Publications by Federico Bolelli

Foundation models serve as the backbone for numerous specialized models developed through fine-tuning. However, when the underlying pretrained model is updated or retrained (e.g., on larger and more curated datasets), the fine-tuned model becomes obsolete, losing its utility and requiring retraining. This raises the question: is it possible to transfer fine-tuning to a new release of the model? In this work, we investigate how to transfer fine-tuning to a new checkpoint without having to re-train, in a data-free manner. To do so, we draw principles from model re-basin and provide a recipe based on weight permutations to re-base the modifications made to the original base model, often called task vector. In particular, our approach tailors model re-basin for Transformer models, taking into account the challenges of residual connections and multi-head attention layers. Specifically, we propose a two-level method rooted in spectral theory, initially permuting the attention heads and subsequently adjusting parameters within select pairs of heads. Through extensive experiments on visual and textual tasks, we achieve the seamless transfer of fine-tuned knowledge to new pre-trained backbones without relying on a single training step or datapoint.

The usage of Multi Instance Learning (MIL) for classifying Whole Slide Images (WSIs) has recently increased. Due to their gigapixel size, the pixel-level annotation of such data is extremely expensive and time-consuming, practically unfeasible. For this reason, multiple automatic approaches have been raised in the last years to support clinical practice and diagnosis. Unfortunately, most state-of-the-art proposals apply attention mechanisms without considering the spatial instance correlation and usually work on a single-scale resolution. To leverage the full potential of pyramidal structured WSI, we propose a graph-based multi-scale MIL approach, DAS-MIL. Our model comprises three modules: i) a self-supervised feature extractor, ii) a graph-based architecture that precedes the MIL mechanism and aims at creating a more contextualized representation of the WSI structure by considering the mutual (spatial) instance correlation both inter and intra-scale. Finally, iii) a (self) distillation loss between resolutions is introduced to compensate for their informative gap and significantly improve the final prediction. The effectiveness of the proposed framework is demonstrated on two well-known datasets, where we outperform SOTA on WSI classification, gaining a +2.7% AUC and +3.7% accuracy on the popular Camelyon16 benchmark.

This article is about Connected Components Labeling (CCL) algorithms developed for GPU accelerators. The task itself is employed in many modern image-processing pipelines and represents a fundamental step in different scenarios, whenever object recognition is required. For this reason, a strong effort in the development of many different proposals devoted to improving algorithm performance using different kinds of hardware accelerators has been made. This paper focuses on GPU-based algorithmic solutions published in the last two decades, highlighting their distinctive traits and the improvements they leverage. The state-of-the-art review proposed is equipped with the source code, which allows to straightforwardly reproduce all the algorithms in different experimental settings. A comprehensive evaluation on multiple environments is also provided, including different operating systems, compilers, and GPUs. Our assessments are performed by means of several tests, including real-case images and synthetically generated ones, highlighting the strengths and weaknesses of each proposal. Overall, the experimental results revealed that block-based oriented algorithms outperform all the other algorithmic solutions on both 2D images and 3D volumes, regardless of the selected environment.

Since their invention in 1949, barcodes have remained the preferred method for automatic data capture, playing a crucial role in supply chain management. To detect a barcode in an image, multiple algorithms have been proposed in the literature, with a significant increase of interest in the topic since the rise of deep learning. However, research in the field suffers from many limitations, including the scarcity of public datasets and code implementations, which hampers the reproducibility and reliability of published results. For this reason, we developed "BarBeR" (Barcode Benchmark Repository), a benchmark designed for testing and comparing barcode detection algorithms. This benchmark includes the code implementation of various detection algorithms for barcodes, along with a suite of useful metrics. It offers a range of test setups and can be expanded to include any localization algorithm. In addition, we provide a large, annotated dataset of 8748 barcode images, combining multiple public barcode datasets with standardized annotation formats for both detection and segmentation tasks. Finally, we share the results obtained from running the benchmark on our dataset, offering valuable insights into the performance of different algorithms.

Implicit biases, subtle and unconscious attitudes, permeate various facets of human decision-making and are similarly pervasive in Artificial Intelligence (AI) systems. These biases can stem from shortcut learning, where models rely on superficial patterns that do not capture the underlying phenomena. Inspired by social psychology studies, we introduce two novel metrics to analyze implicit biases in visual-language models. Our comprehensive analysis of 90 open-clip models reveals widespread anomalies related to ethnicity and gender. The first metric considers the cosine similarity between images and text prompts related to social stereotypes. The second metric adapts the Implicit Association Test (IAT), which evaluates prejudice and hidden discrimination within human behavior. Our findings illustrate that conventional text-based debiasing efforts can inadvertently amplify second-order biases instead of mitigating them. Furthermore, in expanding our evaluation to multimodal Large Language Models (LLMs), we demonstrate disparities in the tendency to generate semantically positive or negative outputs, depending on the ethnicity or gender of the individuals depicted in the input images.

The failures of Deep Networks can sometimes be ascribed to biases in the data or algorithmic choices. Existing debiasing approaches exploit prior knowledge to avoid unintended solutions; we acknowledge that, in real-world settings, it could be unfeasible to gather enough prior information to characterize the bias, or it could even raise ethical considerations. We hence propose a novel debiasing approach, termed ClusterFix, which does not require any external hint about the nature of biases. Such an approach alters the standard empirical risk minimization and introduces a per-example weight, encoding how critical and far from the majority an example is. Notably, the weights consider how difficult it is for the model to infer the correct pseudo-label, which is obtained in a self-supervised manner by dividing examples into multiple clusters. Extensive experiments show that the misclassification error incurred in identifying the correct cluster allows for identifying examples prone to bias-related issues. As a result, our approach outperforms existing methods on standard benchmarks for bias removal and fairness.

Segmentation of the Inferior Alveolar Canal (IAC) is a critical aspect of dentistry and maxillofacial imaging, garnering considerable attention in recent research endeavors. Deep learning techniques have shown promising results in this domain, yet their efficacy is still significantly hindered by the limited availability of 3D maxillofacial datasets. An inherent challenge is posed by the size of input volumes, which necessitates a patch-based processing approach that compromises the neural network performance due to the absence of global contextual information. This study introduces a novel approach that harnesses the spatial information within the extracted patches and incorporates it into a Transformer architecture, thereby enhancing the segmentation process through the use of prior knowledge about the patch location. Our method significantly improves the Dice score by a factor of 4 points, with respect to the previous work proposed by Cipriano et al., while also reducing the training steps required by the entire pipeline. By integrating spatial information and leveraging the power of Transformer architectures, this research not only advances the accuracy of IAC segmentation, but also streamlines the training process, offering a promising direction for improving dental and maxillofacial image analysis.

Background: The growing amount of biomedical knowledge about cancer in combination with genome-scale patient profiling data offers unprecedented opportunities for personalized oncology. However, the large amounts of knowledge and data require scalable approaches to providing actionable information to support clinicians in decision-making [1]. Objective: To develop software and methods that integrate all relevant clinical and genomic data about patients and that enable the discovery of optimal personalized treatment options, together with the supporting literature knowledge and data. Methods: We exploit a Semantic Knowledge Graph (SKG), a type of database that represents medical data in the form of objects and relationships, linking previously unconnected information across several cancer databases. To build up this SKG (OncodashKB), we use the BioCypher library [2]. We then integrate clinical data from patients with high-grade serous ovarian cancer, including information on genome changes collected as part of the DECIDER project (http://deciderproject.eu). The SKG can then be queried to gather evidence paths linking patient-specific alterations to actionable drugs. Results: Our approach provides a fully automated, systematic, and reproducible data integration workflow, along with the use of existing expert-made ontologies to provide interoperability and semantic descriptions. The integrated data is assessed by experts on molecular tumor boards and allows for the exploration of relevant clinical and genomic patient data in a visually accessible format, designed for ease of interpretation by clinicians. Importantly, we expect the system to reveal unexpected evidence paths between patient sequencing data and optimal treatment options based on biomedical knowledge described in the literature and confirmed by high-level evidence. Conclusion: Decision support systems using graph databases emerge as valuable tools by revealing new connections between various patient data and treatment options shown in an easy-to-understand format. References: [1] Reisle, C., Williamson, L.M., Pleasance, E. et al. A platform for oncogenomic reporting and interpretation. Nat Commun 13, 756 (2022). https://doi.org/10.1038/s41467-022-28348-y [2] Lobentanzer, S., Aloy, P., Baumbach, J. et al. Democratizing knowledge representation with BioCypher. Nat Biotechnol 41, 1056–1059 (2023). https://doi.org/10.1038/s41587-023-01848-y.

The presence of visible particles in pharmaceutical products is a critical quality issue that demands strict monitoring. Recently, Convolutional Neural Networks (CNNs) have been widely used in industrial settings to detect defects, but there remains a gap in the literature concerning the detection of particles floating in liquid substances, mainly due to the lack of publicly available datasets. In this study, we focus on the detection of foreign particles in pharmaceutical liquid vials, leveraging two state-of-the-art deep-learning approaches adapted to our specific multiclass problem. The first methodology employs a standard ResNet-18 architecture, while the second exploits a Multi-Instance Learning (MIL) technique to efficiently deal with multiple images (sequences) of the same sample. To address the issue of no data availability, we devised and partially released an annotated dataset consisting of sequences containing 19 images for each sample, captured from rotating vials, both with and without impurities. The dataset comprises 2,426 sequences for a total of 46,094 images labeled at the sequence level and including five distinct classes. The proposed methodologies, trained on this new extensive dataset, represent advancements in the field, offering promising strategies to improve the safety and quality control of pharmaceutical products and setting a benchmark for future comparisons.

Introduction. Length of Stay (LOS) and Prolonged Length of Stay (pLOS) are critical indicators of hospital efficiency. Reducing pLOS is crucial for patient safety, autonomy, and bed allocation. This study investigates different machine learning (ML) models to predict LOS and pLOS. Methods. We analyzed a dataset of patients discharged from a northern Italian hospital between 2022 and 2023 as a retrospective cohort study. We compared sixteen regression algorithms and twelve classification methods for predicting LOS as either a continuous or multi-class variable (1-3 days, 4-10 days, >10 days). We also evaluated pLOS prediction using the same models, having pLOS defined as any hospitalization with LOS longer than 8 days. We further analyzed all models using two versions of the same dataset: one containing only structured data (e.g. demographics and clinical information), whereas the second one also containing features extracted from free-text diagnosis. Results. Our results indicate that ensemble models achieved the highest prediction accuracy for both LOS and pLOS, outperforming traditional single-algorithm models, particularly when using both structured and unstructured data extracted from diagnoses. Discussion. The integration of ML, particularly ensemble models, can significantly improve LOS prediction and identify patients at increased risk of pLOS. This information can guide healthcare professionals and bed managers in making informed decisions to enhance patient care and optimize resource allocation.