E. Fiorina

Large scale validation of the M5L lung CAD on heterogeneous CT datasets

PURPOSE M5L, a fully automated computer-aided detection (CAD) system for the detection and segmentation of lung nodules in thoracic computed tomography (CT), is presented and validated on several image datasets. METHODS M5L is the combination of two independent subsystems, based on the Channeler Ant Model as a segmentation tool [lung channeler ant model (lungCAM)] and on the voxel-based neural approach. The lungCAM was upgraded with a scan equalization module and a new procedure to recover the nodules connected to other lung structures; its classification module, which makes use of a feed-forward neural network, is based of a small number of features (13), so as to minimize the risk of lacking generalization, which could be possible given the large difference between the size of the training and testing datasets, which contain 94 and 1019 CTs, respectively. The lungCAM (standalone) and M5L (combined) performance was extensively tested on 1043 CT scans from three independent datasets, including a detailed analysis of the full Lung Image Database Consortium/Image Database Resource Initiative database, which is not yet found in literature. RESULTS The lungCAM and M5L performance is consistent across the databases, with a sensitivity of about 70% and 80%, respectively, at eight false positive findings per scan, despite the variable annotation criteria and acquisition and reconstruction conditions. A reduced sensitivity is found for subtle nodules and ground glass opacities (GGO) structures. A comparison with other CAD systems is also presented. CONCLUSIONS The M5L performance on a large and heterogeneous dataset is stable and satisfactory, although the development of a dedicated module for GGOs detection could further improve it, as well as an iterative optimization of the training procedure. The main aim of the present study was accomplished: M5L results do not deteriorate when increasing the dataset size, making it a candidate for supporting radiologists on large scale screenings and clinical programs.

Predictive Models Based on Support Vector Machines: Whole-Brain versus Regional Analysis of Structural MRI in the Alzheimer's Disease

Decision‐making systems trained on structural magnetic resonance imaging data of subjects affected by the Alzheimers disease (AD) and healthy controls (CTRL) are becoming widespread prognostic tools for subjects with mild cognitive impairment (MCI). This study compares the performances of three classification methods based on support vector machines (SVMs), using as initial sets of brain voxels (ie, features): (1) the segmented grey matter (GM); (2) regions of interest (ROIs) by voxel‐wise t‐test filtering; (3) parceled ROIs, according to prior knowledge. The recursive feature elimination (RFE) is applied in all cases to investigate whether feature reduction improves the classification accuracy. We analyzed more than 600 AD Neuroimaging Initiative (ADNI) subjects, training the SVMs on the AD/CTRL dataset, and evaluating them on a trial MCI dataset. The classification performance, evaluated as the area under the receiver operating characteristic curve (AUC), reaches AUC = (88.9 ± .5)% in 20‐fold cross‐validation on the AD/CTRL dataset, when the GM is classified as a whole. The highest discrimination accuracy between MCI converters and nonconverters is achieved when the SVM‐RFE is applied to the whole GM: with AUC reaching (70.7 ± .9)%, it outperforms both ROI‐based approaches in predicting the AD conversion.

INSIDE in-beam positron emission tomography system for particle range monitoring in hadrontherapy

Abstract. The quality assurance of particle therapy treatment is a fundamental issue that can be addressed by developing reliable monitoring techniques and indicators of the treatment plan correctness. Among the available imaging techniques, positron emission tomography (PET) has long been investigated and then clinically applied to proton and carbon beams. In 2013, the Innovative Solutions for Dosimetry in Hadrontherapy (INSIDE) collaboration proposed an innovative bimodal imaging concept that combines an in-beam PET scanner with a tracking system for charged particle imaging. This paper presents the general architecture of the INSIDE project but focuses on the in-beam PET scanner that has been designed to reconstruct the particles range with millimetric resolution within a fraction of the dose delivered in a treatment of head and neck tumors. The in-beam PET scanner has been recently installed at the Italian National Center of Oncologic Hadrontherapy (CNAO) in Pavia, Italy, and the commissioning phase has just started. The results of the first beam test with clinical proton beams on phantoms clearly show the capability of the in-beam PET to operate during the irradiation delivery and to reconstruct on-line the beam-induced activity map. The accuracy in the activity distal fall-off determination is millimetric for therapeutic doses.

Full-beam performances of a PET detector with synchrotron therapeutic proton beams

Treatment quality assessment is a crucial feature for both present and next-generation ion therapy facilities. Several approaches are being explored, based on prompt radiation emission or on PET signals by [Formula: see text]-decaying isotopes generated by beam interactions with the body. In-beam PET monitoring at synchrotron-based ion therapy facilities has already been performed, either based on inter-spill data only, to avoid the influence of the prompt radiation, or including both in-spill and inter-spill data. However, the PET images either suffer of poor statistics (inter-spill) or are more influenced by the background induced by prompt radiation (in-spill). Both those problems are expected to worsen for accelerators with improved duty cycle where the inter-spill interval is reduced to shorten the treatment time. With the aim of assessing the detector performance and developing techniques for background reduction, a test of an in-beam PET detector prototype was performed at the CNAO synchrotron-based ion therapy facility in full-beam acquisition modality. Data taken with proton beams impinging on PMMA phantoms showed the system acquisition capability and the resulting activity distribution, separately reconstructed for the in-spill and the inter-spill data. The coincidence time resolution for in-spill and inter-spill data shows a good agreement, with a slight deterioration during the spill. The data selection technique allows the identification and rejection of most of the background originated during the beam delivery. The activity range difference between two different proton beam energies (68 and 72 MeV) was measured and found to be in sub-millimeter agreement with the expected result. However, a slightly longer (2 mm) absolute profile length is obtained for in-spill data when compared to inter-spill data.

Generating a Minimal Set of Templates for the Hippocampal Region in MR Neuroimages

We detail a procedure for generating a set of templates for the hippocampal region in magnetic resonance (MR) images, representative of the clinical conditions of the population under investigation.

The Channeler Ant Model: Object segmentation with virtual ant colonies

3-D object segmentation is an important and challenging topic in computer vision that could be tackled with artificial life models. A Channeler Ant Model (CAM), based on the natural ant capabilities of dealing with 3-D environments through self-organization and emergent behaviours, is proposed. Ant colonies, defined in terms of moving, pheromone laying, reproduction, death and deviating behaviours rules, is able to segment artificially generated objects of different shape, intensity, background. The model depends on few parameters and provides an elegant solution for the segmentation of 3-D structures in noisy environments with unknown range of image intensities: even when there is a partial overlap between the intensity and noise range, it provides a complete segmentation with negligible contamination (i.e., fraction of segmented voxels that do not belong to the object).

Online proton therapy monitoring: clinical test of a Silicon-photodetector-based in-beam PET

Particle therapy exploits the energy deposition pattern of hadron beams. The narrow Bragg Peak at the end of range is a major advantage but range uncertainties can cause severe damage and require online verification to maximise the effectiveness in clinics. In-beam Positron Emission Tomography (PET) is a non-invasive, promising in-vivo technique, which consists in the measurement of the β+ activity induced by beam-tissue interactions during treatment, and presents the highest correlation of the measured activity distribution with the deposited dose, since it is not much influenced by biological washout. Here we report the first clinical results obtained with a state-of-the-art in-beam PET scanner, with on-the-fly reconstruction of the activity distribution during irradiation. An automated time-resolved quantitative analysis was tested on a lacrimal gland carcinoma case, monitored during two consecutive treatment sessions. The 3D activity map was reconstructed every 10 s, with an average delay between beam delivery and image availability of about 6 s. The correlation coefficient of 3D activity maps for the two sessions (above 0.9 after 120 s) and the range agreement (within 1 mm) prove the suitability of in-beam PET for online range verification during treatment, a crucial step towards adaptive strategies in particle therapy.

First results of the INSIDE in-beam PET scanner for the on-line monitoring of particle therapy treatments

Quality assessment of particle therapy treatments by means of PET systems has been carried out since late `90 and it is one of the most promising in-vivo non invasive monitoring techniques employed clinically. It can be performed with a diagnostic PET scanners installed outside the treatment room (off-line monitoring) or inside the treatment room (in-room monitoring). However the most efficient way is by integrating a PET scanner with the treatment delivery system (on-line monitoring) so that the biological wash out and the patient repositioning errors are minimized. In this work we present the performance of the in-beam PET scanner developed within the INSIDE project. The INSIDE PET scanner is made of two planar heads, 10 cm wide (transaxially) and 25 cm long (axially), composed of pixellated LFS crystals coupled to Hamamatsu MPPCs. Custom designed Front-End Electronics (FE) and Data AcQuisition (DAQ) systems allow an on-line reconstruction of PET images from separated in-spill and inter-spill data sets. The INSIDE PET scanner has been recently delivered at the CNAO (Pavia, Italy) hadrontherapy facility and the first experimental measurements have been carried out. Homogeneous PMMA phantoms and PMMA phantoms with small air and bone inserts were irradiated with monoenergetic clinical proton beams. The activity range was evaluated at various benchmark positions within the field of view to assess the homogeneity of response of the PET system. Repeated irradiations of PMMA phantoms with clinical spread out Bragg peak proton beams were performed to evaluate the reproducibility of the PET signal. The results found in this work show that the response of the INSIDE PET scanner is independent of the position within the radiation field. Results also show the capability of the INSIDE PET scanner to distinguish variations of the activity range due to small tissue inhomogeneities. Finally, the reproducibility of the activity range measurement was within 1 mm.

Fully automated hippocampus segmentation with virtual ant colonies

The development of tools for a fully automatic segmentation of the relevant brain structures, such as the hippocampus, is potentially very useful for pathologies detection. In this paper, a method for the automated hippocampal segmentation, based on virtual ant colonies, is proposed. The algorithm used, the Channeler Ant Model (CAM), represents an effective way to segment 3D objects with a complex shape in a noisy background. The CAM was modified by inserting a shape knowledge that is crucial to face the hippocampus segmentation. The algorithm was trained and tested using a database of 56 T1 weighted MRI images with a known manual segmentation of the hippocampus volume. The results are comparable to other methods: an average Dice Index of 0.74 and 0.72 is obtained over the left and right hippocampi, respectively. The lack of a heavy training procedure, because all the model parameters are fixed, and the speed make this approach very effective.

Monte Carlo simulation tool for online treatment monitoring in hadrontherapy with in-beam PET: A patient study

Hadrontherapy is a method for treating cancer with very targeted dose distributions and enhanced radiobiological effects. To fully exploit these advantages, in vivo range monitoring systems are required. These devices measure, preferably during the treatment, the secondary radiation generated by the beam-tissue interactions. However, since correlation of the secondary radiation distribution with the dose is not straightforward, Monte Carlo (MC) simulations are very important for treatment quality assessment. The INSIDE project constructed an in-beam PET scanner to detect signals generated by the positron-emitting isotopes resulting from projectile-target fragmentation. In addition, a FLUKA-based simulation tool was developed to predict the corresponding reference PET images using a detailed scanner model. The INSIDE in-beam PET was used to monitor two consecutive proton treatment sessions on a patient at the Italian Center for Oncological Hadrontherapy (CNAO). The reconstructed PET images were updated every 10 s providing a near real-time quality assessment. By half-way through the treatment, the statistics of the measured PET images were already significant enough to be compared with the simulations with average differences in the activity range less than 2.5 mm along the beam direction. Without taking into account any preferential direction, differences within 1 mm were found. In this paper, the INSIDE MC simulation tool is described and the results of the first in vivo agreement evaluation are reported. These results have justified a clinical trial, in which the MC simulation tool will be used on a daily basis to study the compliance tolerances between the measured and simulated PET images.