Biology subjects

Tozzi, G.

Publications and source records attributed to Tozzi, G..

12 recordsLinked to original sources

bioRxiv2026

BoneGraph: A Domain-Specialised, Self-Correcting Reasoning System for Bone Science Retrieval, Grounded Inference, and Image Mechanics

Bone science literature spans biology, mechanics, materials science, and clinical medicine, and its volume makes reliable knowledge synthesis increasingly difficult. General-purpose large language models (LLMs) answer fluently but under-represent this niche domain, cannot cite specific evidence, and offer no mechanism to be corrected durably. Here we present BoneGraph, a domain-specialised system for bone science delivered as a five-tab web application over a shared substrate: a curated full-text corpus of 7,449 documents embedded into 248,629 passage vectors using SPECTER2, a scientific-paper embedding model, and a bone knowledge graph of 1,597 concepts with 1,699 causal relations. The five tabs are: (I) Chat, retrieval-augmented question answering with server-rebuilt inline citations; (II) Search, raw semantic retrieval with no LLM in the loop; (III) Reasoning, a self-correcting loop in which a deterministic physics check and a literature/knowledge-graph critic constrain the answer, and a user's feedback becomes a durable, per-user rule; (IV) Vision, a bone-region classifier trained on frozen BiomedCLIP features that grounds a vision-language model, guarded against out-of-distribution inputs and augmented with image-embedding correction memory; and (V) Mechanics, integrating our previous data-driven image mechanics (D2IM) model that predicts displacement and strain fields from a single undeformed micro-CT image. All inference is performed locally, without third-party API calls, and the public beta is served at bonegraph.org. Retrieval attains a mean reciprocal rank (MRR) of 0.928 on a 30-question, seven-domain benchmark, and the Vision classifier attains 92.6% accuracy on the held-out MURA (MUsculoskeletal RAdiographs) dataset. A grounded-reasoning benchmark shows that, with the correct passage, BoneGraph raises answer accuracy from 42% to 78%. BoneGraph makes a major contribution to bone-science informatics: to our knowledge it is the first domain-specialised system to unify curated retrieval, deterministic physics-grounded self-correction, and durable per-user learning for bone science.

bioengineering↗

bioRxiv2026

Self-Calibrated Hyperspectral Neural Radiance Fields for 3D Reconstruction of Bone and Bone Analogues

Hyperspectral imaging (HSI) has gained increasing attention for bone assessment because it captures rich wavelength dependent information associated with mineralised tissue. HSI provides detailed spectral information related to material composition, while 3D geometric information supports the analysis of surface morphology and structural detail. However, integrating spectral and geometric information remains challenging, particularly when conventional reconstruction pipelines depend on external pose estimation. To address this challenge, we propose BoNeRF-HS, a self-calibrated hyperspectral neural radiance field for 3D reconstruction. BoNeRF-HS jointly optimises camera intrinsics, volume density, and hyperspectral radiance, removing the need for COLMAP based poses. To improve spectral modelling, we incorporate a gated spectral adapter head that learns wavelength dependent radiance features for hyperspectral view synthesis. We evaluate BoNeRF-HS on a multi-view hyperspectral dataset containing mouse bone, trabecular bone analogue, and cortical bone analogue samples. Experimental results demonstrate that our framework achieves improved reconstruction quality, and better preservation of bone surface details compared with existing approaches.

bioengineering↗

bioRxiv2026

Quantum kernel support vector machines for trabecular bone classification: comparing feature reduction strategies on synthetic micro-CT data

Quantum kernel methods offer a potential advantage for classification tasks in high-dimensional feature spaces, yet their practical benefit critically depends on how input features are prepared. We compare five dimensionality reduction strategies--principal component analysis (PCA), Gaussian random projection (RP Gaussian), sparse random projection (RP Sparse), partial least squares (PLS), and uniform manifold approximation and projection (UMAP) -- as pre-processing steps for quantum kernel support vector machines (SVMs) applied to trabecular bone classification from synthetic micro-computed tomography (micro-CT) data. Using a custom procedural generator based on Gaussian random field zero-crossings, we produced 500 synthetic trabecular bone volumes with controlled morphometric properties such as bone volume fraction (BV/TV), trabecular thickness (Tb.Th), number (Tb.N) and spacing (Tb.Sp). Texture features extracted from grayscale slices are reduced to 8-dimensional quantum circuit inputs via each method, then classified using both classical radial basis function (RBF)-SVMs and quantum kernel SVMs with ZZ feature maps on a statevector simulator, both evaluated with 5 x 5 repeated stratified cross-validation (25 folds). Our results show that UMAP is the only reduction method where the quantum kernel remains competitive with the classical baseline. Under repeated cross-validation, UMAP showed a +0.032 accuracy gap favouring the quantum kernel (Dietterich 5 x 2 CV p = 0.177); however, validation on 10 fully independent datasets--each with independently generated samples, separate reduction fits, and separate kernel matrices -- reversed the sign to -0.030 (paired t-test p = 0.123; Wilcoxon p = 0.193; quantum wins 3/10 datasets), indicating that the apparent advantage was likely inflated by fold dependence. Nevertheless, UMAPs gap remains small and non-significant in both analyses, whereas all linear methods (PCA, RP Gaussian, PLS) show substantial quantum deficits of -0.090 to -0.116 across BV/TV classification, with PCA and PLS remaining significant under corrected tests (5 x 2 CV p = 0.004 and p = 0.007 respectively). We additionally evaluate quantum kernel ridge regression for continuous morphometric prediction, finding that ZZ quantum kernels fail uniformly at regression (negative R2 for all methods except PLS at 4 qubits), suggesting that the ZZ kernel captures decision boundaries but not smooth metric structure. These findings provide practical guidance for feature engineering in near-term quantum machine learning pipelines and demonstrate that the choice of dimensionality reduction can determine whether quantum kernels remain competitive with classical baselines.

biophysics↗

bioRxiv2026

Evaluation of direct strain field prediction in bone with data-driven image mechanics (D2IM-Strain)

Digital volume correlation (DVC) has become the benchmark experimental technique for full-field strain measurement in bone mechanics. In our previous work we developed a novel data-driven image mechanics (D2IM) approach that learns from DVC data and predicts displacement fields directly from undeformed X-ray computed tomography (XCT) images, deriving strain fields from such predictions. However, strain fields derived through numerical differentiation of displacement fields amplify high-frequency noise, and regularization techniques compromise spatial resolution while incurring substantial computational costs. Here we propose the upgrade D2IM-Strain to predict strain fields directly from XCT images of bone. Two prediction strategies were compared: displacement-derived strain and direct strain prediction. The direct strain prediction model significantly improved accuracy particularly for strain magnitudes below 10000{varepsilon}, taken as a representative threshold value for bone tissue yielding in compression. In addition, the direct approach reduced false-positive high-strain classifications by 75%. By eliminating numerical differentiation, the approach reduces noise amplification while maintaining computational efficiency. These findings represent a critical step toward developing robust data-driven volume correlation methods for hierarchical materials.

bioengineering↗

bioRxiv2025

Detecting Temporal Changes in Skin Surrogate Model Using Hyperspectral Imaging

Advancements in biomedical imaging have increasingly explored innovative models for studying human skin conditions. Recent research highlights a biochemical connection between banana peels and human skin, particularly through the enzyme tyrosinase, which plays a key role in pigmentation and oxidative stress. Studies have shown that, much like human skin, banana peels undergo oxidative browning due to tyrosinase activity. Leveraging this similarity, we have developed imaging techniques to track biochemical changes in bananas as a potential surrogate model for the detection of skin changes such as sun spots or melanoma. This paper introduces the application of hyperspectral imaging (HSI) to analyse the spectral properties of banana peels as a method for studying pigmentation disorders and oxidative damage. We present a HSI system designed to capture spectral reflectance characteristics across the 440 nm to 900 nm range, with a hyperspectral cube resolution of 956x952x96. Our approach enables a more targeted spectral assessment by dynamically selecting and investigating specific regions in banana samples over multiple days. Preliminary spectral analysis demonstrates the systems capability to detect reflectance variations linked to biochemical changes such as ripeness, bruising, and oxidative stress--factors that parallel pigmentation alterations in human skin. However, direct hyperspectral imaging can be resource intensive, making it less accessible for applications in dermatology and mobile health diagnostics. To address this, we conducted preliminary experiments to reconstruct 27-band hyperspectral representations in the VIS spectrum from RGB images. The observed spectral variations and successful reconstruction results confirm the potential of HSI for capturing biochemical changes and provide a foundation for future biomedical imaging research, particularly in melanoma detection and dermatology-related applications.

bioengineering↗

bioRxiv2024

RELATIONSHIP BETWEEN ACTUAL STRESS AND IN SITU MORPHOLOGICAL CHANGES IN THE MECHANICAL BEHAVIOR OF BIOMIMETIC POROUS HIERARCHICAL SCAFFOLDS

This study investigates the evolution with strain of the material volume fraction (i.e., porosity) and geometry in porous scaffolds to obtain a more accurate description of their stress-strain behavior. Single bundles and hierarchical structures (8 bundles enveloped by a membrane) were produced by electrospinning as tendon/ligament scaffolds. They underwent a micro-tomography in situ tensile test. Apparent and net stress were obtained using the initial sample cross-section and material volume fraction to normalize axial force. Micro-tomography revealed sample morphology change with strain to calculate the actual stress-strain. Moreover, nanofibers arrangement was revealed by scanning electron microscopy on both bundles and membranes. The description of the mechanical response significantly changed using evolving morphometry (actual stress-strain) instead of initial static one (apparent stress-strain), for both single bundle and hierarchical structure. The actual elastic modulus of the single bundles (583{+/-}97 MPa) was statistically higher than that of the hierarchical structures (163{+/-}107 MPa). This is related to the membrane, membrane-bundle and inter-bundle interactions. In the hierarchical structure, portions of the material resisting traction are constituted by nanofibers not aligned with the load. The different definitions for the stress-strain behavior allow different accuracy levels depending on the experimental complexity. The evolution of morphology with deformation can significantly affect the description of the mechanical response of porous scaffolds. This has a double impact in practical applications: at the body scale, it allows a better comparison between the scaffold behavior and the target tissue; at the cellular scale, it predicts the actual substrate stiffness that cells will face. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=104 SRC="FIGDIR/small/630537v1_ufig1.gif" ALT="Figure 1"> View larger version (24K): org.highwire.dtl.DTLVardef@5d5337org.highwire.dtl.DTLVardef@1027047org.highwire.dtl.DTLVardef@11966a9org.highwire.dtl.DTLVardef@a2e4e3_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioengineering↗

bioRxiv2024

Revealing the Auxetic Behavior of Biomimetic Multi-material and Region-specific Nanofibrous Scaffolds via Synchrotron Multiscale Digital Volume Correlation: Innovative Building Blocks for the Enthesis Regeneration

Enthesis lesions are one of the prevalent causes of injuries in the tendon tissue. The gradient of mineralization, extracellular matrix organization and auxetic mechanical properties, make enthesis regeneration challenging. Innovative electrospun fascicle-inspired nanofibrous poly(L-lactic)acid/collagen type I blend scaffolds were developed. Specifically, a mineralized fibrocartilage-inspired region (with/without nano-mineralization with hydroxyapatite), where random and aligned nanofibers coexist, is connected to a tendon-like region made of aligned nanofibers, through a conical non-mineralized fibrocartilage-inspired junction. Scanning electron microscopy and synchrotron nano-tomography show the morphological biomimicry of scaffolds with the natural tendon fascicles. Human mesenchymal stromal cells spheroids cultures confirm a balanced expression of both tendon, cartilage and bone markers on the non-mineralized scaffolds compared with the mineralized ones. Mechanical tests, at different physiological strain-rates, reveal a biomimetic mechanical behavior of scaffolds and the ability of junctions to tune the mechanics of their surrounding sites. Multiscale synchrotron in situ tensile tests, coupled with Digital Volume Correlation, elucidate the full-field strain distribution of scaffolds from the structural down to the nanofiber level, highlighting the auxetic mechanical behavior of junctions typical of the natural enthesis. The findings and cutting-edge investigations of our study suggest the suitability of these enthesis-inspired fascicles as innovative scaffolds for enhanced enthesis regeneration.

bioengineering↗

bioRxiv2023

Data-driven image mechanics (D2IM): a deep learning approach to predict displacement and strain fields from undeformed X-ray tomography images - Evaluation of bone mechanics.

Experimental measurement of displacement and strain fields using techniques such as digital volume correlation (DVC) from in situ X-ray computed tomography (XCT) has notably advanced the understanding of bone mechanics from organ to tissue level. Being experimental in nature, DVC output has been often employed to validate finite element (FE) models of bone improving their predictive ability. Despite the excellent results achieved, these techniques are complex, time consuming, potentially affecting tissue mechanical properties, and their predictive ability requiring prior knowledge of material properties. The recent advent of deep learning (DL) has enabled data-driven models, paving the way for the full exploitation of rich image datasets from which physics can be learnt and retained. Here we propose a novel data-driven image mechanics (D2IM) approach based on feed forward convolutional neural network (CNN) that learns from DVC displacement fields of vertebrae, predicting displacement and strain fields for undeformed XCT images. D2IM successfully predicted all displacement fields, particularly the one for the z loading axis (w), where high correlation (R2=0.93) and minimal error (as low as less than 1m) were found when comparing measured against predicted displacements. The predicted axial strain field in z ({varepsilon}zz) was also consistent in distribution with the measured one, displaying generally reduced errors (as low as few tens of {varepsilon}) in the regions within the vertebral body where the effect of border outliers was minimal. This is the first study using experimental full-field measurements on bone structures from DVC to inform DL-based model such as D2IM, which represents a major contribution in the prediction of displacement and strain fields only based on the greyscale content of undeformed XCT images. The future development of D2IM will incorporate a wider range of structures and loading scenarios for accurate prediction of physical fields in both hard and soft tissues, aiming at clinical translation for improved diagnostics.

bioengineering↗

bioRxiv2023

Full-field strain distribution in hierarchical electrospun nanofibrous Poly-L(lactic) acid and Collagen based scaffolds for tendon and ligament tissue regeneration: a multiscale study

Regeneration of injured tendons and ligaments (T/L) is a worldwide need. In this study electrospun hierarchical scaffolds made of a poly (L-lactic) acid and collagen blend were developed reproducing all the multiscale levels of aggregation of these tissues. Scanning electron microscopy, microCT and tensile mechanical tests were carried out, including a multiscale digital volume correlation analysis to measure the full-field strain distribution of electrospun structures. The principal tensile and compressive strains detected the pattern of strains caused by the nanofibers rearrangement, while the deviatoric strains revealed the related internal sliding of nanofibers and bundles. The results of this study confirmed the biomimicry of such electrospun hierarchical scaffolds, paving the way to further tissue engineering and clinical applications. Graphical Abstract O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=92 SRC="FIGDIR/small/543145v1_ufig1.gif" ALT="Figure 1"> View larger version (30K): org.highwire.dtl.DTLVardef@1a67baorg.highwire.dtl.DTLVardef@190149aorg.highwire.dtl.DTLVardef@38968borg.highwire.dtl.DTLVardef@123bb49_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioengineering↗

bioRxiv2023

Biofabrication of nanocomposite-based scaffolds containing human bone extracellular matrix for the differentiation of skeletal stem and progenitor cells

Autograft or metal implants are routinely used in skeletal repair but can fail to provide a long-term clinical resolution, emphasising the need for a functional biomimetic tissue engineering alternative. An attractive sustainable opportunity for tissue regeneration would be the application of human bone waste tissue for the synthesis of a material ink for 3D bioprinting of skeletal tissue. The use of human bone extracellular matrix (bone-ECM) offers an exciting potential for the development of an appropriate micro-environment for human bone marrow stromal cells (HBMSCs) to proliferate and differentiate along the osteogenic lineage. Extrusion-based deposition was mediated by the blending of human bone-ECM (B) with nanoclay (L, Laponite(R)) and alginate (A) polymer, to engineer a novel material ink (LAB). The inclusion of nanofiller and polymeric material increased the rheological, printability, and drug retention properties and, critically, the preservation of HBMSCs viability upon printing. The composite human bone-ECM-based 3D constructs containing vascular endothelial growth factor (VEGF) enhanced vascularisation following implantation in an ex vivo chick chorioallantoic membrane (CAM) model. Addition of bone morphogenetic protein-2 (BMP-2) with HBMSCs further enhanced vascularisation together with mineralisation after only 7 days. The current study demonstrates the synergistic combination of nanoclay with biomimetic materials, (alginate and bone-ECM) to support the formation of osteogenic tissue both in vitro and ex vivo and offers a promising novel 3D bioprinting approach to personalised skeletal tissue repair. Graphical AbstractEngineering nanoclay-based bone ECM novel bioink for bone regeneration. Human bone trabecular tissue was demineralised, decellularised and blended with nanoclay (Laponite(R)) and alginate after digestion. The resulting ink was investigated for printability following rheological and filament fusion investigation. The microstructural arrangement of the blends was examined together with viability and functionality of bioprinted HBMSCs. Finally, the ability of the novel blend to support drug release ex vivo in a CAM model was determined confirming the potential of the bone ECM ink to support bone formation. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=191 SRC="FIGDIR/small/536074v1_ufig1.gif" ALT="Figure 1"> View larger version (64K): org.highwire.dtl.DTLVardef@b9dc13org.highwire.dtl.DTLVardef@1f6f0dorg.highwire.dtl.DTLVardef@1b3c098org.highwire.dtl.DTLVardef@295192_HPS_FORMAT_FIGEXP M_FIG C_FIG

bioengineering↗

bioRxiv2022

Nonlinear micro finite element models based on digital volume correlation measurements predict early microdamage in newly formed bone

Bone regeneration in critical-sized defects is a clinical challenge, with biomaterials under constant development aiming at enhancing the natural bone healing process. The delivery of bone morphogenetic proteins (BMPs) in appropriate carriers represents a promising strategy for bone defect treatment but optimisation of the spatial-temporal release is still needed for the regeneration of bone with biological, structural, and mechanical properties comparable to the native tissue. Nonlinear micro finite element (FE) models can address some of these challenges by providing a tool able to predict the biomechanical strength and microdamage onset in newly formed bone when subjected to physiological or supraphysiological loads. Yet, these models need to be validated against experimental data. In this study, experimental local displacements in newly formed bone induced by osteoinductive biomaterials subjected to in situ X-ray computed tomography compression in the apparent elastic regime and measured using digital volume correlation (DVC) were used to validate FE models. Displacement predictions from homogeneous linear FE models were highly correlated to DVC-measured local displacements, while tissue heterogeneity capturing mineralisation differences showed negligible effects. Nonlinear FE models improved the correlation and showed that tissue microdamage occurs at low apparent strains. Microdamage seemed to occur next to large cavities or in biomaterial-induced thin trabeculae, independent of the mineralisation. While localisation of plastic strain accumulation was similar, the amount of damage accumulated in these locations was slightly higher when including material heterogeneity. These results demonstrate the ability of the nonlinear FE model to capture local microdamage in newly formed bone tissue and can be exploited to improve the current understanding of healing bone and mechanical competence. This will ultimately aid the development of BMPs delivery systems for bone defect treatment able to regenerate bone with optimal biological, mechanical, and structural properties.

bioengineering↗

bioRxiv2021

Fibrillar-Level Strain Gradients Across the Calcified Bone-Cartilage Interface

The bone-cartilage unit (BCU) is a universal feature in diarthrodial joints, which is mechanically-graded and subjected to shear and compressive strains. Changes in the BCU have been linked to osteoarthritis progression. Here we report existence of a physiological internal strain gradient (pre-strain) across the BCU at the ultrastructural scale of the extracellular matrix constituents, specifically the collagen fibril. We use X-ray scattering that probes changes in the axial periodicity of fibril-level D-stagger of tropocollagen molecules in the matrix fibrils, as a measure of microscopic pre-strain. We find that mineralized collagen nanofibrils in the calcified plate are in tensile pre-strain relative to the underlying trabecular bone. This behaviour contrasts with the previously accepted notion that fibrillar pre-strain (or D-stagger) in collagenous tissues always reduces with mineralization, via reduced hydration and associated swelling pressure. Within the calcified part of the BCU, a finer-scale gradient in pre-strain (0.6% increase over [~]50m) is observed. The increased fibrillar pre-strain is linked to prior research reporting large tissue-level residual strains under compression. The findings may have biomechanical adaptative significance: higher in-built molecular level resilience/damage resistance to physiological compression, and disruption of the molecular-level pre-strains during remodelling of the BCI may be potential factors in osteoarthritis-based degeneration.

bioengineering↗