Clinical Use of a Commercial Artificial Intelligence-Based Software for Autocontouring in Radiation Therapy: Geometric Performance and Dosimetric Impact.

PURPOSE: When autocontouring based on artificial intelligence (AI) is used in the radiotherapy (RT) workflow, the contours are reviewed and eventually adjusted by a radiation oncologist before an RT treatment plan is generated, with the purpose of improving dosimetry and reducing both interobserver variability and time for contouring. The purpose of this study was to evaluate the results of application of a commercial AI-based autocontouring for RT, assessing both geometric accuracies and the influence on optimized dose from automatically generated contours after review by human operator. MATERIALS AND METHODS: A commercial autocontouring system was applied to a retrospective database of 40 patients, of which 20 were treated with radiotherapy for prostate cancer (PCa) and 20 for head and neck cancer (HNC). Contours resulting from AI were compared against AI contours reviewed by human operator and human-only contours using Dice similarity coefficient (DSC), Hausdorff distance (HD), and relative volume difference (RVD). Dosimetric indices such as Dmean, D0.03cc, and normalized plan quality metrics were used to compare dose distributions from RT plans generated from structure sets contoured by humans assisted by AI against plans from manual contours. The reduction in contouring time obtained by using automated tools was also assessed. A Wilcoxon rank sum test was computed to assess the significance of differences. Interobserver variability of the comparison of manual vs. AI-assisted contours was also assessed among two radiation oncologists for PCa. RESULTS: For PCa, AI-assisted segmentation showed good agreement with expert radiation oncologist structures with average DSC among patients ≥ 0.7 for all structures, and minimal radiation oncology adjustment of structures (DSC of adjusted versus AI structures ≥ 0.91). For HNC, results of comparison between manual and AI contouring varied considerably e.g., 0.77 for oral cavity and 0.11-0.13 for brachial plexus, but again, adjustment was generally minimal (DSC of adjusted against AI contours 0.97 for oral cavity, 0.92-0.93 for brachial plexus). The difference in dose for the target and organs at risk were not statistically significant between human and AI-assisted, with the only exceptions of D0.03cc to the anal canal and Dmean to the brachial plexus. The observed average differences in plan quality for PCa and HNC cases were 8% and 6.7%, respectively. The dose parameter changes due to interobserver variability in PCa were small, with the exception of the anal canal, where large dose variations were observed. The reduction in time required for contouring was 72% for PCa and 84% for HNC. CONCLUSIONS: When an autocontouring system is used in combination with human review, the time of the RT workflow is significantly reduced without affecting dose distribution and plan quality.

Predicting Local Failure after Partial Prostate Re-Irradiation Using a Dosiomic-Based Machine Learning Model.

The aim of this study is to predict local failure after partial prostate re-irradiation for the treatment of isolated locally recurrent prostate cancer by using a machine learning classifier based on radiomic features from pre-treatment computed tomography (CT), positron-emission tomography (PET) and biological effective dose distribution (BED) of the radiotherapy plan. The analysis was conducted on a monocentric dataset of 43 patients with evidence of isolated intraprostatic recurrence of prostate cancer after primary external beam radiotherapy. All patients received partial prostate re-irradiation delivered by volumetric modulated arc therapy. The gross tumor volume (GTV) of each patient was manually contoured from planning CT, choline-PET and dose maps. An ensemble machine learning pipeline including unbalanced data correction and feature selection was trained using the radiomic and dosiomic features as input for predicting occurrence of local failure. The model performance was assessed using sensitivity, specificity, accuracy and area under receiver operating characteristic curves of the score function in 10-fold cross validation repeated 100 times. Local failure was observed in 13 patients (30%), with a median time to recurrence of 36.7 months (range = 6.1-102.4 months). A four variables ensemble machine learning model resulted in accuracy of 0.62 and AUC 0.65. According to our results, a dosiomic machine learning classifier can predict local failure after partial prostate re-irradiation.

A Novel Benchmarking Approach to Assess the Agreement among Radiomic Tools.

Background The translation of radiomic models into clinical practice is hindered by the limited reproducibility of features across software and studies. Standardization is needed to accelerate this process and to bring radiomics closer to clinical deployment. Purpose To assess the standardization level of seven radiomic software programs and investigate software agreement as a function of built-in image preprocessing (eg, interpolation and discretization), feature aggregation methods, and the morphological characteristics (ie, volume and shape) of the region of interest (ROI). Materials and Methods The study was organized into two phases: In phase I, the two Image Biomarker Standardization Initiative (IBSI) phantoms were used to evaluate the IBSI compliance of seven software programs. In phase II, the reproducibility of all IBSI-standardized radiomic features across tools was assessed with two custom Italian multicenter Shared Understanding of Radiomic Extractors (ImSURE) digital phantoms that allowed, in conjunction with a systematic feature extraction, observations on whether and how feature matches between program pairs varied depending on the preprocessing steps, aggregation methods, and ROI characteristics. Results In phase I, the software programs showed different levels of completeness (ie, the number of computable IBSI benchmark values). However, the IBSI-compliance assessment revealed that they were all standardized in terms of feature implementation. When considering additional preprocessing steps, for each individual program, match percentages fell by up to 30%. In phase II, the ImSURE phantoms showed that software agreement was dependent on discretization and aggregation as well as on ROI shape and volume factors. Conclusion The agreement of radiomic software varied in relation to factors that had already been standardized (eg, interpolation and discretization methods) and factors that need standardization. Both dependences must be resolved to ensure the reproducibility of radiomic features and to pave the way toward the clinical adoption of radiomic models. Published under a CC BY 4.0 license. Online supplemental material is available for this article. See also the editorial by Steiger in this issue. An earlier incorrect version appeared online and in print. This article was corrected on March 2, 2022.

Combining computed tomography and biologically effective dose in radiomics and deep learning improves prediction of tumor response to robotic lung stereotactic body radiation therapy.

PURPOSE: The aim of this study is to improve the performance of machine learning (ML) models in predicting response of non-small cell lung cancer (NSCLC) to stereotactic body radiation therapy (SBRT) by integrating image features from pre-treatment computed tomography (CT) with features from the biologically effective dose (BED) distribution. MATERIALS AND METHODS: Image features, consisting of crafted radiomic features or machine-learned features extracted using a convolutional neural network, were calculated from pre-treatment CT data and from dose distributions converted into BED for 80 NSCLC lesions over 76 patients treated with robotic guided SBRT. ML models using different combinations of features were trained to predict complete or partial response according to response criteria in solid tumors, including radiomics CT (RadCT ), radiomics CT and BED (RadCT,BED ), deep learning (DL) CT (DLCT ), and DL CT and BED (DLCT,BED ). Training of ML included feature selection by neighborhood component analysis followed by ensemble ML using robust boosting. A model was considered as acceptable when the sum of average sensitivity and specificity on test data in repeated cross validations was at least 1.5. RESULTS: Complete or partial response occurred in 58 out of 80 lesions. The best models to predict the tumor response were those using BED variables, achieving significantly better area under curve (AUC) and accuracy than those using only features from CT, including a RadCT,BED model using three radiomic features from BED, which scored an accuracy of 0.799 (95% confidence intervals (0.75-0.85)) and AUC of 0.773 (0.688-0.846), and a DLCT,BED model also using three variables with an accuracy of 0.798 (0.649-0.829) and AUC of 0.812 (0.755-0.867). CONCLUSION: According to our results, the inclusion of BED features improves the response prediction of ML models for lung cancer patients undergoing SBRT, regardless of the use of radiomic or DL features.

Artificial intelligence applications in medical imaging: A review of the medical physics research in Italy.

PURPOSE: To perform a systematic review on the research on the application of artificial intelligence (AI) to imaging published in Italy and identify its fields of application, methods and results. MATERIALS AND METHODS: A Pubmed search was conducted using terms Artificial Intelligence, Machine Learning, Deep learning, imaging, and Italy as affiliation, excluding reviews and papers outside time interval 2015-2020. In a second phase, participants of the working group AI4MP on Artificial Intelligence of the Italian Association of Physics in Medicine (AIFM) searched for papers on AI in imaging. RESULTS: The Pubmed search produced 794 results. 168 studies were selected, of which 122 were from Pubmed search and 46 from the working group. The most used imaging modality was MRI (44%) followed by CT(12%) ad radiography/mammography (11%). The most common clinical indication were neurological diseases (29%) and diagnosis of cancer (25%). Classification was the most common task for AI (57%) followed by segmentation (16%). 65% of studies used machine learning and 35% used deep learning. We observed a rapid increase of research in Italy on artificial intelligence in the last 5 years, peaking at 155% from 2018 to 2019. CONCLUSIONS: We are witnessing an unprecedented interest in AI applied to imaging in Italy, in a diversity of fields and imaging techniques. Further initiatives are needed to build common frameworks and databases, collaborations among different types of institutions, and guidelines for research on AI.

Radiomics and deep learning in lung cancer.

Lung malignancies have been extensively characterized through radiomics and deep learning. By providing a three-dimensional characterization of the lesion, models based on radiomic features from computed tomography (CT) and positron-emission tomography (PET) have been developed to detect nodules, distinguish malignant from benign lesions, characterize their histology, stage, and genotype. Deep learning models have been applied to automatically segment organs at risk in lung cancer radiotherapy, stratify patients according to the risk for local and distant recurrence, and identify patients candidate for molecular targeted therapy and immunotherapy. Moreover, radiomics has also been applied successfully to predict side effects such as radiation- and immunotherapy-induced pneumonitis and differentiate lung injury from recurrence. Radiomics could also untap the potential for further use of the cone beam CT acquired for treatment image guidance, four-dimensional CT, and dose-volume data from radiotherapy treatment plans. Radiomics is expected to increasingly affect the clinical practice of treatment of lung tumors, optimizing the end-to-end diagnosis-treatment-follow-up chain. The main goal of this article is to provide an update on the current status of lung cancer radiomics.

Electron Density and Biologically Effective Dose (BED) Radiomics-Based Machine Learning Models to Predict Late Radiation-Induced Subcutaneous Fibrosis.

Purpose: to predict the occurrence of late subcutaneous radiation induced fibrosis (RIF) after partial breast irradiation (PBI) for breast carcinoma by using machine learning (ML) models and radiomic features from 3D Biologically Effective Dose (3D-BED) and Relative Electron Density (3D-RED). Methods: 165 patients underwent external PBI following a hypo-fractionation protocol consisting of 40 Gy/10 fractions, 35 Gy/7 fractions, and 28 Gy/4 fractions, for 73, 60, and 32 patients, respectively. Physicians evaluated toxicity at regular intervals by the Common Terminology Adverse Events (CTAE) version 4.0. RIF was assessed every 3 months after the completion of radiation course and scored prospectively. RIF was experienced by 41 (24.8%) patients after average 5 years of follow up. The Hounsfield Units (HU) of the CT-images were converted into relative electron density (3D-RED) and Dose maps into Biologically Effective Dose (3D-BED), respectively. Shape, first-order and textural features of 3D-RED and 3D-BED were calculated in the planning target volume (PTV) and breast. Clinical and demographic variables were also considered (954 features in total). Imbalance of the dataset was addressed by data augmentation using ADASYN technique. A subset of non-redundant features that best predict the data was identified by sequential feature selection. Support Vector Machines (SVM), ensemble machine learning (EML) using various aggregation algorithms and Naive Bayes (NB) classifiers were trained on patient dataset to predict RIF occurrence. Models were assessed using sensitivity and specificity of the ML classifiers and the area under the receiver operator characteristic curve (AUC) of the score functions in repeated 5-fold cross validation on the augmented dataset. Results: The SVM model with seven features was preferred for RIF prediction and scored sensitivity 0.83 (95% CI 0.80-0.86), specificity 0.75 (95% CI 0.71-0.77) and AUC of the score function 0.86 (0.85-0.88) on cross-validation. The selected features included cluster shade and Run Length Non-uniformity of breast 3D-BED, kurtosis and cluster shade from PTV 3D-RED, and 10th percentile of PTV 3D-BED. Conclusion: Textures extracted from 3D-BED and 3D-RED in the breast and PTV can predict late RIF and may help better select patient candidates to exclusive PBI.

Electron radiotherapy (IOERT) for applications outside of the breast: Dosimetry and influence of tissue inhomogeneities.

PURPOSE: The purpose of study is to investigate the dosimetry of electron intraoperative radiotherapy (IOERT) of the Intraop Mobetron 2000 mobile LINAC in treatments outside of the breast. After commissioning and external validation of dosimetry, we report in vivo results of measurements for treatments outside the breast in a large patient cohort, and investigate if the presence of inhomogeneities can affect in vivo measurements. METHODS AND MATERIALS: Applicator factors and profile curves were measured with a stereotactic diode. The applicators factors of the 6 cm flat and beveled applicators were also confirmed with radiochromic films, parallel-plate ion chamber and by an external audit performed with ThermoLuminescent Dosimeters (TLDs). The influence of bone on dose was investigated by using radiochromic films attached to an insert equivalent to cortical bone, immersed in the water phantom. In vivo dosimetry was performed on 126 patients treated with IOERT using metal oxide-silicon semiconductor field effect transistors (MOSFETs) placed on the tumor bed. RESULTS: Relatively small differences were found among different detectors for measurements of applicator factors. In the external audit, the agreement with the TLD was mostly within ±0.2%. The largest increase of dose due to the presence of cortical bone insert was +6.0% with energy 12 MeV and 3 cm applicator. On average, in vivo dose was significantly (+3.1%) larger than prescribed dose. CONCLUSION: IOERT in applications outside the breast results in low discrepancies between in vivo and prescribed doses, which can be also explained with the presence of tissue inhomogeneity.

Prediction of skin dose in low-kV intraoperative radiotherapy using machine learning models trained on results of in vivo dosimetry.

PURPOSE: The purpose of this study was to implement a machine learning model to predict skin dose from targeted intraoperative (TARGIT) treatment resulting in timely adoption of strategies to limit excessive skin dose. METHODS: A total of 283 patients affected by invasive breast carcinoma underwent TARGIT with a prescribed dose of 6 Gy at 1 cm, after lumpectomy. Radiochromic films were used to measure the dose to the skin for each patient. Univariate statistical analysis was performed to identify correlation of physical and patient variables with measured dose. After feature selection of predictors of in vivo skin dose, machine learning models stepwise linear regression (SLR), support vector regression (SVR), ensemble with bagging or boosting, and feed forward neural networks were trained on results of in vivo dosimetry to derive models to predict skin dose. Models were evaluated by tenfold cross validation and ranked according to root mean square error (RMSE) and adjusted correlation coefficient of true vs predicted values (adj-R2 ). RESULTS: The predictors correlated with in vivo dosimetry were the distance of skin from source, depth-dose in water at depth of the applicator in the breast, use of a replacement source, and irradiation time. The best performing model was SVR, which scored RMSE and adj-R2 , equal to 0.746 [95% confidence intervals (CI), 95% CI 0.737,0.756] and 0.481 (95% CI 0.468,0.494), respectively, on the tenfold cross validation. CONCLUSION: The model trained on results of in vivo dosimetry can be used to predict skin dose during setup of patient for TARGIT and this allows for timely adoption of strategies to prevent of excessive skin dose.

Voxel-by-voxel correlation between radiologically radiation induced lung injury and dose after image-guided, intensity modulated radiotherapy for lung tumors.

PURPOSE: To correlate radiation dose to the risk of severe radiologically-evident radiation-induced lung injury (RRLI) using voxel-by-voxel analysis of the follow-up computed tomography (CT) of patients treated for lung cancer with hypofractionated helical Tomotherapy. METHODS AND MATERIALS: The follow-up CT scans from 32 lung cancer patients treated with various regimens (5, 8, and 25 fractions) were registered to pre-treatment CT using deformable image registration (DIR). The change in density was calculated for each voxel within the combined lungs minus the planning target volume (PTV). Parameters of a Probit formula were derived by fitting the occurrences of changes of density in voxels greater than 0.361gcm-3 to the radiation dose. The model's predictive capability was assessed using the area under receiver operating characteristic curve (AUC), the Kolmogorov-Smirnov test for goodness-of-fit, and the permutation test (Ptest). RESULTS: The best-fit parameters for prediction of RRLI 6months post RT were D50 of 73.0 (95% CI 59.2.4-85.3.7)Gy, and m of 0.41 (0.39-0.46) for hypofractionated (5 and 8 fractions) and D50 of 96.8 (76.9-123.9)Gy, and m of 0.36 (0.34-0.39) for 25 fractions RT. According to the goodness-of-fit test the null hypothesis of modeled and observed occurrence of RRLI coming from the same distribution could not be rejected. The AUC was 0.581 (0.575-0.583) for fractionated and 0.579 (0.577-0.581) for hypofractionated patients. The predictive models had AUC>upper 95% band of the Ptest. CONCLUSIONS: The correlation of voxel-by-voxel density increase with dose can be used as a support tool for differential diagnosis of tumor from benign changes in the follow-up of lung IMRT patients.

Depth of interaction determination in monolithic scintillator with double side SiPM readout.

BACKGROUND: Monolithic scintillators read out by arrays of photodetectors represent a promising solution to obtain high spatial resolution and the depth of interaction (DOI) of the annihilation photon. We have recently investigated a detector geometry composed of a monolithic scintillator readout on two sides by silicon photomultiplier (SiPM) arrays, and we have proposed two parameters for the DOI determination: the difference in the number of triggered SiPMs on the two sides of the detector and the difference in the maximum collected signal on a single SiPM on each side. This work is focused on the DOI calibration and on the determination of the capability of our detector. For the DOI calibration, we studied a method which can be implemented also in detectors mounted in a full PET scanner. We used a PET detector module composed of a monolithic 20 × 20 × 10 mm3 LYSO scintillator crystal coupled on two opposite faces to two arrays of SiPMs. On each side, the scintillator was coupled to 6 × 6 SiPMs. In this paper, the two parameters previously proposed for the DOI determination were calibrated with two different methods. The first used a lateral scan of the detector with a collimated 511 keV pencil beam at steps of 0.5 mm to study the detector DOI capability, while the second used the background radiation of the 176Lu in the scintillator. The DOI determination capability was tested on different regions of the detector using each parameter and the combination of the two. RESULTS: With both parameters for the DOI determination, in the lateral scan, the bias between the mean reconstructed DOI and the real beam position was lower than 0.3 mm, and the DOI distribution had a standard deviation of about 1.5 mm. When using the calibration with the radioactivity of the LYSO, the mean bias increased of about 0.2 mm but with no degradation of the standard deviation of the DOI distribution. CONCLUSIONS: The two parameters allow to achieve a DOI resolution comparable with the state of the art, giving a continuous information about the three-dimensional interaction position of the scintillation. These results were obtained by using simple estimators and a detector scalable to a whole PET system. The DOI calibration obtained using lutetium natural radioactivity gives results comparable to the other standard method but appears more readily applicable to detectors mounted in a full PET scanner.

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

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.

[Massive hematemesis resulting from stomach rupture caused by pseudoaneurysm of the splenic artery in a patient with chronic pancreatitis. A case report]. / Ematemesi massiva da rottura nello stomaco di pseudo-aneurisma dell'arteria splenica in paziente con pancreatite cronica. Osservazione personale.

The authors report a case of a pancreatic pseudocyst, due to alcoholic chronic pancreatitis, that was transformed into a pseudoaneurysm of the splenic artery as a result of vascular erosion and that manifested itself with massive haematemesis due to spontaneous fistulisation in the stomach. After defining the incidence of the pancreatic disease and of this unusual form of gastric bleeding, particular attention is devoted to the clinical data and to the aetiopathogenic and physiopathological mechanisms involved in the vascular glandular and periglandular damage, outlining the sources and sites of bleeding. The authors go on to discuss the rationale in using imaging techniques, which cannot ignore the haemodynamic conditions of the patient and the conviction that the execution time of a selective coeliac arteriography never represents an unacceptable delay in the management of a life-threatening ruptured pancreatic pseudoaneurysm. This conviction is due both to the therapeutic potential inherent in the method itself and to the greater safety with which the following operation can be performed, owing to the topographical guidance the procedure provides. After a review of the conditions that make the treatment difficult, the authors stress the importance of a certain measure of eclecticism and careful planning to obtain effective and safe results. Only the combined, integrated efforts of the interventional radiologist and the surgeon can ensure rapid stabilisation of the bleeding and the desired improvement in survival.