of the most significant causes of permanent functional impairment and severe motor disability. Hemiplegia or hemiparesis are common consequences of the acute event, which negatively impacts daily life and requires continuous rehabilitation treatments to favor partial or complete recovery and, consequently, to regain autonomy, independence, and safety in daily activities. Gait impairments are frequent in stroke survivors. The accurate assessment of gait anomalies is therefore crucial and a major focus of neurorehabilitation programs to prevent falls or injuries. This study aims to estimate, using a single RGB-D sensor, gait patterns and parameters on a short walkway. This solution may be suitable for monitoring the improvement or worsening of gait disorders, including in domestic and unsupervised scenarios. For this purpose, some of the most relevant spatiotemporal parameters, estimated by the proposed solution on a cohort of post-stroke individuals, were compared with those estimated by a gold standard system for a simultaneous instrumented 3D gait analysis. Preliminary results indicate good agreement, accuracy, and correlation between the gait parameters estimated by the two systems. This suggests that the proposed solution may be employed as an intermediate tool for gait analysis in environments where gold standard systems are impractical, such as home and ecological settings in real-life contexts.

Validation of a Laser Ranged Scanner-Based Detection of Spatio-Temporal Gait Parameters Using the aTUG Chair.

This article covers the suitability to measure gait-parameters via a Laser Range Scanner (LRS) that was placed below a chair during the walking phase of the Timed Up&Go Test in a cohort of 92 older adults (mean age 73.5). The results of our study demonstrated a high concordance of gait measurements using a LRS in comparison to the reference GAITRite walkway. Most of aTUG's gait parameters demonstrate a strong correlation coefficient with the GAITRite, indicating high measurement accuracy for the spatial gait parameters. Measurements of velocity had a correlation coefficient of 99%, which can be interpreted as an excellent measurement accuracy. Cadence showed a slightly lower correlation coefficient of 96%, which is still an exceptionally good result, while step length demonstrated a correlation coefficient of 98% per leg and stride length with an accuracy of 99% per leg. In addition to confirming the technical validation of the aTUG regarding its ability to measure gait parameters, we compared results from the GAITRite and the aTUG for several parameters (cadence, velocity, and step length) with results from the Berg Balance Scale (BBS) and the Activities-Specific Balance Confidence-(ABC)-Scale assessments. With confidence coefficients for BBS and velocity, cadence and step length ranging from 0.595 to 0.798 and for ABC ranging from 0.395 to 0.541, both scales demonstrated only a medium-sized correlation. Thus, we found an association of better walking ability (represented by the measured gait parameters) with better balance (BBC) and balance confidence (ABC) overall scores via linear regression. This results from the fact that the BBS incorporates both static and dynamic balance measures and thus, only partly reflects functional requirements for walking. For the ABC score, this effect was even more pronounced. As this is to our best knowledge the first evaluation of the association between gait parameters and these balance scores, we will further investigate this phenomenon and aim to integrate further measures into the aTUG to achieve an increased sensitivity for balance ability.

Automated assessment of cortical mastoidectomy performance in virtual reality.

INTRODUCTION: Cortical mastoidectomy is a core skill that Otolaryngology trainees must gain competency in. Automated competency assessments have the potential to reduce assessment subjectivity and bias, as well as reducing the workload for surgical trainers. OBJECTIVES: This study aimed to develop and validate an automated competency assessment system for cortical mastoidectomy. PARTICIPANTS: Data from 60 participants (Group 1) were used to develop and validate an automated competency assessment system for cortical mastoidectomy. Data from 14 other participants (Group 2) were used to test the generalisability of the automated assessment. DESIGN: Participants drilled cortical mastoidectomies on a virtual reality temporal bone simulator. Procedures were graded by a blinded expert using the previously validated Melbourne Mastoidectomy Scale: a different expert assessed procedures by Groups 1 and 2. Using data from Group 1, simulator metrics were developed to map directly to the individual items of this scale. Metric value thresholds were calculated by comparing automated simulator metric values to expert scores. Binary scores per item were allocated using these thresholds. Validation was performed using random sub-sampling. The generalisability of the method was investigated by performing the automated assessment on mastoidectomies performed by Group 2, and correlating these with scores of a second blinded expert. RESULTS: The automated binary score compared with the expert score per item had an accuracy, sensitivity and specificity of 0.9450, 0.9547 and 0.9343, respectively, for Group 1; and 0.8614, 0.8579 and 0.8654, respectively, for Group 2. There was a strong correlation between the total scores per participant assigned by the expert and calculated by the automatic assessment method for both Group 1 (r = .9144, P < .0001) and Group 2 (r = .7224, P < .0001). CONCLUSION: This study outlines a virtual reality-based method of automated assessment of competency in cortical mastoidectomy, which proved comparable to the assessment provided by human experts.

Using fuzzy string matching for automated assessment of listener transcripts in speech intelligibility studies.

Many studies of speech perception assess the intelligibility of spoken sentence stimuli by means of transcription tasks ('type out what you hear'). The intelligibility of a given stimulus is then often expressed in terms of percentage of words correctly reported from the target sentence. Yet scoring the participants' raw responses for words correctly identified from the target sentence is a time-consuming task, and hence resource-intensive. Moreover, there is no consensus among speech scientists about what specific protocol to use for the human scoring, limiting the reliability of human scores. The present paper evaluates various forms of fuzzy string matching between participants' responses and target sentences, as automated metrics of listener transcript accuracy. We demonstrate that one particular metric, the token sort ratio, is a consistent, highly efficient, and accurate metric for automated assessment of listener transcripts, as evidenced by high correlations with human-generated scores (best correlation: r = 0.940) and a strong relationship to acoustic markers of speech intelligibility. Thus, fuzzy string matching provides a practical tool for assessment of listener transcript accuracy in large-scale speech intelligibility studies. See https://tokensortratio.netlify.app for an online implementation.

An interpretable natural language processing system for written medical examination assessment.

OBJECTIVE: The assessment of written medical examinations is a tedious and expensive process, requiring significant amounts of time from medical experts. Our objective was to develop a natural language processing (NLP) system that can expedite the assessment of unstructured answers in medical examinations by automatically identifying relevant concepts in the examinee responses. MATERIALS AND METHODS: Our NLP system, Intelligent Clinical Text Evaluator (INCITE), is semi-supervised in nature. Learning from a limited set of fully annotated examples, it sequentially applies a series of customized text comparison and similarity functions to determine if a text span represents an entry in a given reference standard. Combinations of fuzzy matching and set intersection-based methods capture inexact matches and also fragmented concepts. Customizable, dynamic similarity-based matching thresholds allow the system to be tailored for examinee responses of different lengths. RESULTS: INCITE achieved an average F1-score of 0.89 (precision = 0.87, recall = 0.91) against human annotations over held-out evaluation data. Fuzzy text matching, dynamic thresholding and the incorporation of supervision using annotated data resulted in the biggest jumps in performances. DISCUSSION: Long and non-standard expressions are difficult for INCITE to detect, but the problem is mitigated by the use of dynamic thresholding (i.e., varying the similarity threshold for a text span to be considered a match). Annotation variations within exams and disagreements between annotators were the primary causes for false positives. Small amounts of annotated data can significantly improve system performance. CONCLUSIONS: The high performance and interpretability of INCITE will likely significantly aid the assessment process and also help mitigate the impact of manual assessment inconsistencies.

Feasibility of Home-Based Automated Assessment of Postural Instability and Lower Limb Impairments in Parkinson's Disease.

A self-managed, home-based system for the automated assessment of a selected set of Parkinson's disease motor symptoms is presented. The system makes use of an optical RGB-Depth device both to implement its gesture-based human computer interface and for the characterization and the evaluation of posture and motor tasks, which are specified according to the Unified Parkinson's Disease Rating Scale (UPDRS). Posture, lower limb movements and postural instability are characterized by kinematic parameters of the patient movement. During an experimental campaign, the performances of patients affected by Parkinson's disease were simultaneously scored by neurologists and analyzed by the system. The sets of parameters which best correlated with the UPDRS scores of subjects' performances were then used to train supervised classifiers for the automated assessment of new instances of the tasks. Results on the system usability and the assessment accuracy, as compared to clinical evaluations, indicate that the system is feasible for an objective and automated assessment of Parkinson's disease at home, and it could be the basis for the development of neuromonitoring and neurorehabilitation applications in a telemedicine framework.

Cost-Effectiveness of a Technology-Facilitated Depression Care Management Adoption Model in Safety-Net Primary Care Patients with Type 2 Diabetes.

BACKGROUND: The Diabetes-Depression Care-Management Adoption Trial is a translational study of safety-net primary care predominantly Hispanic/Latino patients with type 2 diabetes in collaboration with the Los Angeles County Department of Health Services. OBJECTIVES: To evaluate the cost-effectiveness of an information and communication technology (ICT)-facilitated depression care management program. METHODS: Cost-effectiveness of the ICT-facilitated care (TC) delivery model was evaluated relative to a usual care (UC) and a supported care (SC) model. TC added automated low-intensity periodic depression assessment calls to patients. Patient-reported outcomes included the 12-Item Short Form Health Survey converted into quality-adjusted life-years (QALYs) and the 9-Item Patient Health Questionnaire-calculated depression-free days (DFDs). Costs and outcomes data were collected over a 24-month period (-6 to 0 months baseline, 0 to 18 months study intervention). RESULTS: A sample of 1406 patients (484 in UC, 480 in SC, and 442 in TC) was enrolled in the nonrandomized trial. TC had a significant improvement in DFDs (17.3; P = 0.011) and significantly greater 12-Item Short Form Health Survey utility improvement (2.1%; P = 0.031) compared with UC. Medical costs were statistically significantly lower for TC (-$2328; P = 0.001) relative to UC but not significantly lower than for SC. TC had more than a 50% probability of being cost-effective relative to SC at willingness-to-pay thresholds of more than $50,000/QALY. CONCLUSIONS: An ICT-facilitated depression care (TC) delivery model improved QALYs, DFDs, and medical costs. It was cost-effective compared with SC and dominant compared with UC.

A Self-Managed System for Automated Assessment of UPDRS Upper Limb Tasks in Parkinson's Disease.

A home-based, reliable, objective and automated assessment of motor performance of patients affected by Parkinson's Disease (PD) is important in disease management, both to monitor therapy efficacy and to reduce costs and discomforts. In this context, we have developed a self-managed system for the automated assessment of the PD upper limb motor tasks as specified by the Unified Parkinson's Disease Rating Scale (UPDRS). The system is built around a Human Computer Interface (HCI) based on an optical RGB-Depth device and a replicable software. The HCI accuracy and reliability of the hand tracking compares favorably against consumer hand tracking devices as verified by an optoelectronic system as reference. The interface allows gestural interactions with visual feedback, providing a system management suitable for motor impaired users. The system software characterizes hand movements by kinematic parameters of their trajectories. The correlation between selected parameters and clinical UPDRS scores of patient performance is used to assess new task instances by a machine learning approach based on supervised classifiers. The classifiers have been trained by an experimental campaign on cohorts of PD patients. Experimental results show that automated assessments of the system replicate clinical ones, demonstrating its effectiveness in home monitoring of PD.

Automated assessment of joint synovitis activity from medical ultrasound and power doppler examinations using image processing and machine learning methods.

OBJECTIVES: Rheumatoid arthritis is the most common rheumatic disease with arthritis, and causes substantial functional disability in approximately 50% patients after 10 years. Accurate measurement of the disease activity is crucial to provide an adequate treatment and care to the patients. The aim of this study is focused on a computer aided diagnostic system that supports an assessment of synovitis severity. MATERIAL AND METHODS: This paper focus on a computer aided diagnostic system that was developed within joint Polish-Norwegian research project related to the automated assessment of the severity of synovitis. Semiquantitative ultrasound with power Doppler is a reliable and widely used method of assessing synovitis. Synovitis is estimated by ultrasound examiner using the scoring system graded from 0 to 3. Activity score is estimated on the basis of the examiner's experience or standardized ultrasound atlases. The method needs trained medical personnel and the result can be affected by a human error. RESULTS: The porotype of a computer-aided diagnostic system and algorithms essential for an analysis of ultrasonic images of finger joints are main scientific output of the MEDUSA project. Medusa Evaluation System prototype uses bone, skin, joint and synovitis area detectors for mutual structural model based evaluation of synovitis. Finally, several algorithms that support the semi-automatic or automatic detection of the bone region were prepared as well as a system that uses the statistical data processing approach in order to automatically localize the regions of interest. CONCLUSIONS: Semiquantitative ultrasound with power Doppler is a reliable and widely used method of assessing synovitis. Activity score is estimated on the basis of the examiner's experience and the result can be affected by a human error. In this paper we presented the MEDUSA project which is focused on a computer aided diagnostic system that supports an assessment of synovitis severity.

Computer-Aided Assessment of Repigmentation Rates in Vitiligo Patients: Implications for Treatment Efficacy - A Retrospective Study.

Precise evaluation of repigmentation in vitiligo patients is crucial for monitoring treatment efficacy and enhancing patient satisfaction. This study aimed to develop a computer-aided system for assessing repigmentation rates in vitiligo patients, providing valuable insights for clinical practice. A retrospective study was conducted at the Dermatology Department of Shenzhen People's Hospital between June 2019 and November 2022. Pre- and post-treatment images of vitiligo lesions under Wood's lamp were collected, involving 833 participants stratified by sex, age, and pigmentation patterns. Our results demonstrated that the marginal pigmentation pattern exhibited a higher repigmentation rate of 72% compared with the central non-follicular pattern at 45%. Males had a slightly higher average repigmentation rate of 0.37 in comparison to females at 0.33. Among age groups, individuals aged 0-20 years showed the highest average repigmentation rate at 0.41, while the oldest age group (61-80 years) displayed the lowest rate at 0.25. Analysis of multiple visits identified the marginal pattern as the most prevalent (60%), with a mean repigmentation rate of 40%. This study introduced a computational system for evaluating vitiligo repigmentation rates, enhancing our comprehension of patient responses, and ultimately contributing to enhanced clinical care.

Development and characterization of an automated behavioral assessment platform for the Göttingen minipig.

The Göttingen minipig is fast becoming the standard for assessing dermal chemical hazards because, like most swine, its skin is predictive of human skin response and because this strain's smaller size makes laboratory manipulations and husbandry easier. Unfortunately, standard behavioral tests and apparatus have not been developed for behavioral assessments of this swine strain. Indeed, computer-controlled automated behavioral testing procedures are much needed. The present research advanced this goal by producing a home-cage behavioral testing system that could accommodate minipigs of various sizes (ages). An aluminum frame housed three levers for recording operant responses, and LEDs above and below each lever served as discriminative stimuli. A commercially available food pellet dispenser was attached to a specialized pellet receptacle capable of measuring pellet retrieval. Two behavioral tests were selected and adapted from our commonly used non-human primate behavioral assessments: delayed match-to-sample (a memory test) and temporal response differentiation (a time-estimation test). Minipigs were capable of learning both tests and attaining stable performance. Next, scopolamine was used to validate the sensitivity of the behavioral tests for gauging behavioral perturbations in this swine strain. Scopolamine dose-effect functions were comparable to those observed in other species, including non-human primates, wherein 37.5 µg/kg of scopolamine (administered intramuscularly) reduced responding approximately 50%. Thus, we were successful in developing the apparatus and automated operant behavioral tests necessary to characterize drug safety in this swine strain. This capability will be valuable for characterizing chemical agent toxicity as well as the safety and efficacy of medical countermeasures.

A Fully Automated Mini-Mental State Examination Assessment Model Using Computer Algorithms for Cognitive Screening.

Background: Rapidly growing healthcare demand associated with global population aging has spurred the development of new digital tools for the assessment of cognitive performance in older adults. Objective: To develop a fully automated Mini-Mental State Examination (MMSE) assessment model and validate the model's rating consistency. Methods: The Automated Assessment Model for MMSE (AAM-MMSE) was an about 10-min computerized cognitive screening tool containing the same questions as the traditional paper-based Chinese MMSE. The validity of the AAM-MMSE was assessed in term of the consistency between the AAM-MMSE rating and physician rating. Results: A total of 427 participants were recruited for this study. The average age of these participants was 60.6 years old (ranging from 19 to 104 years old). According to the intraclass correlation coefficient (ICC), the interrater reliability between physicians and the AAM-MMSE for the full MMSE scale AAM-MMSE was high [ICC (2,1)=0.952; with its 95% CI of (0.883,0.974)]. According to the weighted kappa coefficients results the interrater agreement level for audio-related items showed high, but for items "Reading and obey", "Three-stage command", and "Writing complete sentence" were slight to fair. The AAM-MMSE rating accuracy was 87%. A Bland-Altman plot showed that the bias between the two total scores was 1.48 points with the upper and lower limits of agreement equal to 6.23 points and -3.26 points. Conclusions: Our work offers a promising fully automated MMSE assessment system for cognitive screening with pretty good accuracy.

A multimodal deep learning architecture for smoking detection with a small data approach.

Covert tobacco advertisements often raise regulatory measures. This paper presents that artificial intelligence, particularly deep learning, has great potential for detecting hidden advertising and allows unbiased, reproducible, and fair quantification of tobacco-related media content. We propose an integrated text and image processing model based on deep learning, generative methods, and human reinforcement, which can detect smoking cases in both textual and visual formats, even with little available training data. Our model can achieve 74% accuracy for images and 98% for text. Furthermore, our system integrates the possibility of expert intervention in the form of human reinforcement. Using the pre-trained multimodal, image, and text processing models available through deep learning makes it possible to detect smoking in different media even with few training data.

Three-Dimensional Mesh Recovery from Common 2-Dimensional Pictures for Automated Assessment of Body Posture in Camptocormia.

Background: Three-dimensional (3D) human body estimation from common photographs is an evolving method in the field of computer vision. It has not yet been evaluated on postural disorders. We generated 3D models from 2-dimensional pictures of camptocormia patients to measure the bending angle of the trunk according to recommendations in the literature. Methods: We used the Part Attention Regressor algorithm to generate 3D models from photographs of camptocormia patients' posture and validated the resulting angles against the gold standard. A total of 2 virtual human models with camptocormia were generated to evaluate the performance depending on the camera angle. Results: The bending angle assessment using the 3D mesh correlated highly with the gold standard (R = 0.97, P < 0.05) and is robust to deviations of the camera angle. Conclusions: The generation of 3D models offers a new method for assessing postural disorders. It is automated and robust to nonperfect pictures, and the result offers a comprehensive analysis beyond the bending angle.