Subset Of Data Research Articles

Adapting Gaussian Mixture Model Training to Embedded/Edge Devices: A Low I/O, Deadline-Aware and Energy Efficient Design

This work focuses on the Gaussian Mixture Model (GMM), a machine learning model used for density estimation and cluster analysis in healthcare, networking, etc. The Expectation-Maximization (EM) algorithm is commonly used to train GMMs. One of the main challenges facing this algorithm when running on embedded systems is the crippling memory constraints. In fact, EM requires several scans of the dataset and we observed that when the dataset cannot fully reside in the main memory, its execution is dramatically slowed down by I/O movements. In this paper, we present an optimization of the EM algorithm for GMMs that reduces the number of I/O operations thanks to two main contributions: (1) A divide-and-conquer strategy that divides the dataset into chunks, learns the GMM separately on each chunk and combines the results incrementally. By doing so, we prevent data from being swapped several times during the learning process. (2) Restricting the training on a subset of data whose volume is inferred online using data properties while producing good accuracy. On average, our results show a 63% improvement in overall execution time with comparable accuracy. We also adapted GMM learning to run in a limited time budget while hitting a good trade-off between execution time and energy consumption. This solution succeeded in meeting the fixed deadline in 100% of the cases and in reducing the energy consumption by up to 68.77%.

The chained difference-in-differences

This paper studies the identification, estimation, and inference of long-term (binary) treatment effect parameters when balanced panel data is not available, or consists of only a subset of the available data. We develop a new estimator: the chained difference-in-differences, which leverages the overlapping structure of many unbalanced panel data sets. This approach consists in aggregating a collection of short-term treatment effects estimated on multiple incomplete panels. Our estimator accommodates (1) multiple time periods, (2) variation in treatment timing, (3) treatment effect heterogeneity, (4) general missing data patterns, and (5) sample selection on observables. We establish the asymptotic properties of the proposed estimator and discuss identification and efficiency gains in comparison to existing methods. Finally, we illustrate its relevance through (i) numerical simulations, and (ii) an application about the effects of an innovation policy in France.

Synthetic orbit uncertainty generation through regression analysis of historical Conjunction Data Messages

In the last decades, the Earth-orbiting population of both active and non-active objects has grown significantly, leading to a substantial increase in number of possible in-orbit collisions. It is therefore crucial to monitor the orbit of space resident objects to assess in advance the threat of risky conjunctions. Within this framework, the 18th Space Defense Squadron (SDS) is consistently updating the orbit of thousands of tracked objects by processing observations of the U.S. Space Surveillance Network (SSN). The determined orbital data is continuously maintained in the Special Perturbation (SP) catalogue and used by the 19th SDS to issue close approach warnings to satellite operators around the globe in the form of Conjunction Data Messages (CDM). The Flight Dynamics (FD) group of the German Space Operation Centre (GSOC) receives on regular basis a subset of the SP catalogue data along with CDMs associated to the fleet of its controlled satellites. The SP ephemerides are in fact provided without any covariance information preventing any computation of the Probability of Collision (Pc). In GSOC FD we are implementing a service to link a series of synthetic orbital error covariance matrices to a given SP ephemeris by statistically analyzing historical CDMs of past events. More than 30 GB of past conjunction data are processed to extract state vector, covariance matrix and object size parameter of already encountered secondary objects. The orbital errors of these last are subsequently categorized and divided into orbital classes to decouple the high correlation the covariance has with respect to solar flux, object dimension, altitude of perigee, eccentricity and orbit inclination. The classification aims at collecting similar CDMs regarding the aforementioned dependencies, and approximates the predicted 1-sigma position errors in the orbital frame by optimal curve-fitting techniques. By evaluation of the curve fitting coefficients of a requested orbit class a covariance matrix can be generated for any prediction time in upcoming CDM refinements and other analyses. The work discusses the limiting cases of the classification approach, bringing possible solutions to the scenario of empty classes. An in-depth characterization of the parameters that affect the orbital errors is in fact performed to individualize the neighboring class that provides the closest and most meaningful covariance timeline. Successively, the effect of using synthetic covariance in a conjunction risk assessment is also explored, adapting the problem on real operations. Lastly, the entire data processing pipeline and how the described service fits into the GSOC Flight Dynamics System (FDS) framework is described.

LC3AB protein-based gene (PbG) signatures as an indicator of relapse even after achieving pathological complete response.

e12552 Background: The application of tissue-based quantitative protein and morphological profiling of breast cancer tissues combined with RNA sequencing provides a flexible and generally applicable and scalable method to develop Protein-based Gene (PbG) signatures with significant prognostic value. Methods: A racially and subtype diverse breast cancer cohort (N=555) with available subset of RNA-seq data was used to develop gene expression-based signatures by stratifying patients according to protein expression (IHC) based on survival information and tumor lymphocyte infiltration (TILs) using survival-derived cutoff values. The predictive value of these signatures was then compared with other popular biomarkers scoring systems such as ROR.score, and OncotypeDx (Gene 21). Here we describe the predictive value of these signatures is evaluated by logistic regression modeling of publicly available gene expression data (N=996) from patients receiving neoadjuvant chemotherapy (anthracycline +/- taxane) with a final validation in an independent cohort (N=443) receiving similar therapy. Results: A model of PbG-derived signature from Kaiso, LC3A/B IHC and quantitative TIL-measurements together with ROR score, MAPK, PTEN, Immune and Gene21 signatures show significant prediction of distant metastasis-free survival in all patients: (DMFS, AUC= 0.737, 95% CI: 0.665- 0.819); and in non-pCR patients DMSF (AUC= 0.71, 95% CI: 0.633-0.773). When examining the role of all those markers among the patients who achieved the pCR and had relapse, the LC3A/B PbG-derived score was found to be highly predictive of those patients who would not achieve pCR and ultimately have recurrent disease. Conclusions: PbG-based signatures utilizing on LC3A/B protein values offer a novel and clinically relevant predictive biomarker of response to therapy and relapse in breast cancer patients.

Complementary medicine in patients with head and neck cancer

ObjectivesLiterature shows that up to 40 % of cancer patients use complementary medicine treatments. Most patients do not communicate the usage with their practitioner. There are no studies investigating the prevalence and extent of complementary medicine in head and neck cancer in Germany. Material and methodsPatients undergoing tumor aftercare in our department answered a questionnaire, that investigates subsets of demographic data, health behavior, cancerogenesis and complementary medicine. Data was evaluated in total and relative numbers. Also, Chi-squared test and Phi coefficient were performed to evaluate significant correlations. ResultsA total of 102 patients could be included in this study. In the analyzed collective 30 % of patients were interested in complementary medicine. Of those, 71 % have experienced complementary medicine treatments. There was a wide variety of modalities being used. None of the patients has talked with the doctors at tumor aftercare about this usage. Correlation of data showed that gender influenced i.a. health behavior and the usage of complementary medicine before cancer. ConclusionInvestigation revealed that a relevant number of head and neck cancer patients used complementary treatments, but no one talked about it during tumor aftercare. As a “blind spot” in anamnesis and diagnostic, evaluation of complementary medicine before tumor treatment and in aftercare are necessary to improve therapeutic outcome and to prevent potential negative influences like medication interactions.

Adaptive K values and training subsets selection for optimal K-NN performance on FPGA

This study introduces an Adaptive K-Nearest Neighbors methodology designed for FPGA platforms, offering substantial improvements over traditional K-Nearest Neighbors implementations. By integrating a dynamic classifier selection system, our approach enhances adaptability, enabling on-the-fly adjustments of K values and subsets of training data. This flexibility results in up to a 10.66% improvement in accuracy and significantly reduces latency, rendering our system up to 3.918 times more efficient than conventional K-Nearest Neighbors techniques. The methodology’s efficacy is validated through experiments across multiple datasets, demonstrating its potential in optimizing both classification accuracy and system efficiency. The adaptive approach’s ability to improve response times, along with its flexibility, positions it as an ideal solution for real-time applications and highlights the advantages of the adaptive K-Nearest Neighbors methodology in overcoming the constraints of hardware-accelerated machine learning.

Private Doctors' Perspective towards "Patient First" in TB Diagnostic Cascade, Hisar, India.

TB diagnosis has been simplified in India following advances in available diagnostic tools. This facilitates private doctors' "patient first" approach toward early diagnosis; however, costs remain high. India's NTEP established a TB diagnostic network, which is free for patients and incentivizes private doctors to participate. Drawing from this context led to the design and implementation of the One-Stop TB Diagnostic Solution model, which was conducted in the Hisar district, Haryana, allowing specimens from presumptive TB patients from private doctors to be collected and tested as per NTEPs diagnostic algorithm. A subset of data pertaining to private doctors was analyzed for the project period. Qualitative data were also collected by interviewing doctors using a snowball method to capture doctors' perception about the model. Out of 1159 specimens collected from 60 facilities, MTB was detected in 32% and rifampicin resistance was detected in 7% specimens. All specimens went through the diagnostic algorithm. Thirty doctors interviewed were satisfied with the services offered and were appreciative of the program that implements this "patient centric" model. Results from implementation indicate the need to strengthen private diagnostics through a certification process to ensure provision of quality TB diagnostic services.

Assessing the Performance of Logistic Regression in Heart Disease Detection through 5-Fold Cross-Validation

This study explores the effectiveness of Logistic Regression in predicting heart disease using a dataset derived from multiple international databases. Employing a 5-fold cross-validation method, the research aimed to evaluate the model's accuracy, precision, recall, and F1-score. Results indicated that Logistic Regression performs robustly, with accuracy ranging from 80% to 88.29%, and high recall rates, highlighting its potential as a valuable tool in medical diagnostics. Despite some variability in precision, which may lead to higher false positive rates, the model's high recall is crucial in clinical settings where missing a diagnosis can have dire consequences. The research confirmed the applicability of Logistic Regression to binary classification problems in healthcare, aligning with existing literature that supports its use in similar contexts. The study contributes to the field by demonstrating the model's consistency and reliability across diverse data subsets, reinforcing the potential for machine learning applications in healthcare diagnostics. Future research should focus on integrating Logistic Regression with other models to improve accuracy and testing the model on more current, varied datasets to enhance its generalizability and effectiveness in real-world settings.

Optimizing Large-Scale Educational Assessment with a "Divide-and-Conquer" Strategy: Fast and Efficient Distributed Bayesian Inference in IRT Models.

With the growing attention on large-scale educational testing and assessment, the ability to process substantial volumes of response data becomes crucial. Current estimation methods within item response theory (IRT), despite their high precision, often pose considerable computational burdens with large-scale data, leading to reduced computational speed. This study introduces a novel "divide- and-conquer" parallel algorithm built on the Wasserstein posterior approximation concept, aiming to enhance computational speed while maintaining accurate parameter estimation. This algorithm enables drawing parameters from segmented data subsets in parallel, followed by an amalgamation of these parameters via Wasserstein posterior approximation. Theoretical support for the algorithm is established through asymptotic optimality under certain regularity assumptions. Practical validation is demonstrated using real-world data from the Programme for International Student Assessment. Ultimately, this research proposes a transformative approach to managing educational big data, offering a scalable, efficient, and precise alternative that promises to redefine traditional practices in educational assessments.

Noise filtering options for conically scanning Doppler lidar measurements with low pulse accumulation

Abstract. Doppler lidar (DL) applications with a focus on turbulence measurements sometimes require measurement settings with a relatively small number of accumulated pulses per ray in order to achieve high sampling rates. Low pulse accumulation comes at the cost of the quality of DL radial velocity estimates and increases the probability of outliers, also referred to as “bad” estimates or noise. Careful filtering is therefore the first important step in a data processing chain that begins with radial velocity measurements as DL output variables and ends with turbulence variables as the target variable after applying an appropriate retrieval method. It is shown that commonly applied filtering techniques have weaknesses in distinguishing between “good” and “bad” estimates with the sensitivity needed for a turbulence retrieval. For that reason, new ways of noise filtering have been explored, taking into account that the DL background noise can differ from generally assumed white noise. It is shown that the introduction of a new coordinate frame for a graphical representation of DL radial velocities from conical scans offers a different perspective on the data when compared to the well-known velocity–azimuth display (VAD) and thus opens up new possibilities for data analysis and filtering. This new way of displaying DL radial velocities builds on the use of a phase-space perspective. Following the mathematical formalism used to explain a harmonic oscillator, the VAD’s sinusoidal representation of the DL radial velocities is transformed into a circular arrangement. Using this kind of representation of DL measurements, bad estimates can be identified in two different ways: either in a direct way by singular point detection in subsets of radial velocity data grouped in circular rings or indirectly by localizing circular rings with mostly good radial velocity estimates by means of the autocorrelation function. The improved performance of the new filter techniques compared to conventional approaches is demonstrated through both a direct comparison of unfiltered with filtered datasets and a comparison of retrieved turbulence variables with independent measurements.

More is better? The role of strategic data management in a lean manufacturing process

Purpose Lean manufacturing has been pivotal in emphasizing the reduction of cycle times, minimizing manufacturing costs and diminishing inventories. This research endeavors to formulate a lean data management paradigm, through the design and execution of a strategic edge-cloud data governance approach. This study aims to discern anomalous or unforeseen patterns within data sets, enabling an efficacious examination of product shortcomings within manufacturing processes, while concurrently minimizing the redundancy associated with the storage, access and processing of nonvalue-adding data. Design/methodology/approach Adopting a lean data management framework within both edge and cloud computing contexts, this study ensures the preservation of significant time series sequences, while ascertaining the optimal quantity of normal time series data to retain. The efficacy of detected anomalous patterns, both at the edge and in the cloud, is assessed. A comparative analysis between traditional data management practices and the strategic edge-cloud data governance approach facilitates an exploration into the equilibrium between anomaly detection and space conservation in cloud environments for aggregated data analysis. Findings Evaluation of the proposed framework through a real-world inspection case study revealed its capability to navigate alternative strategies for harmonizing anomaly detection with data storage efficiency in cloud-based analysis. Contrary to the conventional belief that retaining comprehensive data in the cloud maximizes anomaly detection rates, our findings suggest that a strategic edge-cloud data governance model, which retains a specific subset of normal data, can achieve comparable or superior accuracy with less normal data relative to traditional methods. This approach further demonstrates enhanced space efficiency and mitigates various forms of waste, including temporal delays, storage of noncontributory normal data, costs associated with the analysis of such data and excess data transmission. Originality/value By treating inspected normal data as nonvalue-added, this study probes the intricacies of maintaining an optimal balance of such data in light of anomaly detection performance from aggregated data sets. Our proposed methodology augments existing research by integrating a strategic edge-cloud data governance model within a lean data analytical framework, thereby ensuring the retention of adequate data for effective anomaly detection.

ICEED: Integrated customized extraction of enzyme data.

Enzymes catalyze diverse biochemical reactions and are building blocks of cellular and metabolic pathways. Data and metadata of enzymes are distributed across databases and are archived in various formats. The enzyme databases provide utilities for efficient searches and downloading enzyme records in batch mode but do not support organism-specific extraction of subsets of data. Users are required to write scripts for parsing entries for customized data extraction prior to downstream analysis. Integrated Customized Extraction of Enzyme Data (iCEED) has been developed to provide organism-specific customized data extraction utilities for seven commonly used enzyme databases and brings these resources under an integrated portal. iCEED provides dropdown menus and search boxes using typehead utility for submission of queries as well as enzyme class-based browsing utility. A utility to facilitate mapping and visualization of functionally important features on the three-dimensional (3D) structures of enzymes is integrated. The customized data extraction utilities provided in iCEED are expected to be useful for biochemists, biotechnologists, computational biologists, and life science researchers to build curated datasets of their choice through an easy to navigate web-based interface. The integrated feature visualization system is useful for a fine-grained understanding of the enzyme structure-function relationship. Desired subsets of data, extracted and curated using iCEED can be subsequently used for downstream processing, analyses, and knowledge discovery. iCEED can also be used for training and teaching purposes.

Testing the generalizability and effectiveness of deep learning models among clinics: sperm detection as a pilot study

BackgroundDeep learning has been increasingly investigated for assisting clinical in vitro fertilization (IVF). The first technical step in many tasks is to visually detect and locate sperm, oocytes, and embryos in images. For clinical deployment of such deep learning models, different clinics use different image acquisition hardware and different sample preprocessing protocols, raising the concern over whether the reported accuracy of a deep learning model by one clinic could be reproduced in another clinic. Here we aim to investigate the effect of each imaging factor on the generalizability of object detection models, using sperm analysis as a pilot example.MethodsAblation studies were performed using state-of-the-art models for detecting human sperm to quantitatively assess how model precision (false-positive detection) and recall (missed detection) were affected by imaging magnification, imaging mode, and sample preprocessing protocols. The results led to the hypothesis that the richness of image acquisition conditions in a training dataset deterministically affects model generalizability. The hypothesis was tested by first enriching the training dataset with a wide range of imaging conditions, then validated through internal blind tests on new samples and external multi-center clinical validations.ResultsAblation experiments revealed that removing subsets of data from the training dataset significantly reduced model precision. Removing raw sample images from the training dataset caused the largest drop in model precision, whereas removing 20x images caused the largest drop in model recall. by incorporating different imaging and sample preprocessing conditions into a rich training dataset, the model achieved an intraclass correlation coefficient (ICC) of 0.97 (95% CI: 0.94-0.99) for precision, and an ICC of 0.97 (95% CI: 0.93-0.99) for recall. Multi-center clinical validation showed no significant differences in model precision or recall across different clinics and applications.ConclusionsThe results validated the hypothesis that the richness of data in the training dataset is a key factor impacting model generalizability. These findings highlight the importance of diversity in a training dataset for model evaluation and suggest that future deep learning models in andrology and reproductive medicine should incorporate comprehensive feature sets for enhanced generalizability across clinics.

Two-Stage Training Framework Using Multicontrast MRI Radiomics for IDH Mutation Status Prediction in Glioma.

Purpose To develop a radiomics framework for preoperative MRI-based prediction of isocitrate dehydrogenase (IDH) mutation status, a crucial glioma prognostic indicator. Materials and Methods Radiomics features (shape, first-order statistics, and texture) were extracted from the whole tumor or the combination of nonenhancing, necrosis, and edema regions. Segmentation masks were obtained via the federated tumor segmentation tool or the original data source. Boruta, a wrapper-based feature selection algorithm, identified relevant features. Addressing the imbalance between mutated and wild-type cases, multiple prediction models were trained on balanced data subsets using random forest or XGBoost and assembled to build the final classifier. The framework was evaluated using retrospective MRI scans from three public datasets (The Cancer Imaging Archive [TCIA, 227 patients], the University of California San Francisco Preoperative Diffuse Glioma MRI dataset [UCSF, 495 patients], and the Erasmus Glioma Database [EGD, 456 patients]) and internal datasets collected from the University of Texas Southwestern Medical Center (UTSW, 356 patients), New York University (NYU, 136 patients), and University of Wisconsin-Madison (UWM, 174 patients). TCIA and UTSW served as separate training sets, while the remaining data constituted the test set (1617 or 1488 testing cases, respectively). Results The best performing models trained on the TCIA dataset achieved area under the receiver operating characteristic curve (AUC) values of 0.89 for UTSW, 0.86 for NYU, 0.93 for UWM, 0.94 for UCSF, and 0.88 for EGD test sets. The best performing models trained on the UTSW dataset achieved slightly higher AUCs: 0.92 for TCIA, 0.88 for NYU, 0.96 for UWM, 0.93 for UCSF, and 0.90 for EGD. Conclusion This MRI radiomics-based framework shows promise for accurate preoperative prediction of IDH mutation status in patients with glioma. Keywords: Glioma, Isocitrate Dehydrogenase Mutation, IDH Mutation, Radiomics, MRI Supplemental material is available for this article. Published under a CC BY 4.0 license. See also commentary by Moassefi and Erickson in this issue.

Optimal processing of surface facial EMG to identify emotional expressions: A data-driven approach

Surface facial electromyography (EMG) is commonly used to detect emotions from subtle facial expressions. Although there are established procedures for collecting EMG data and some aspects of their processing, there is little agreement among researchers about the optimal way to process the EMG signal, so that the study-unrelated variability (noise) is removed, and the emotion-related variability is best detected. The aim of the current paper was to establish an optimal processing pipeline for EMG data for identifying emotional expressions in facial muscles. We identified the most common processing steps from existing literature and created 72 processing pipelines that represented all the different processing choices. We applied these pipelines to a previously published dataset from a facial mimicry experiment, where 100 adult participants observed happy and sad facial expressions, whilst the activity of their facial muscles, zygomaticus major and corrugator supercilii, was recorded with EMG. We used a resampling approach and subsets of the original data to investigate the effect and robustness of different processing choices on the performance of a logistic regression model that predicted the mimicked emotion (happy/sad) from the EMG signal. In addition, we used a random forest model to identify the most important processing steps for the sensitivity of the logistic regression model. Three processing steps were found to be most impactful: baseline correction, standardisation within muscles, and standardisation within subjects. The chosen feature of interest and the signal averaging had little influence on the sensitivity to the effect. We recommend an optimal processing pipeline, share our code and data, and provide a step-by-step walkthrough for researchers.

NaTech triggered by lightning: Novel insights from past events in the process industry

Lightning strikes, a prominent meteorological event, pose a significant risk of triggering technological disruptions within the process industry. To better understand this phenomenon, an analysis focused on past lightning-triggered events was carried out, examining open-source industrial-accident databases to compile a new NaTech-driven dataset of 689 records. First, an overall quantitative analysis revealed that over 80 % of these events involved incidents or loss of containment. Notably, 83.3 % of them occurred during the spring and summer, indicating a seasonal pattern. Based on the frequency of functional attributes, the chemical and petrochemical macro-sector was the most vulnerable, followed by storage and warehousing. About 40 % of all classifiable events happened on storage equipment, while 21 % happened on electric and electronic devices. Given the lack of valuable information for the principal source of data (NRC), the technological scenarios triggered were characterized using a refined subset of 127 observations, obtained considering the “other sources” of data. Fire scenarios predominated at 56 %; coincidentally, roughly 70 % of all scenarios involved hazardous substances classified as physical hazards. Estimated losses for the available information underscored the adverse consequences of lightning-triggered NaTech events, highlighting their major impact on both safety and the environment. An analysis of the event tree showed the logical path from the lightning strike to the final ignition scenarios (considering a subset of 107 records). This path accounted for 36 % of the classifiable records that directly affected the structure, while more than 50 % of them did not. Bayesian network structures made it possible to get conditional probabilities from the event tree and improved the model by adding attributes for vulnerable equipment and macro-sectors. In order to deal with the uncertain data, algorithms were used to generalize the models that were obtained from smaller subsets of data with more accurate information to the whole dataset. It provides an important additional view of unclassifiable data that otherwise remained in the dark. This novel insight contributes to increase the vulnerability awareness of industrial assets against lightning strikes.

Multi-view fusion for instruction mining of large language model

Large Language Models (LLMs) obtain their instruction-following ability through instruction tuning. While the quality of instruction data is considered critical for a successful LLM, the selection of high-quality datasets for finetuning still lacks clear guidelines and quantitative analyses. In this work, we introduce three analytical views for instruction mining: diversity, complexity, and accuracy, which can aid in selecting an optimal subset of instruction data for fine-tuning. Based on these views, we propose a multi-view fusion framework for efficient instruction selection, including diversity sampling based on LoRA representation distribution, complexity scoring based on uncertainty quantification, and accuracy scoring based on reward modeling. We perform the framework on various open-sourced instruction datasets, and achieved enhanced performance of LLMs with a carefully curated subset, underscoring the effectiveness of our proposed framework.

Guided Docking as a Data Generation Approach Facilitates Structure-Based Machine Learning on Kinases.

Drug discovery pipelines nowadays rely on machine learning models to explore and evaluate large chemical spaces. While including 3D structural information is considered beneficial, structural models are hindered by the availability of protein-ligand complex structures. Exemplified for kinase drug discovery, we address this issue by generating kinase-ligand complex data using template docking for the kinase compound subset of available ChEMBL assay data. To evaluate the benefit of the created complex data, we use it to train a structure-based E(3)-invariant graph neural network. Our evaluation shows that binding affinities can be predicted with significantly higher precision by models that take synthetic binding poses into account compared to ligand- or drug-target interaction models alone.

Selecting subsets of source data for transfer learning with applications in metal additive manufacturing

Selecting subsets of source data for transfer learning with applications in metal additive manufacturing

BACK-to-MOVE: Machine learning and computer vision model automating clinical classification of non-specific low back pain for personalised management.

Low back pain (LBP) is a major global disability contributor with profound health and socio-economic implications. The predominant form is non-specific LBP (NSLBP), lacking treatable pathology. Active physical interventions tailored to individual needs and capabilities are crucial for its management. However, the intricate nature of NSLBP and complexity of clinical classification systems necessitating extensive clinical training, hinder customised treatment access. Recent advancements in machine learning and computer vision demonstrate promise in characterising NSLBP altered movement patters through wearable sensors and optical motion capture. This study aimed to develop and evaluate a machine learning model (i.e., 'BACK-to-MOVE') for NSLBP classification trained with expert clinical classification, spinal motion data from a standard video alongside patient-reported outcome measures (PROMs). Synchronised video and three-dimensional (3D) motion data was collected during forward spinal flexion from 83 NSLBP patients. Two physiotherapists independently classified them as motor control impairment (MCI) or movement impairment (MI), with conflicts resolved by a third expert. The Convolutional Neural Networks (CNNs) architecture, HigherHRNet, was chosen for effective pose estimation from video data. The model was validated against 3D motion data (subset of 62) and trained on the freely available MS-COCO dataset for feature extraction. The Back-to-Move classifier underwent fine-tuning through feed-forward neural networks using labelled examples from the training dataset. Evaluation utilised 5-fold cross-validation to assess accuracy, specificity, sensitivity, and F1 measure. Pose estimation's Mean Square Error of 0.35 degrees against 3D motion data demonstrated strong criterion validity. Back-to-Move proficiently differentiated MI and MCI classes, yielding 93.98% accuracy, 96.49% sensitivity (MI detection), 88.46% specificity (MCI detection), and an F1 measure of .957. Incorporating PROMs curtailed classifier performance (accuracy: 68.67%, sensitivity: 91.23%, specificity: 18.52%, F1: .800). This study is the first to demonstrate automated clinical classification of NSLBP using computer vision and machine learning with standard video data, achieving accuracy comparable to expert consensus. Automated classification of NSLBP based on altered movement patters video-recorded during routine clinical examination could expedite personalised NSLBP rehabilitation management, circumventing existing healthcare constraints. This advancement holds significant promise for patients and healthcare services alike.