Search bioRxiv⌕ Search
Home/Archive/bioRxiv

bioRxiv · 10.1101/421172

Can machine learning assess trunk alignment directly from raw video?

Sanchez, M. B.·Loram, I.·Cunningham, R.·Butler, P. B.
View original source ↗Export metadata ↓

Abstract

BackgroundThe Segmental Assessment of Trunk Control (SATCo) evaluates sitting control at seven separate trunk segments, making a judgement based on their position in space relative to a defined, aligned posture. SATCo is in regular clinical and research use and is a Recommended Instrument for Cerebral Palsy and Spinal Cord Injury-Paediatric by The National Institute of Neurological Disorders and Stroke (US). However, SATCo remains a subjective assessment.\n\nResearch questionThis study tests the feasibility of providing an objective, automated identification of frames containing the aligned, reference trunk posture using deep convolutional neural network (DCNN) analysis of raw high definition and depth (HD+D) images.\n\nMethodsA SATCo was conducted on sixteen healthy male adults and recorded using a Kinect V2. For each of seven segments tested, two different trials were collected (control and no-control) to simulate a range of alignment configurations. For all images, classification of alignment obtained from a trained and validated DCNN was compared to expert clinicians labelling.\n\nResultsUsing leave-one-out testing, at the optimal operating threshold, the DCNN correctly classified individual images (alignment v misaligned) with average precision 92.7{+/-}16% (mean{+/-}SD).\n\nSignificanceThese results show for the first time, automation of a key component of the SATCo test, namely identification of aligned trunk posture directly from raw images (HD+D). This demonstrates the potential of machine learning to provide a fully automated, objective SATCo test to enhance assessment of trunk control in children and adults for research and treatment of various conditions including neurodisability and stroke.

Explore related subjects

bioengineeringPreprint

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Sanchez, M. B., Loram, I., Cunningham, R., Butler, P. B.. 2018-09-19. Can machine learning assess trunk alignment directly from raw video?. https://doi.org/10.1101/421172

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

Preserving Nanotherapy-Induced Allogeneic Murine Islet Graft Tolerance During Distal Methicillin-Resistant Staphylococcus aureus Skin Infection

Eliminating lifelong immunosuppression can make islet transplantation a durable alternative to insulin therapy for type 1 diabetes. As short-course tolerogenic regimens are developed to replace lifelong immunosuppression, it is critical to test their robustness against perturbations such as hospital-acquired infections, which transplant recipients routinely encounter. Herein, we investigated whether rapamycin-loaded polymersomes (rPS), a subcutaneous nanotherapy that induces long-term murine islet allograft survival, can preserve host defenses in a Methicillin-resistant Staphylococcus aureus (MRSA) skin infection, and whether the resulting immune response from this distal infection relative to graft location erodes rPS-induced tolerance. rPS maintained neutrophil antibacterial activity despite inhibiting oxidative mechanisms in vitro, and rPS-treated mice demonstrated normal recovery following MRSA challenge. However, MRSA skin infection caused a decline of ~43% in recipient survival for rPS-treated mice, despite surviving recipients retaining glucose-stimulated insulin responsiveness. This erosion of allograft acceptance coincided with a burst of IL-6 within 24h of the infection followed by elevated TNF-[a] and neutrophilia at 13 days post-infection. Combining rPS with a single dose of methylprednisolone reduced IL-6 and TNF- without altering neutrophilia, nearly doubling survival among infected recipients to 80%. Our findings demonstrate that partial dampening of inflammatory insults can restore superior islet allograft acceptance following perioperative infection.

bioengineering↗
bioRxiv2026

Extensional Flow Microfluidics Reveal Size-Dependent Core-Shell Mechanical Heterogeneity in Pancreatic Tumor Spheroids

Understanding the mechanical behavior of cancer cell aggregates is important for investigating metastatic potential and developing effective therapeutics. Tumor cell clusters are major contributors to metastasis and can undergo extreme deformation to migrate through capillaries smaller than the cluster size, suggesting complex multicellular mechanics. Despite their clinical significance, the mechanics of multicellular spheroids remain underexplored. In this study, the mechanical behavior of spheroids formed from the pancreatic cancer cell line PANC-1 was investigated using a cross-slot microfluidic device. Suspended spheroids from two size ranges were stretched in non-contact extensional flow (small spheroids mostly 40 < D0 < 75 m, large spheroids mostly 90 < D0 < 145 m). Digital image correlation analyses of intra-spheroid strains revealed strong deformation differences for small and large spheroids while analyses based on 2D projected contours of the spheroid shape did not show trends. Smaller spheroids deformed relatively homogeneously and exhibited significant overall stretching. In contrast, larger spheroids displayed heterogeneous deformation characterized by a stiff central core and a deformable outer shell. To investigate the origin of these mechanical differences, cellular morphology and the expression of stiffness-associated proteins, including vimentin and E-cadherin, were assessed from confocal microscopy. Larger spheroids display significantly lower nuclear sphericity than small spheroids, indicating more cellular packing with increasing spheroid size. Additionally, vimentin expression was concentrated in the outer shell of large spheroids, whereas small spheroids exhibit vimentin expression throughout the aggregate. Thus individual cell morphology and cytoskeletal networks seemed to contribute to the core-shell biomechanical behavior of the large spheroids and lack thereof for small spheroids. Taken together, this study suggests that as spheroids grow, they transition from an irregular and disorganized cluster at low cell numbers to a coordinated and hierarchical structure at sufficient cell numbers that corresponds to diameters {gtrsim} 100 m.

bioengineering↗
bioRxiv2026

Material context and TGF-β blockade timing determine retained matrix in patient-derived bioprinted keratoconus stroma

Stromal repair requires newly formed matrix to persist while the contractile response to TGF-{beta} is controlled. We developed patient-derived bioprinted gelatin methacryloyl (GelMA) stroma to connect material-dependent cell responses with matrix retention after timed receptor blockade. In constructs from six keratoconus (KC) and six normal donors, KC cells retained disease-associated transcription while secreting procollagen I and lumican at normal-donor levels per unit DNA. Comparisons with two-dimensional culture and matched cast GelMA showed that the three-dimensional environment increased matrix induction and lowered resting contractile expression. Printing added a stromal identity response that persisted as retained lumican. Delayed blockade after a TGF-{beta}1 pulse preserved matrix transcription and late secretion. Eleven days after withdrawal, this finite course left more native collagen and lower contractile-gene expression than vehicle in 9/12 donors. Across the tested blockade regimens, at least one regimen achieved both outcomes at the final endpoint in 11/12 donors, including every KC donor. Culture format changed the preferred regimen. Choosing finite blockade by retained collagen instead of secretion changed the schedule in half the donors; the selected arms contained 22.2% more collagen. Retained collagen and contractile expression therefore provide joint endpoints for designing when TGF-{beta}-pathway courses, including losartan and TGF-{beta}3 treatment for keratoconic and scarred corneas, should start and stop.

bioengineering↗