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.

Typical values for pediatric interventional cardiology catheterizations: A standardized approach towards Diagnostic Reference Level establishment.

PURPOSE: To define weight-stratified Diagnostic Reference Levels (DRL) typical values for pediatric interventional cardiology (IC) procedures adopting standardized methodologies proposed by ICRP135 and RP185. METHODS: Procedures performed at the pediatric catheterization room of the University-Hospital of Padua were analysed. Patients were stratified into body weight (BW) classes and DRL quantities were analysed for the most performed procedures. Typical values are defined as median PKA and Ka,r. For database consistency, sampling and exclusion methods were precisely defined. The DRL-curve methodology by means of quantile regression median curves was investigated to assess the relationship between PKA and weight. A like-to-like comparison with literature was made. RESULTS: 385 procedures were analysed. A large PKA variability was observed in each weight group. PKA differences across BW groups were not always statistically significant. When stratifying by procedure, PKA variability decreased while correlations of PKA and PKA/FT with weight increased. The established typical values are generally lower than DRLs published data, whatever stratification method adopted. The highest PKA median values were observed for Angioplasty (4.9 and 11.6 Gycm2 for 5-<15 kg and 15-<30 kg, respectively). The DRL-curve approach shows promising results for Valvuloplasty and Angioplasty. CONCLUSIONS: Typical values for pediatric IC DRL quantities were determined according to ICRP135 and RP185 methodologies. Stratification by BW classification does not reduce the variability of the PKA values, unlike what happens when stratifying by procedure type. Results seem to corroborate that variability and exposure are more affected by procedure type and complexity than by patient weight. DRL-curve is a feasible approach.

Technical Note: An IBEX adaption toward image biomarker standardization.

PURPOSE: Interest in the field of radiomics is rapidly growing because of its potential to characterize tumor phenotype and provide predictive and prognostic information. Nevertheless, the reproducibility and robustness of radiomics studies are hampered by the lack of standardization in feature definition and calculation. In the context of the image biomarker standardization initiative (IBSI), we investigated the grade of compliance of the image biomarker explorer (IBEX), a free open-source radiomic software, and we developed and validated standardized-IBEX (S-IBEX), an adaptation of IBEX to IBSI. METHODS: Image biomarker explorer source code was checked against IBSI standard. Both the feature implementation and the overall image preprocessing chain were evaluated. Sections were re-implemented wherever differences emerged: in particular, contour-to-binary-mask conversion, image sub-portion extraction, re-segmentation, gray-level discretization and interpolation were aligned to IBSI. All reported IBSI features were implemented in S-IBEX. On a patient phantom, S-IBEX was validated by benchmarking five different preprocessing configurations proposed by IBSI. RESULTS: Most IBEX feature definitions are IBSI compliant; however, IBEX preprocessing introduces non-negligible nonconformities, resulting in feature values not aligned with the corresponding IBSI benchmarks. On the contrary, S-IBEX features are in agreement with the standard regardless of preprocessing configurations: the percentage of features equal to their benchmark values ranges from 98.1% to 99.5%, with overall maximum percentage error below 1%. Moreover, the impact of noncompliant preprocessing steps has been assessed: in these cases, the percentage of features equal to the standard drops below 35%. CONCLUSIONS: The use of standardized software for radiomic feature extraction is essential to ensure the reproducibility of results across different institutions, easing at the same time their external validation. This work presents and validates S-IBEX, a free IBSI-compliant software, developed upon IBEX, for feature extraction that is both easy to use and quantitatively accurate.

Monitoring skin dose changes during image-guided helical tomotherapy for head and neck cancer patients.

PURPOSE: An increase of skin dose during head and neck cancer (HNC) radiotherapy is potentially dangerous. Aim of this study was to quantify skin dose variation and to assess the need of planning adaptation (ART) to counteract it. METHODS: Planning CTs of 32 patients treated with helical tomotherapy (HT) according to a Simultaneous Integrated Boost (SIB) technique delivering 54/66 Gy in 30 fractions were deformably co-registered to MVCTs taken at fractions 15 and 30; in addition, the first fraction was also considered. The delivered dose-of-the-day was calculated on the corresponding deformed images. Superficial body layers (SL) were considered as a surrogate for skin, considering a layer thickness of 2 mm. Variations of SL DVH (∆SL) during therapy were quantified, focusing on ∆SL95% (i.e., 62.7 Gy). RESULTS: Small changes (within ± 1 cc for ∆SL95%) were seen in 15/32 patients. Only 2 patients experienced ∆SL95% > 1 cc in at least one of the two monitored fractions. Negative ∆SL95% > 1 cc (up to 17 cc) were much more common (15/32 patients). The trend of skin dose changes was mostly detected at the first fraction. Negative changes were correlated with the presence of any overlap between PTV and SL at planning and were explained in terms of how the planning system optimizes the PTV dose coverage near the skin. Acute toxicity was associated with planning DVH and this association was not improved if considering DVHs referring to fractions 15/30. CONCLUSION: About half of the patients treated with SIB with HT for HNC experienced a skin-sparing effect during therapy; only 6% experienced an increase. Our findings do not support skin-sparing ART, while suggesting the introduction of improved skin-sparing planning techniques.

Impact of acquisition count statistics reduction and SUV discretization on PET radiomic features in pediatric 18F-FDG-PET/MRI examinations.

PURPOSE: The evaluation of features robustness with respect to acquisition and post-processing parameter changes is fundamental for the reliability of radiomics studies. The aim of this study was to investigate the sensitivity of PET radiomic features to acquisition statistics reduction and standardized-uptake-volume (SUV) discretization in PET/MRI pediatric examinations. METHODS: Twenty-seven lesions were detected from the analysis of twenty-one 18F-FDG-PET/MRI pediatric examinations. By decreasing the count-statistics of the original list-mode data (3 MBq/kg), injected activity reduction was simulated. Two SUV discretization approaches were applied: 1) resampling lesion SUV range into fixed bins numbers (FBN); 2) rounding lesion SUV into fixed bin size (FBS). One hundred and six radiomic features were extracted. Intraclass Correlation Coefficient (ICC), Spearman correlation coefficient and coefficient-of-variation (COV) were calculated to assess feature reproducibility between low tracer activities and full tracer activity feature values. RESULTS: More than 70% of Shape and first order features, and around 70% and 40% of textural features, when using FBS and FBN methods respectively, resulted robust till 1.2 MBk/kg. Differences in median features reproducibility (ICC) between FBS and FBN datasets were statistically significant for every activity level independently from bin number/size, with higher values for FBS. Differences in median Spearman coefficient (i.e. patient ranking according to feature values) were not statistically significant, varying the intensity resolution (i.e. bin number/size) for either FBS and FBN methods. CONCLUSIONS: For each simulated count-statistic level, robust PET radiomic features were determined for pediatric PET/MRI examinations. A larger number of robust features were detected when using FBS methods.

Quantitative analysis of image metrics for reduced and standard dose pediatric 18F-FDG PET/MRI examinations.

OBJECTIVE:: The study performs a comprehensive analysis of image metrics to objectively support the reduction of injected activity in pediatric oncology 18F-FDG PET/MR (18F-fludeoxyglucose PET/MR) examinations. Contrast-to-Noise Ratio (CNR), Normalized Noise (NN), tumor burden, and standardized uptake value (SUV) parameters stability were investigated to robustly define the acceptable reduced activity level that preserves the clinical utility of images, considering different PET applications. METHODS:: 21 PET/MRI examinations performed on a 3-Tesla Biograph mMR scanner were analyzed. Tracer activity reduction was stimulated by decreasing the count statistics of the original list-mode data (3 MBq kg-1). In addition to the already studied SUV metrics and subjective scoring on lesion detectability, a thorough analysis of CNR, NN, Metabolic Tumor Volume (MTV), and Total Lesion Glycolysis (TLG) was performed. RESULTS:: SUVmax and SUVmean increased more than 5% only in 0.6 MBq kg-1 reconstructed images (+10% and +9%, respectively), while SUVpeak was almost unaffected (average variations < 2%). The quantified CNR, NN, MTV, and TLG behavior with the decrease of the injected activity clearly defines 1.5 MBq kg-1 as a threshold of activity after which the quality of the image degrades. Subjective and objective analyses yielded consistent results. All 56 lesions were detected until activity of 1.2 MBq kg-1, whereas five lesions were missed on the 0.6 MBq kg-1 image. Perceived image quality (IQ) decreased in Lower Tracer Activity (LTA) images but remained acceptable until 1.5 MBq kg-1. CONCLUSION:: Results about the stability of image metrics beyond the semi-quantitative SUV parameters and subjective analysis, rigorously proves the feasibility of the reduction of injected activity to 1.5 MBqkg-1 for pediatric patients aged between 7 and 17 years. ADVANCES IN KNOWLEDGE:: This is the first report on the quantitative evaluation of the effect of activity reduction on image quality in pediatric PET/MR. The findings offer objective corroboration to the feasibility of a significant dose reduction without consequences on clinical image reading and tumor burden metrics.

Skin dose calculation during radiotherapy of head and neck cancer using deformable image registration of planning and mega-voltage computed tomography scans.

BACKGROUND AND PURPOSE: Head-Neck (HN) patients may experience severe acute skin complications that can cause treatment interruption and increase the risk of late fibrosis. This study assessed a method for accurately monitoring skin dose changes during helical tomotherapy for HN cancer based on deformable image registration of planning computed tomography (CT) and mega-voltage CT (MVCT). MATERIALS AND METHODS: Planning CTs of nine patients were deformably registered to mid-treatment MVCT (MV15) images resulting in CTdef images. The original plans were recalculated on both CTdef and mid-treatment kilo-voltage CT (CT15) taken as ground truth. Superficial layers (SL) of the body with thicknesses of 2, 3 and 5 mm (SL2, SL3, SL5) were considered as derma surrogates. SL V95%, V97%, V98%, V100%, V102%, V105% and V107% of the prescribed PTV dose were extracted for CT15/CTdef and compared (considering patients with skin dose > 95%). For comparison, doses were calculated directly on the calibrated MVCT and analyzed in the same way. RESULTS: Differences between SL2/SL3/SL5 V95%-V107% in CT15/CTdef were very small: for eight of nine patients the difference between the considered SL2 Vd% computed on CTdef and CT15 was less than 1.4 cm3 for all d%. A larger value was found when using MVCT for skin dose calculation (4.8 cm3 for SL2), although CTdef body contour matched CT15 body with accuracy similar to that of MV15. CONCLUSIONS: Deforming the planning CT-to-MVCT was shown to be accurate considering external body contours and skin DVHs. The method was able to accurately identify superficial overdosing.

Organ doses and lifetime attributable risk evaluations for scoliosis examinations of adolescent patients with the EOS imaging system.

PURPOSE: The goal of this work is to evaluate organ doses and lifetime attributable risk of cancer incidence and mortality in scoliosis examinations of adolescent patients performed with EOS imaging system, in order to optimize patient dose and protocols. METHODS: An anthropomorphic phantom of a normal patient, with thermoluminescent dosimeters in correspondence with the main organs at risk, was imaged with both EOS and computed radiography (CR). For each modality, effective dose was calculated from the measured organ doses. Lifetime attributable risk was computed accordingly to the Committee on the Biological Effects of Ionizing Radiation (BEIR VII) and Public Health England (HPA) publications. RESULTS: Except for testes and eyes, which were excluded from the scan in CR protocol, for all the other organs the doses delivered with CR examination were higher than these delivered by EOS system. The effective dose in EOS examination (0.43 ± 0.04 mSv) is about two times less than the dose in computed radiography with anti-scatter grid examination (0.87 ± 0.09 mSv), and, consequently, also the cancer probability is lower (5.4 vs 9.7 number of any cancers induction cases per 100,000 person examined, for a 20-year-old male patient). CONCLUSIONS: The EOS system is efficient in limiting patient dose. The shielding of testes and the exclusion of eyes from the scan could allow to further reduce the dose.