arXiv subjects

Salvador Tortajada

Publications and source records attributed to Salvador Tortajada.

3 recordsLinked to original sources

arXiv2026

Real-Time Requirements and Transferability in Compton Imaging: From the Detector Chain to the Application

Compton cameras are proposed for tasks whose value decays with delay: verifying a range during irradiation, guiding an intervention, characterising an inaccessible volume, surveying a band no telescope covers. Whether a device can serve such a task is settled by the composition of its whole chain, while the literature that would answer the question is organised by stage -- so claims made at application level routinely rest on evidence obtained at component level. This review walks that chain, asking at each stage what binds first and what a reader can determine from the published text, and then asks what governs whether a capability transfers between groups and between application domains. The evidence is of two kinds: 83 full texts scored in context against defined markers, and, for five application domains measured alike, the size of the receiving literature, of the need it states, of the incumbent and of deployment commitment. Performance is reported at one operating point in 65 of 83 full texts, with count rate swept in none: the field reports values where transfer requires gradients. An accelerator is used in 39 works and a learned model in 31, while separability is discussed in 28, a memory footprint given in 9, and the cost of a precomputation or the inference time of a model in none. Across domains, neither the size of the receiving literature, nor of the incumbent, nor of the stated need orders the domains as deployment commitment does, while whether the incumbent can serve the task at all, and how many domain boundaries the output must cross, do so consistently. The distribution of publication is close to the inverse of the distribution of deployment commitment. We give the quantities a report must contain for a third party to judge whether a method fits an application it was not built for.

physics.med-ph↗

arXiv2026

Bounded-Latency Spherical-Histogram Reconstruction for Compton Cameras

Gamma-ray imaging with Compton cameras is computationally demanding because conventional reconstruction retains the list-mode acquisition inside the inversion loop: event-dependent cone/voxel interactions must be recomputed as the event count grows. We present a spherical-histogram framework in which each Compton event is encoded online into detector-centred angular histograms. Volumetric reconstruction is then performed from coherent histogram snapshots using a precomputed sparse projection operator. This turns the event stream into a bounded reconstruction state, decoupling event accumulation from iterative inversion. The method supports multi-view and multi-resolution operation, non-blocking acquisition, and iterative forward/backward reconstruction whose dominant cost depends.

physics.ins-det↗

arXiv2019

Design of one-year mortality forecast at hospital admission based: a machine learning approach

Background: Palliative care is referred to a set of programs for patients that suffer life-limiting illnesses. These programs aim to guarantee a minimum level of quality of life (QoL) for the last stage of life. They are currently based on clinical evaluation of risk of one-year mortality. Objectives: The main objective of this work is to develop and validate machine-learning based models to predict the exitus of a patient within the next year using data gathered at hospital admission. Methods: Five machine learning techniques were applied in our study to develop machine-learning predictive models: Support Vector Machines, K-neighbors Classifier, Gradient Boosting Classifier, Random Forest and Multilayer Perceptron. All models were trained and evaluated using the retrospective dataset. The evaluation was performed with five metrics computed by a resampling strategy: Accuracy, the area under the ROC curve, Specificity, Sensitivity, and the Balanced Error Rate. Results: All models for forecasting one-year mortality achieved an AUC ROC from 0.858 to 0.911. Specifically, Gradient Boosting Classifier was the best model, producing an AUC ROC of 0.911 (CI 95%, 0.911 to 0.912), a sensitivity of 0.858 (CI 95%, 0.856 to 0.86) and a specificity of 0.807 (CI 95%, 0.806 to 0808) and a BER of 0.168 (CI 95%, 0.167 to 0.169). Conclusions: The analysis of common information at hospital admission combined with machine learning techniques produced models with competitive discriminative power. Our models reach the best results reported in state of the art. These results demonstrate that they can be used as an accurate data-driven palliative care criteria inclusion.

cs.LG↗