Semi-supervised Machine Learning Approach Research Articles

Local field potential-based brain-machine interface to inhibit epileptic seizures by spinal cord electrical stimulation.

This study proposes a closed-loop brain-machine interface (BMI) based on spinal cord stimulation to inhibit epileptic seizures, applying a semi-supervised machine learning approach that learns from Local Field Potential (LFP) patterns acquired on the pre-ictal (preceding the seizure) condition.
 
Approach. LFP epochs from the hippocampus and motor cortex are band-pass filtered from 1 to 13 Hz, to obtain the time-frequency representation using the continuous Wavelet transform, and successively calculate the phase lock values (PLV). As a novelty, the Z-score-based PLV normalization using both modified k-means and Davies-Bouldin's measure for clustering is proposed here. Consequently, a generic seizure's detector is calibrated for detecting seizures on the normalized PLV, and enables the spinal cord stimulation for periods of 30 s in a closed-loop, while the BMI system detects seizure events. To calibrate the proposed BMI, a dataset with LFP signals recorded on five Wistar rats during basal state and epileptic crisis was used. The epileptic crisis was induced by injecting pentylenetetrazol (PTZ). Afterwards, two experiments without/with our BMI were carried out, inducing epileptic crisis by PTZ in Wistar rats. 

Main Results. Stronger seizure events of high LFP amplitudes and long time periods were observed in the rat, when the BMI system was not used. In contrast, short-time seizure events of relative low intensity were observed in the rat, using the proposed BMI. The proposed system detected on unseen data the synchronized seizure activity in the hippocampus and motor cortex, provided stimulation appropriately, and consequently decreased seizure symptoms. 

Significance. Low-frequency LFP signals from the hippocampus and motor cortex, and cord spinal stimulation can be used to develop accurate closed-loop BMIs for early epileptic seizures inhibition, as an alternative treatment.

Automated PIM Detection in Wireless Networks via Smart Data-Driven Techniques

As the number of base stations keeps increasing due to ever-growing wireless services and pervasive connectivity, the operational costs for network service providers rises proportionally. In addition, with the evolution of wireless technologies and the development of next generation networks such as 5G and 6G, network operators are inevitably expected to encounter more complex challenges. One such network issue is the Passive Intermodulation (PIM) problem that is observed in both 4G and 5G networks. It degrades the user experience and radio resource efficiency while leading to an operational overhead for detecting and mitigating it on the operator side. Although there is a significant body of work regarding PIM detection and cancellation methods, the majority of such studies depends on hardware solutions and manual investigation by network engineers, which is costly in terms of time and labor. In this paper, we propose two methods using unsupervised and semi-supervised Machine Learning (ML) approaches, namely a time-series-based anomaly detection technique and an autoencoder-based one, for identifying PIM problems in network sites. The proposed solutions utilize a set of Key Performance Indicator (KPI) data of base stations obtained from network management systems for a significantly long time interval and detect possible PIM problems without the need for a human in the loop. We measure and analyze the performance of our solutions, with the guidance of experienced network engineers, on our collected dataset.

Discovery and validation of a 10-gene predictive signature for response to adjuvant chemotherapy in stage II and III colon cancer

Identifying patients with stage II and III colon cancer who will benefit from 5-fluorouracil (5-FU)-based adjuvant chemotherapy is crucial for the advancement of personalized cancer therapy. We employ a semi-supervised machine learning approach to analyze a large dataset with 933 stage II and III colon cancer samples. Our analysis leverages gene regulatory networks to discover an 18-gene prognostic signature and to explore a 10-gene signature that potentially predicts chemotherapy benefits. The 10-gene signature demonstrates strong prognostic power and shows promising potential to predict chemotherapy benefits. We establish a robust clinical assay on the NanoString nCounter platform, validated in a retrospective formalin-fixed paraffin-embedded (FFPE) cohort, which represents an important step toward clinical application. Our study lays the groundwork for improving adjuvant chemotherapy and potentially expanding into immunotherapy decision-making in colon cancer. Future prospective studies are needed to validate and establish the clinical utility of the 10-gene signature in clinical settings.

Identifying tectonic settings of porphyry copper deposits using zircon trace elements – A semi-supervised machine learning method

Porphyry copper deposits (PCDs), as main sources of Cu resources worldwide, have attracted great attention for both scientific research and mining purposes in the past decades. The formation of PCDs is closely related to magmatic-hydrothermal system formed in both (continental/oceanic) arc and non-arc environments. Many low-dimensional geochemistry discriminants were developed for identifying the tectonic environments where ore-related magmas formed. However, long-term geological processes may obscure the geochemistry information recorded in rocks and minerals, which inevitably introduces uncertainties for identifying tectonic environments when using the low-dimensional discriminant diagrams. Here, we proposed a semi-supervised machine learning (ML) approach based on zircon chemistry data (spanning 16 elements) to identify the tectonic environments of PCDs. The results show that the semi-supervised models perform better in identifying the tectonic environments compared to the conventional low-dimensional discriminants. Moreover, as another benefit of big data analysis, ML model can clarify the systematic geochemistry differences of igneous zircons formed in different tectonic environments. The results suggest that (U+Th)/Yb and LREE/HREE are the top two important zircon chemical features for distinguishing PCDs formed in arc or non-arc environments. Specifically, zircons from non-arc show higher (U+Th)/Yb compared to that of continental arc, while zircons from island arc show lowest (U+Th)/Yb and LREE/HREE. Finally, the semi-supervised learning model was applied to the Dexing porphyry copper deposit to approach the longstanding debate regarding the underlying metallogenic setting. The results support the viewpoint that the Dexing porphyry copper deposit was formed in an intracontinental (non-arc) environment related to asthenosphere upwelling.

Identifying type II quasars at intermediate redshift with few-shot learning photometric classification

A sub-population of AGNs where the central engine is obscured are known as type II quasars (QSO2s). These luminous AGNs have a thick and dusty torus that obscures the accretion disc from our line of sight. Thus, their special orientation allows for detailed studies of the AGN-host co-evolution. Increasing the sample size of QSO2 sources in critical redshift ranges is crucial for understanding the interplay of AGN feedback, the AGN-host relationship, and the evolution of active galaxies. We aim to identify QSO2 candidates in the `redshift desert' using optical and infrared photometry. At this intermediate redshift range (i.e. $1 z 2$), most of the prominent optical emission lines in QSO2 sources (e.g. CIV$ OIII 4959, 5008$) fall either outside the wavelength range of the SDSS optical spectra or in particularly noisy wavelength ranges, making QSO2 identification challenging. Therefore, we adopted a semi-supervised machine learning approach to select candidates in the SDSS galaxy sample. Recent applications of machine learning in astronomy focus on problems involving large data sets, with small data sets often being overlooked. We developed a `few-shot' learning approach for the identification and classification of rare-object classes using limited training data (200 sources). The new AMELIA pipeline uses a transfer-learning based approach with decision trees, distance-based, and deep learning methods to build a classifier capable of identifying rare objects on the basis of an observational training data set. We validated the performance of AMELIA by addressing the problem of identifying QSO2s at $1 z 2$ using SDSS and WISE photometry, obtaining an F1-score above 0.8 in a supervised approach. We then used AMELIA to select new QSO2 candidates in the `redshift desert' and examined the nature of the candidates using SDSS spectra, when available. In particular, we identified a sub-population of NeV emitters at $z which are highly likely to contain obscured AGNs. We used X-ray and radio cross-matching to validate our classification and investigated the performance of photometric criteria from the literature showing that our candidates have an inherent dusty nature. Finally, we derived physical properties for our QSO2 sample using photoionisation models and verified the AGN classification using an SED fitting. Our results demonstrate the potential of few-shot learning applied to small data sets of rare objects, in particular QSO2s, and confirms that optical-IR information can be further explored to search for obscured AGNs. We present a new sample of candidates to be further studied and validated using multi-wavelength observations.

In search of prosociality in rodents: A scoping review.

Studying prosociality in rodents can provide insight into brain mechanisms potentially related to neurodevelopmental disorders known to impact social behaviors (e.g., autism spectrum disorder). While many studies have been published suggesting promising models, current knowledge remains scattered, including potential factors mediating prosocial behaviors in rodents. Prosocial behavior is characterized by an action done to benefit another or promote their well-being. The goal of this scoping review is to characterize current findings regarding prosocial paradigms in rodents, highlight current gaps in reporting, and identify factors shown to be important in mediating prosocial responses in rodents. Five databases were consulted in search of relevant studies published between 2000 and 2020 (APA PsycInfo, Embase, MEDLINE, Scopus, Web of Science). An update using a semi-supervised machine learning approach (ASReview) was then conducted to collect studies from 2021-2023. In total, 80 articles were included. Findings were the following: (1) Three categories of prosocial paradigm were extracted: cooperation, helping, and sharing tasks, (2) Rodents showed the ability to perform prosocial actions in all three categories, (3) Significant gaps in reported methodologies (e.g., failure to report animals' characteristics, housing conditions, and/or experimental protocol) and mediating factors (e.g., sex, strain, housing, food restriction) were found, and (4) Behaviors are determinant when investigating prosociality in rodents, however many studies omitted to include such analyses. Together these results inform future studies on the impact of mediating factors and the importance of behavioral analyses on the expression of prosocial behaviors in rodents.

Semi-supervised machine learning approach for reaction stoichiometry and kinetic model identification using spectral data from flow reactors

The semi-supervised machine learning approach is an integrated calibration-free modelling framework for identifying reaction systems from spectral data using minimal prior information and it is validated with experimental data obtained in a micro-reactor.

Making Systems Fail-Aware: A Semi-Supervised Machine Learning Approach for Identifying Failures by Learning the Correct Behavior of a System

Observing the interaction between a system, its environment, and its internal state is vital to detect failures during operation. Monitoring systems often use predefined system properties to detect such failures, and violations indicate potential failures. However, obtaining these properties is work-intensive and error-prone. Therefore, we describe an approach to obtain a system model by learning only the correct behavior using machine learning. Monitoring systems can use such models to predict correct future behavior. A potential failure is raised if real-world data deviate significantly from this prediction. We use a semi-supervised LSTM-based forecasting approach with a simple architecture, apply our approach to simulation data from a battery control system, and discuss our experimental results.

Semi-Supervised Machine Learning Approaches for Thyroid Disease Prediction and its Integration With the Internet of Everything

Semi-Supervised Machine Learning Approaches for Thyroid Disease Prediction and its Integration With the Internet of Everything

Semi-Supervised Machine Learning Approaches for DDOS Attack Detection

Network infrastructures are the target of several attacks. These include intrusions into the confidentiality, integrity, and availability of the network. The network's availability is impacted by a persistent attack known as a distributed denial-of-service (DDoS) attack. Such an assault is carried out using a command and control (C & C) technique. To detect these assaults, numerous researchers have put forth various machine learning-based solutions. In this paper, we are going to detect different DDoS attacks by various methods and evaluate their performance. This experiment made use of the KD99 dataset. The normal and assault samples were classified using the random forest technique. The classification of 99.76% of the samples was accurate. By strategically selecting clusters and incorporating the insights gained from the small labelled dataset, a portion of the unlabelled clusters can be assigned labels, effectively converting raw data into useful training examples. This enriched dataset is then used to train an improved classifier that can better generalize and adapt to the dynamic nature of DDoS attacks.

A Machine Learning Approach to Predicting Physical Activity Levels in Adolescents

The ongoing evolution of technology has had both positive and negative effects on modern society. On the positive side, it has significantly improved the ease with which various activities can be performed. However, it has also had a negative impact by reducing physical activity. This reduction in physical activity, in turn, increases the risk of chronic diseases that contribute to global mortality rates. This research aims to assess the effectiveness of machine learning in predicting the physical activity levels of adolescents. The study utilizes data from accelerometers, specifically the ActiGraph GT3X. The research methodology employs a semi-supervised machine learning approach, using the support vector machine and decision tree algorithms to make these predictions. The sample comprises 61 adolescents (males = 17, female = 44), including high school students and university students aged 18-21, from the West Java region. The results from the machine learning model using the decision tree algorithm indicated a model accuracy of 97.50% in predicting physical activity levels. In contrast, the accuracy obtained from the performance analysis using the confusion matrix for the support vector machine model was 92.5%. Based on these accuracy levels, the decision tree algorithm outperforms the support vector machine algorithm's accuracy. Further analyses involving different models are necessary to determine which algorithm offers the highest level of accuracy.

Comparing supervised and semi-supervised machine learning approaches in NTCP modeling to predict complications in head and neck cancer patients

Background and purposeHead and neck cancer (HNC) patients treated with radiotherapy often suffer from radiation-induced toxicities. Normal Tissue Complication Probability (NTCP) modeling can be used to determine the probability to develop these toxicities based on patient, tumor, treatment and dose characteristics. Since the currently used NTCP models are developed using supervised methods that discard unlabeled patient data, we assessed whether the addition of unlabeled patient data by using semi-supervised modeling would gain predictive performance. Materials and methodsThe semi-supervised method of self-training was compared to supervised regression methods with and without prior multiple imputation by chained equation (MICE). The models were developed for the most common toxicity outcomes in HNC patients, xerostomia (dry mouth) and dysphagia (difficulty swallowing), measured at six months after treatment, in a development cohort of 750 HNC patients. The models were externally validated in a validation cohort of 395 HNC patients. Model performance was assessed by discrimination and calibration. ResultsMICE and self-training did not improve performance in terms of discrimination or calibration at external validation compared to current regression models. In addition, the relative performance of the different models did not change upon a decrease in the amount of (labeled) data available for model development. Models using ridge regression outperformed the logistic models for the dysphagia outcome. ConclusionSince there was no apparent gain in the addition of unlabeled patient data by using the semi-supervised method of self-training or MICE, the supervised regression models would still be preferred in current NTCP modeling for HNC patients.

On Evaluating IoT Data Trust via Machine Learning

Data trust in IoT is crucial for safeguarding privacy, security, reliable decision-making, user acceptance, and complying with regulations. Various approaches based on supervised or unsupervised machine learning (ML) have recently been proposed for evaluating IoT data trust. However, assessing their real-world efficacy is hard mainly due to the lack of related publicly available datasets that can be used for benchmarking. Since obtaining such datasets is challenging, we propose a data synthesis method, called random walk infilling (RWI), to augment IoT time-series datasets by synthesizing untrustworthy data from existing trustworthy data. Thus, RWI enables us to create labeled datasets that can be used to develop and validate ML models for IoT data trust evaluation. We also extract new features from IoT time-series sensor data that effectively capture its autocorrelation as well as its cross-correlation with the data of the neighboring (peer) sensors. These features can be used to learn ML models for recognizing the trustworthiness of IoT sensor data. Equipped with our synthesized ground-truth-labeled datasets and informative correlation-based features, we conduct extensive experiments to critically examine various approaches to evaluating IoT data trust via ML. The results reveal that commonly used ML-based approaches to IoT data trust evaluation, which rely on unsupervised cluster analysis to assign trust labels to unlabeled data, perform poorly. This poor performance is due to the underlying assumption that clustering provides reliable labels for data trust, which is found to be untenable. The results also indicate that ML models, when trained on datasets augmented via RWI and using the proposed features, generalize well to unseen data and surpass existing related approaches. Moreover, we observe that a semi-supervised ML approach that requires only about 10% of the data labeled offers competitive performance while being practically more appealing compared to the fully supervised approaches. The related Python code and data are available online.

A machine learning-based score for precise echocardiographic assessment of cardiac remodelling in hypertensive young adults.

Accurate staging of hypertension-related cardiac changes, before the development of significant left ventricular hypertrophy, could help guide early prevention advice. We evaluated whether a novel semi-supervised machine learning approach could generate a clinically meaningful summary score of cardiac remodelling in hypertension. A contrastive trajectories inference approach was applied to data collected from three UK studies of young adults. Low-dimensional variance was identified in 66 echocardiography variables from participants with hypertension (systolic ≥160 mmHg) relative to a normotensive group (systolic < 120 mmHg) using a contrasted principal component analysis. A minimum spanning tree was constructed to derive a normalized score for each individual reflecting extent of cardiac remodelling between zero (health) and one (disease). Model stability and clinical interpretability were evaluated as well as modifiability in response to a 16-week exercise intervention. A total of 411 young adults (29 ± 6 years) were included in the analysis, and, after contrastive dimensionality reduction, 21 variables characterized >80% of data variance. Repeated scores for an individual in cross-validation were stable (root mean squared deviation = 0.1 ± 0.002) with good differentiation of normotensive and hypertensive individuals (area under the receiver operating characteristics 0.98). The derived score followed expected hypertension-related patterns in individual cardiac parameters at baseline and reduced after exercise, proportional to intervention compliance (P = 0.04) and improvement in ventilatory threshold (P = 0.01). A quantitative score that summarizes hypertension-related cardiac remodelling in young adults can be generated from a computational model. This score might allow more personalized early prevention advice, but further evaluation of clinical applicability is required.

Solo tourism: exploration and conceptualization – a semi-supervised machine learning approach

ABSTRACT This study aims at understanding and conceptualizing solo tourism. By using a semi-supervised machine learning approach and scrutinizing 27,208 solo tourist tweets. Based on implicit self-theory we capture solo consumers’ self-development and self-discovery. We provide an all-inclusive slow tourism conceptualization and show that the solo tourism framework is a three-stage journey of self-discovery. This study not only provides tourism scholars and providers with an evidence-based solo tourism conceptualization, but also with a marketing, psychological, and operation tool to manage this segment.

Automation of the Timed Up and Go Test Using a Doppler Radar System for Gait and Balance Analysis in Elderly People.

The timed up and go (TUG) test is a simple, valid, and reliable clinical tool that is widely used to assess mobility in elderly people. Several research studies have been conducted to automate the TUG test using wearable sensors or motion-tracking systems. Despite their promising results, the adopted technological systems present inconveniences in terms of acceptability and privacy protection. In this work, we propose to overcome these problems by using a Doppler radar system set into the backrest of a chair in order to automate the TUG test and extract additional information from its phases (i.e., transfer, walk, and turn). We intend to segment its phases and extract spatiotemporal gait parameters automatically. Our methodology is mainly based on a multiresolution analysis of radar signals. We proposed a segmentation technique based on the extraction of limbs oscillations signals through a semisupervised machine learning approach, on the one hand, and the application of the DARC algorithm on the other hand. Once the speed signals of torso and limbs oscillations were detected, we suggested estimating 14 gait parameters. All our approaches were validated by comparing outcomes to those obtained from a reference Vicon system. High correlation coefficients were obtained by comparing the speed signals of the torso (ρ=0.8), the speed signals of limbs oscillations (ρ=0.91), the initial and final indices of TUG phases (ρ=0.95), and the extracted parameters (percentage error < 4.8) obtained after radar signal processing to those obtained from the Vicon system.

Abstract 650: A machine learning model accurately predicts response to immune checkpoint inhibitors in colorectal cancer using the gut microbiome

Abstract Background: Immune checkpoint inhibitors (ICI) have shown a long-lasting response in many cancers such as colorectal cancer. However, only patients with microsatellite instability (MSI-High) or mismatch repair deficiency (dMMR) benefit from ICI. One of the many factors that can influence ICI-response is the gut microbiome, however its effect remains elusive. Here we aim to investigate the role of the gut microbiome in ICI-response using machine learning (ML). Methods: Clinical and genomic data of ICI-treated CRC patients were retrieved from the KEYNOTE-177 trial. Microbiome abundances from The Cancer Genome Atlas (TCGA) were downloaded from the cBioportal database. A label propagation semi-supervised ML approach was conducted to label patients’ responses to ICI treatment in the TCGA cohort. Tumor mutational burden (TMB), MSI status, treatment, histological subtype, TNM staging and prior chemotherapy were fitted into the semi-supervised model to predict ICI-response in the combination cohort. The normalized abundance of 1406 microbial signatures were regularized using a LASSO model to avoid multicollinearity. Signatures with a non-zero LASSO coefficients were fitted into a supervised ML Random Forest Classification model (RFC) to evaluate their prediction for ICI-response. Samples were split with 80:20 training-testing ratio, and model’s performance was evaluated on the testing set using mean bootstrap estimate, 10-fold cross-validation (CV), and area under curve (AUC). Results: A total of 538 CRC patients from the TCGA and 29 ICI-treat patients from the KEYNOTE-177 trial were fitted into the semi-supervised model to label ICI-response in the TCGA cohort. A total of 95 patients were labeled as responder (R) while 488 patients were labeled as non-responder (NR). Responders had significantly higher TMB levels than NR (mean TMB: 22.5 vs. 12.2, p-value &amp;lt;0.0001). Of the 95 R, 42 were MSS, and 53 patients were MSI-High. While 417 of the NR were MSS. A total of 25 microbial signatures were screened after LASSO for their ICI prediction and were fitted into the RFC model along with TMB and MSI status (Hereafter called RFC27). The RFC27 model performed on the testing set with mean bootstrap estimate of 0.88 and 95%CI: [0.83-0.94], 10-fold CV of 0.84 and AUC of 0.97. TMB and MSI showed highest feature contribution followed by Actinopolyspora and Bacteroides Eggerthia. The relative abundance of both microbiotas was significantly different between R and NR (mean: 0.99 vs. 0.47, 2.0 vs. 1.6; p-value &amp;lt;0.001 respectively). Conclusions: Our results show that the gut microbiome can accurately predict ICI-response and their higher abundance contributed to a better response. Previous studies also reported a pro-inflammatory response of Bacteroides in addition to their association with ICI-response in melanoma patients, thus providing a potential therapeutic marker for ICI. Citation Format: Ayah N. Al-Bzour. A machine learning model accurately predicts response to immune checkpoint inhibitors in colorectal cancer using the gut microbiome [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 650.

Automatic classification of construction safety reports using semi-supervised YAKE-Guided LDA approach

Literature on supervised Machine-Learning (ML) approaches for classifying text-based safety reports for the construction sector has been growing. Recent studies have emphasized the need to build ML approaches that balance high classification accuracy and performance on management criteria, such as resource intensiveness. However, despite being highly accurate, the extensively focused, supervised ML approaches may not perform well on management criteria as many factors contribute to their resource intensiveness. Alternatively, the potential for semi-supervised ML approaches to achieve balanced performance has rarely been explored in the construction safety literature. The current study contributes to the scarce knowledge on semi-supervised ML approaches by demonstrating the applicability of a state-of-the-art semi-supervised learning approach, i.e., Yet, Another Keyword Extractor (YAKE) integrated with Guided Latent Dirichlet Allocation (GLDA) for construction safety report classification. Construction-safety-specific knowledge is extracted as keywords through YAKE, relying on accessible literature with minimal manual intervention. Keywords from YAKE are then seeded in the GLDA model for the automatic classification of safety reports without requiring a large quantity of prelabeled datasets. The YAKE-GLDA classification performance (F1 score of 0.66) is superior to existing unsupervised methods for the benchmark data containing injury narratives from Occupational Health and Safety Administration (OSHA). The YAKE-GLDA approach is also applied to near-miss safety reports from a construction site. The study demonstrates a high degree of generality of the YAKE-GLDA approach through a moderately high F1 score of 0.86 for a few categories in the near-miss data. The current research demonstrates that, unlike the existing supervised approaches, the semi-supervised YAKE-GLDA approach can achieve a novel possibility of consistently achieving reasonably good classification performance across various construction-specific safety datasets yet being resource-efficient. Results from an objective comparative and sensitivity analysis contribute to much-required knowledge-contesting insights into the functioning and applicability of the YAKE-GLDA. The results from the current study will help construction organizations implement and optimize an efficient ML-based knowledge-mining strategy for domains beyond safety and across sites where the availability of a pre-labeled dataset is a significant limitation.

Leveraging Active Learning for Failure Mode Acquisition

Identifying failure modes is an important task to improve the design and reliability of a product and can also serve as a key input in sensor selection for predictive maintenance. Failure mode acquisition typically relies on experts or simulations which require significant computing resources. With the recent advances in Natural Language Processing (NLP), efforts have been made to automate this process. However, it is not only time consuming, but extremely challenging to obtain maintenance records that list failure modes. Unsupervised learning methods such as topic modeling, clustering, and community detection are promising approaches for automatic processing of maintenance records to identify failure modes. However, the nascent state of NLP tools combined with incompleteness and inaccuracies of typical maintenance records pose significant technical challenges. As a step towards addressing these challenges, this paper proposes a framework in which online active learning is used to identify failure modes from maintenance records. Active learning provides a semi-supervised machine learning approach, allowing for a human in the training stage of the model. The hypothesis of this paper is that the use of a human to annotate part of the data and train a machine learning model to annotate the rest is more efficient than training unsupervised learning models. Results demonstrate that the model is trained with annotating less than ten percent of the total available data. The framework is able to achieve ninety percent (90%) accuracy in the identification of failure modes in test cases with an F-1 score of 0.89. This paper also demonstrates the effectiveness of the proposed framework with both qualitative and quantitative measures.

Biphasic burst and sustained transdermal delivery in vivo using an AI-optimized 3D-printed MN patch

The objective of the present study was to fabricate microneedles for delivering lipophilic active ingredients (APIs) using digital light processing (DLP) printing technology and quality by design (QbD) supplemented by artificial intelligence (AI) algorithms. In the present study, dissolvable microneedle (MN) patches using ibuprofen (IBU) as a model drug were successfully fabricated with DLP printing technology at ∼ 750 μm height, ∼250 μm base diameter, and tip with radius of curvature (RoC) of ∼ 15 μm. MN patches were comprised of IBU, photoinitiator, Lithium phenyl (2,4,6-trimethylbenzoyl) phosphinate (LAP), polyethylene glycol dimethacrylate (PEGDAMA)550 and distilled water and were developed using the QbD optimization approach. Optimization of print fidelity and needle morphology were achieved using AI implementing a semi-supervised machine learning approach. Mechanical strength tests demonstrated that IBU MNs formed pores both on Parafilm M® and human cadaver skin. IBU-MNs consisting of 0.23 %w/v and 0.49 %w/v LAP with 10 %w/v water showed ∼ 2 mg/cm2 sustained drug permeation at 72 h in skin permeation experiments with flux of ∼ 40 μg/cm2/h. Pharmacokinetic studies in rats displayed biphasic rapid first-order absorption with sustained zero-order input of Ko = 150ug/hr, AUC0-48h = 62812.02 ± 11128.39 ng/ml*h, Tmax = 2.66 ± 1.12 h, and Cmax = 3717.43 ± 782.25 ng/ml (using 0.23 %w/v LAP IBU MN patch). An in vitro in vivo relation (IVIVR) was conducted identifying a polynomial relationship between patch release and fraction absorbed in vivo. This study demonstrates fabrication of dissolvable DLP-printed microneedle patches for lipophilic API delivery with biphasic rapid first-order and sustained zero-order release.