Point spread function (PSF) engineering in an imaging system involves the introduction of additional optical components to efficiently encode object information. It typically relies on a well-established mapping relation between object space, PSF engineering plane, and image plane. However, this reliance can limit its application in imaging systems with unknown (“black box”) attributes, e. g. commercial photographic lenses, where a theoretical imaging model is not readily available. To address this limitation, we conduct wave-optics analysis of general lens systems, explicitly derive a general imaging model based on scaled Fourier transform, and develop an overall pipeline for automatic parameter determination of the model. When applied to a photographic lens, this pipeline demonstrates high PSF modeling accuracy, showcasing the capability of “seeing what you design”. We then implemented PSF engineering in this lens to find its optimal PSF for 3D localization. This work improves computational imaging by enabling accurate wavefront shaping design in arbitrary imaging systems.

Reliable extraction of structured data from radiology reports using Large Language Models (LLMs) remains challenging, especially for complex, non-English texts like Hebrew. This study introduces an agent-based uncertainty-aware approach to improve the trustworthiness of LLM predictions in medical applications. We analyzed 9,683 Hebrew radiology reports from Crohn's disease patients (from 2010 to 2023) across three medical centers. A subset of 512 reports was manually annotated for six gastrointestinal organs and 15 pathological findings, while the remaining reports were automatically annotated using HSMP-BERT. Structured data extraction was performed using Llama 3.1 (Llama 3-8b-instruct) with Bayesian Prompt Ensembles (BayesPE), which employed six semantically equivalent prompts to estimate uncertainty. An Agent-Based Decision Model integrated multiple prompt outputs into five confidence levels for calibrated uncertainty and was compared against three entropy-based models. Performance was evaluated using accuracy, F1 score, precision, recall, and Cohen's Kappa before and after filtering high-uncertainty cases. The agent-based model outperformed the baseline across all metrics, achieving an F1 score of 0.3967, recall of 0.6437, and Cohen's Kappa of 0.3006. After filtering high-uncertainty cases (greater than or equal to 0.5), the F1 score improved to 0.4787, and Kappa increased to 0.4258. Uncertainty histograms demonstrated clear separation between correct and incorrect predictions, with the agent-based model providing the most well-calibrated uncertainty estimates. By incorporating uncertainty-aware prompt ensembles and an agent-based decision model, this approach enhances the performance and reliability of LLMs in structured data extraction from radiology reports, offering a more interpretable and trustworthy solution for high-stakes medical applications.

Cancer is one of the leading causes of death worldwide and stained tissues’ biopsy analysis remains the standard method for pathology diagnostics. Major optical developments have improved pathological diagnostics lately. High quality microscopic optical scanners now allow whole-slide imaging of tissue sections and with the accessibility of datasets, digital imaging becomes attractive for biopsy analysis. Beyond its user-friendly features for image review and telepathology, digital imaging also enables the utilization of image processing and artificial intelligence (AI) which are advantageous for numerous applications. Nevertheless, AI faces significant barriers to widespread adoption. In order to extend the use of AI for clinical use in pathology, here we present an advanced approach for analyzing Hematoxylin and Eosin-stained biopsies and identify cancer cells. Our approach relies on the fusion of rapid spectral imaging measurements of biopsies with tailored machine learning algorithms designed explicitly for spectral images. The spectrum measured at each pixel provides much more information than standard color, that contains only three values of red, green, and blue intensities. We lately found that the spectral information provides high separability of normal and cancerous cells in Hematoxylin and Eosin-stained biopsies. This breakthrough surpasses previous obstacles, marking the potential for utilizing spectral imaging for cancer identification. Nevertheless, the method required identifying the nuclei in the tissue, a complex task that has not yet been addressed. Here, we demonstrate a rapid spectral imaging system combining artificial intelligence procedures for spectral-based nuclear segmentation using U-net models, with an adjusted input layer for the spectral imaging size. We trained two models; one with the full measured spectrum, and one based on color produced from the spectra. Excellent segmentation performance is achieved with both models, but the model trained on the full spectrum achieved significantly better results. The high performance of spectral image-based segmentation together with the simplicity of the system makes it applicable for tissue analysis and cancer classification in the clinical arena.

Background and objective: Treatment approaches for colorectal cancer (CRC) are highly dependent on the molecular subtype, as immunotherapy has shown efficacy in cases with microsatellite instability (MSI) but is ineffective for the microsatellite stable (MSS) subtype. There is promising potential in utilizing deep neural networks (DNNs) to automate the differentiation of CRC subtypes by analyzing hematoxylin and eosin (H&E) stained whole-slide images (WSIs). Due to the extensive size of WSIs, multiple instance learning (MIL) techniques are typically explored. However, existing MIL methods focus on identifying the most representative image patches for classification, which may result in the loss of critical information. Additionally, these methods often overlook clinically relevant information, like the tendency for MSI class tumors to predominantly occur on the proximal (right side) colon. Methods: We introduce ‘CIMIL-CRC’, a DNN framework that: (1) solves the MSI/MSS MIL problem by efficiently combining a pre-trained feature extraction model with principal component analysis (PCA) to aggregate information from all patches, and (2) integrates clinical priors, particularly the tumor location within the colon, into the model to enhance patient-level classification accuracy. We assessed our CIMIL-CRC method using the average area under the receiver operating characteristic curve (AUROC) from a 5-fold cross-validation experimental setup for model development on the TCGA-CRC-DX cohort, contrasting it with a baseline patch-level classification, a MIL-only approach, and a clinically-informed patch-level classification approach. Results: Our CIMIL-CRC outperformed all methods (AUROC: (95% CI 0.91–0.92), vs. (95% CI 0.76–0.82), (95% CI 0.85–0.88), and (95% CI 0.86–0.88), respectively). The improvement was statistically significant. To the best of our knowledge, this is the best result achieved for MSI/MSS classification on this dataset. Conclusion: Our CIMIL-CRC method holds promise for offering insights into the key representations of histopathological images and suggests a straightforward implementation.

Volumetric muscle loss (VML) refers to muscle tissue loss exceeding 20% within a functional area due to trauma or surgery, often leading to physical disabilities. VML treatment relies on the transplantation of autologous flaps harvested from a healthy-donor site while minimizing the probability of immune rejection. However, this approach often leads to donor-site morbidity and relies on a restricted supply of muscle tissue. Current solutions in tissue engineering focus on engineered grafts lacking hierarchical vasculature with a feeding vessel, thus limited by diffusion. This study expanded upon a new approach of multimodal bioprinting which enabled the fabrication of thick hierarchical vascular muscle flaps composed of bioprinted and vascularized skeletal muscle tissue, and a 3D-printed engineered macrovessel, which successfully repaired VML injury in-vivo. The flaps are implanted by anastomosing the macrovessel via microsurgery to the femoral artery in proximity to an induced VML injury in Sprague-Dawley rat hindlimbs. Immediate perfusion of the flaps is demonstrated, as is flap endurance to physiological blood pressure, flow, and shear stress. Flap implantation enhanced myocyte differentiation, and vascular ingrowth and facilitated tissue viability and integration. These results obtained by utilizing human-origin cells provide a foundation for fabricating patient-specific flaps for the treatment of extensive soft tissue defects.

In this contribution to the Orations - New Horizons of the Journal of Controlled Release, I present a personal perspective on the complexities of cancer nanomedicine and the approaches to master them. This oration draws mainly from my lab's journey to explore three transformative approaches to master complexities in the field: (1) leveraging text mining to construct dynamic knowledge bases for hypothesis generation in cell-specific drug delivery, (2) introducing the concept of meta-synergy to further optimize and classify multi-drug combinations across dimensions such as chemical loading, pharmacodynamics, and pharmacokinetics (3) utilizing automation to accelerate nanoparticle discovery with advanced screening methodologies such as aggregation-induced emission (AIE). I argue that by embracing complexity in nanomedicine, we can manifest new therapeutic possibilities, paving the way for more effective, precise, and adaptive treatment strategies.

Best cosmetic outcomes of breast reconstruction using tissue engineering techniques rely on the scaffold architecture and material, which are currently both to be determined. This study suggests an approach for a rational design of breast-shaped scaffold architecture, in which structural analysis is implemented to predict its stiffness and adjust it to that of the native tissue. This approach can help achieve the goal of optimal scaffold architecture for breast tissue engineering. 
Based on specifications defined in a preliminary implantation study of a non-rationally designed scaffold, and using analytical modeling and finite element analysis (FEA), we rationally designed a polycaprolactone (PCL) made, 3D-printed, highly porous, breast-shaped scaffold with a stiffness similar to the breast adipose tissue. This scaffold had an architecture of a double-shelled dome connected by pillars, with no bottom to allow direct contact of its fat graft with the host's blood vessels (Shelled Hemisphere Adaptive Design (SHAD)). To demonstrate the potential of the SHAD scaffold in breast tissue engineering, a proof-of-concept study was performed, in which SHAD scaffolds were embedded with human adipose derived mesenchymal stem cells (hAdMSCs), isolated from lipoaspirates, and implanted in Nod-Scid-Gamma (NSG) mouse model with a delayed fat graft injection. After 4 weeks of implantation, the SHAD implants were vascularized with a viable fat graft, indicating the suitability of the SHAD scaffold for breast tissue engineering. 


Positron emission tomography (PET) imaging plays a pivotal role in oncology for the early detection of metastatic tumors and response to therapy assessment due to its high sensitivity compared to anatomical imaging modalities. The balance between image quality and radiation exposure is critical, as reducing the administered dose results in a lower signal-to-noise ratio (SNR) and information loss, which may significantly affect clinical diagnosis. Deep learning (DL) algorithms have recently made significant progress in low-dose (LD) PET reconstruction. Nevertheless, a successful clinical application requires a thorough evaluation of uncertainty to ensure informed clinical judgment. We propose NPB-LDPET, a DL-based non-parametric Bayesian framework for LD PET reconstruction and uncertainty assessment. Our framework utilizes an Adam optimizer with stochastic gradient Langevin dynamics (SGLD) to sample from the underlying posterior distribution. We employed the Ultra-low-dose PET Challenge dataset to assess our framework’s performance relative to the Monte Carlo dropout benchmark. We evaluated global reconstruction accuracy utilizing SSIM, PSNR, and NRMSE, local lesion conspicuity using mean absolute error (MAE) and local contrast, and the clinical relevance of uncertainty maps employing correlation between the uncertainty measures and the dose reduction factor (DRF). Our NPB-LDPET reconstruction method exhibits a significantly superior global reconstruction accuracy for various DRFs (paired t-test, $$, N=10, 631). Moreover, we demonstrate a 21% reduction in MAE (573.54 vs. 723.70, paired t-test, $$, N=28) and an 8.3% improvement in local lesion contrast (2.077 vs. 1.916, paired t-test, $$, N=28). Furthermore, our framework exhibits a stronger correlation between the predicted uncertainty 95th percentile score and the DRF ($$ vs. $$, N=10, 631). The proposed framework has the potential to improve clinical decision-making for LD PET imaging by providing a more accurate and informative reconstruction while reducing radiation exposure.

Objective. Diabetic retinopathy (DR) is a serious diabetes complication that can lead to vision loss, making timely identification crucial. Existing data-driven algorithms for DR staging from digital fundus images (DFIs) often struggle with generalization due to distribution shifts between training and target domains. Approach. To address this, DRStageNet, a deep learning model, was developed using six public and independent datasets with 91 984 DFIs from diverse demographics. Five pretrained self-supervised vision transformers (ViTs) were benchmarked, with the best further trained using a multi-source domain (MSD) fine-tuning strategy. Main results. DINOv2 showed a 27.4% improvement in L-Kappa versus other pretrained ViT. MSD fine-tuning improved performance in four of five target domains. The error analysis revealing 60% of errors due to incorrect labels, 77.5% of which were correctly classified by DRStageNet. Significance. We developed DRStageNet, a DL model for DR, designed to accurately stage the condition while addressing the challenge of generalizing performance across target domains. The model and explainability heatmaps are available at www.aimlab-technion.com/lirot-ai.

Aims/Purpose: Deriving vascular features of retinal images is a proposed noninvasive method to assess vascular health. Although several studies have linked cardiovascular risk to retinal image features, they often rely on limited datasets or have used deep learning approaches with limited explainability. This research introduces an end-to-end method for analyzing the retinal vasculature in large retinal image datasets, applied for primary open angle glaucoma (POAG).

Methods: 115,237 retinal images of the UZ Leuven Glaucoma Clinic were extracted. 4858 unique images remain of POAG patients (n = 3010) and controls (n = 1848) after image quality assessment, optic disk detection, region of interest definition, automated segmentation of arterioles and venules (LUNet algorithm), and parametrization of the vascular biomarkers (PVBM toolbox). Analysis of covariance and linear mixed models were used to adjust for age, sex and disc size.

Results: Both arteriolar and venular diameter, area, length, tortuosity, branching angle, endpoints, intersection points and both mono- as multifractal dimension levels are independently lower in POAG patients and older patients (all p < 0.001), after adjustment for sex, disc size and multiple testing. The multivariate linear mixed models additionally show that the vascular features are significantly influenced by sex and disc size, which necessitates correct adjustment.

Conclusions: This is the first time a fully automated end-to-end pipeline is published for the analysis of retinal vascular geometry in a large cohort (n = 4858). Given the independent similarities in retinal vascular geometry changes between older age and POAG prevalence the theory of pronounced vascular ageing in POAG patients is proposed. This novel approach enables quantitative analysis of eye vasculature and supports the study of its correlation with specific diseases, facilitating reproducible analysis of large datasets and providing explainability to deep learning approaches.