•Interest has grown in texture analysis and machine learning in medical imaging.•Computerized systems play a growing role in thyroid nodule characterization.•Quantitative image analysis can improve ...diagnostic accuracy of thyroid nodules.•Methodological issues need to be solved in texture analysis and machine learning.•Quantitative image analysis should be standardized and validated on large scale.
In thyroid imaging, “texture” refers to the echographic appearence of the parenchyma or a nodule. However, definition of the image characteristics is operator dependent and influenced by the operator’s experience. In a more objective texture analysis, a variety of mathematical methods are used to describe image inhomogeneity, allowing assessment of an image by means of quantitative parameters. Moreover, this approach may be used to develop an efficient computer-aided diagnosis (CAD) system to yield a second opinion when differentiating malignant and benign thyroid lesions. The aim of this review is to summarize the available literature data on texture analysis, with and without CAD, in patients with suspected thyroid nodules or differentiated thyroid cancer, and to assess the current state of the approach.
Purpose
Radiomic features derived from the texture analysis of different imaging modalities e show promise in lesion characterisation, response prediction, and prognostication in lung cancer ...patients. The present study aimed to identify an images-based radiomic signature capable of predicting disease-free survival (DFS) in non-small cell lung cancer (NSCLC) patients undergoing surgery.
Methods
A cohort of 295 patients was selected. Clinical parameters (age, sex, histological type, tumour grade, and stage) were recorded for all patients. The endpoint of this study was DFS. Both computed tomography (CT) and fluorodeoxyglucose positron emission tomography (PET) images generated from the PET/CT scanner were analysed. Textural features were calculated using the LifeX package. Statistical analysis was performed using the R platform. The datasets were separated into two cohorts by random selection to perform training and validation of the statistical models. Predictors were fed into a multivariate Cox proportional hazard regression model and the receiver operating characteristic (ROC) curve as well as the corresponding area under the curve (AUC) were computed for each model built.
Results
The Cox models that included radiomic features for the CT, the PET, and the PET+CT images resulted in an AUC of 0.75 (95%CI: 0.65–0.85), 0.68 (95%CI: 0.57–0.80), and 0.68 (95%CI: 0.58–0.74), respectively. The addition of clinical predictors to the Cox models resulted in an AUC of 0.61 (95%CI: 0.51–0.69), 0.64 (95%CI: 0.53–0.75), and 0.65 (95%CI: 0.50–0.72) for the CT, the PET, and the PET+CT images, respectively.
Conclusions
A radiomic signature, for either CT, PET, or PET/CT images, has been identified and validated for the prediction of disease-free survival in patients with non-small cell lung cancer treated by surgery.
Purpose
To evaluate the ability of CT and PET radiomics features to classify lung lesions as primary or metastatic, and secondly to differentiate histological subtypes of primary lung cancers.
...Methods
A cohort of 534 patients with lung lesions were retrospectively studied. Radiomics texture features were extracted using the LIFEx package from semiautomatically segmented PET and CT images. Histology data were recorded in all patients. The patient cohort was divided into a training and a validation group and linear discriminant analysis (LDA) was performed to classify the lesions using both direct and backward stepwise methods. The robustness of the procedure was tested by repeating the entire process 100 times with different assignments to the training and validation groups. Scoring metrics included analysis of the receiver operating characteristic curves in terms of area under the curve (AUC), sensitivity, specificity and accuracy.
Results
Radiomics features extracted from CT and PET datasets were able to differentiate primary tumours from metastases in both the training and the validation group (AUCs 0.79 ± 0.03 and 0.70 ± 0.04, respectively, from the CT dataset; AUCs 0.92 ± 0.01 and 0.91 ± 0.03, respectively, from the PET dataset). The AUC cut-off thresholds identified by LDA using direct and backward elimination strategies were −0.79 ± 0.06 and −0.81 ± 0.08, respectively (CT dataset) and −0.69 ± 0.05 and −0.68 ± 0.04, respectively (PET dataset). For differentiation between primary subgroups based on CT features, the AUCs in the training and validation groups were 0.81 ± 0.02 and 0.69 ± 0.04 for adenocarcinoma (Adc) vs. squamous cell carcinoma (Sqc) or “Other”, 0.85 ± 0.02 and 0.70 ± 0.05 for Sqc vs. Adc or Other, and 0.77 ± 0.03 and 0.57 ± 0.05 for Other vs. Adc or Sqc. The same analyses for the PET data revealed AUCs of 0.90 ± 0.10 and 0.80 ± 0.04, 0.80 ± 0.02 and 0.61 ± 0.06, and 0.97 ± 0.01 and 0.88 ± 0.04, respectively.
Conclusion
PET radiomics features were able to differentiate between primary and metastatic lung lesions and showed the potential to identify primary lung cancer subtypes.
Objectives
We aimed to assess the ability of radiomics, applied to not-enhanced computed tomography (CT), to differentiate mediastinal masses as thymic neoplasms
vs
lymphomas.
Methods
The present ...study was an observational retrospective trial. Inclusion criteria were pathology-proven thymic neoplasia or lymphoma with mediastinal localization, availability of CT. Exclusion criteria were age < 16 years and mediastinal lymphoma lesion < 4 cm. We selected 108 patients (
M
:
F
= 47:61, median age 48 years, range 17–79) and divided them into a training and a validation group. Radiomic features were used as predictors in linear discriminant analysis. We built different radiomic models considering segmentation software and resampling setting. Clinical variables were used as predictors to build a clinical model. Scoring metrics included sensitivity, specificity, accuracy and area under the curve (AUC). Wilcoxon paired test was used to compare the AUCs.
Results
Fifty-five patients were affected by thymic neoplasia and 53 by lymphoma. In the validation analysis, the best radiomics model sensitivity, specificity, accuracy and AUC resulted 76.2 ± 7.0, 77.8 ± 5.5, 76.9 ± 6.0 and 0.84 ± 0.06, respectively. In the validation analysis of the clinical model, the same metrics resulted 95.2 ± 7.0, 88.9 ± 8.9, 92.3 ± 8.5 and 0.98 ± 0.07, respectively. The AUCs of the best radiomic and the clinical model not differed.
Conclusions
We developed and validated a CT-based radiomic model able to differentiate mediastinal masses on non-contrast-enhanced images, as thymic neoplasms or lymphoma. The proposed method was not affected by image postprocessing. Therefore, the present image-derived method has the potential to noninvasively support diagnosis in patients with prevascular mediastinal masses with major impact on management of asymptomatic cases.
Highlights • The literature regarding the lung SBRT dose calculation algorithm accuracy is reviewed. • The small field and low density problems are analysed. • A summarizing example focuses on the ...main differences and aspects of the main three classes of dose calculation algorithms.
Purpose
To assess the role of radiomics parameters in predicting pathological complete response (pCR) to neoadjuvant chemotherapy (NAC) in patients with locally advanced breast cancer.
Methods
...Seventy-nine patients who had undergone pretreatment staging
18
F-FDG PET/CT and treatment with NAC between January 2010 and January 2018 were included in the study. Primary lesions on PET images were delineated, and extraction of first-, second-, and higher-order imaging features was performed using LIFEx software. The relationship between these parameters and pCR to NAC was analyzed by multiple logistic regression models.
Results
Nineteen patients (24%) had pCR to NAC. Different models were generated on complete information and imputed datasets, using univariable and multivariable logistic regression and least absolute shrinkage and selection operator (lasso) regression. All models could predict pCR to NAC, with area under the curve values ranging from 0.70 to 0.73. All models agreed that tumor molecular subtype is the primary predictor of the primary endpoint.
Conclusions
Our models predicted that patients with subtype 2 and subtype 3 (HER2+ and triple negative, respectively) are more likely to have a pCR to NAC than those with subtype 1 (luminal). The association between PET imaging features and pCR suggested that PET imaging features could be considered as potential predictors of pCR in locally advanced breast cancer patients.
To evaluate the feasibility of high-dose stereotactic body radiation therapy (SBRT) in the treatment of unresectable liver metastases.
Patients with 1 to 3 liver metastases, with maximum individual ...tumor diameters less than 6 cm and a Karnofsky Performance Status of at least 70, were enrolled and treated by SBRT on a phase 2 clinical trial. Dose prescription was 75 Gy on 3 consecutive days. SBRT was delivered using the volumetric modulated arc therapy by RapidArc (Varian, Palo Alto, CA) technique. The primary end-point was in-field local control. Secondary end-points were toxicity and survival.
Between February 2010 and September 2011, a total of 61 patients with 76 lesions were treated. Among the patients, 21 (34.3%) had stable extrahepatic disease at study entry. The most frequent primary sites were colorectal (45.9%) and breast (18%). Of the patients, 78.7% had 1 lesion, 18.0% had 2 lesions, and 3.3% had 3 lesions. After a median of 12 months (range, 2-26 months), the in-field local response rate was 94%. The median overall survival rate was 19 months, and actuarial survival at 12 months was 83.5%. None of the patients experienced grade 3 or higher acute toxicity. No radiation-induced liver disease was detected. One patient experienced G3 late toxicity at 6 months, resulting from chest wall pain.
SBRT for unresectable liver metastases can be considered an effective, safe, and noninvasive therapeutic option, with excellent rates of local control and a low treatment-related toxicity.
To assess the clinical impact of the Acuros XB algorithm (implemented in the Varian Eclipse treatment-planning system) in non-small-cell lung cancer (NSCLC) cases.
A CT dataset of 10 patients ...presenting with advanced NSCLC was selected and contoured for planning target volume, lungs, heart, and spinal cord. Plans were created for 6-MV and 15-MV beams using three-dimensional conformal therapy, intensity-modulated therapy, and volumetric modulated arc therapy with RapidArc. Calculations were performed with Acuros XB and the Anisotropic Analytical Algorithm. To distinguish between differences coming from the different heterogeneity management and those coming from the algorithm and its implementation, all the plans were recalculated assigning Hounsfield Unit (HU) = 0 (Water) to the CT dataset.
Differences in dose distributions between the two algorithms calculated in Water were <0.5%. This suggests that the differences in the real CT dataset can be ascribed mainly to the different heterogeneity management, which is proven to be more accurate in the Acuros XB calculations. The planning target dose difference was stratified between the target in soft tissue, where the mean dose was found to be lower for Acuros XB, with a range of 0.4% ± 0.6% (intensity-modulated therapy, 6 MV) to 1.7% ± 0.2% (three-dimensional conformal therapy, 6 MV), and the target in lung tissue, where the mean dose was higher for 6 MV (from 0.2% ± 0.2% to 1.2% ± 0.5%) and lower for 15 MV (from 0.5% ± 0.5% to 2.0% ± 0.9%). Mean doses to organs at risk presented differences up to 3% of the mean structure dose in the worst case. No particular or systematic differences were found related to the various modalities. Calculation time ratios between calculation time for Acuros XB and the Anisotropic Analytical Algorithm were 7 for three-dimensional conformal therapy, 5 for intensity-modulated therapy, and 0.2 for volumetric modulated arc therapy with RapidArc.
The availability of Acuros XB could improve patient dose estimation, increasing the data consistency of clinical trials.
A study was realised to evaluate and determine relative figures of merit of a new algorithm for photon dose calculation when applied to inhomogeneous media.
The new Acuros XB algorithm implemented in ...the Varian Eclipse treatment planning system was compared against a Monte Carlo method (VMC++), and the Analytical Anisotropic Algorithm (AAA). The study was carried out in virtual phantoms characterized by simple geometrical structures. An insert of different material and density was included in a phantom built of skeletal-muscle and HU = 0 (setting "A"): Normal Lung (lung, 0.198 g/cm3); Light Lung (lung, 0.035 g/cm3); Bone (bone, 1.798 g/cm3); another phantom (setting "B") was built of adipose material and including thin layers of bone (1.85 g/cm3), adipose (0.92 g/cm3), cartilage (1.4745 g/cm3), air (0.0012 g/cm3). Investigations were performed for 6 and 15 MV photon beams, and for a large (13 × 13 cm2) and a small (2.8 × 13 cm2) field.
Results are provided in terms of depth dose curves, transverse profiles and Gamma analysis (3 mm/3% and 2 mm/2% distance to agreement/dose difference criteria) in planes parallel to the beam central axis; Monte Carlo simulations were assumed as reference. Acuros XB gave an average gamma agreement, with a 3 mm/3% criteria, of 100%, 86% and 100% for Normal Lung, Light Lung and Bone settings, respectively, and dose to medium calculations. The same figures were 86%, 11% and 100% for AAA, where only dose rescaled to water calculations are possible.
In conclusion, Acuros XB algorithm provides a valid and accurate alternative to Monte Carlo calculations for heterogeneity management.
Objectives
To evaluate the feasibility and efficacy of stereotactic body radiation therapy (SBRT) in the treatment of hepatocellular carcinoma (HCC) unsuitable for standard loco-regional therapies.
...Materials and methods
Patients with 1–3 inoperable HCC lesions with diameter ≤6 cm were treated by SBRT. According to lesions size and liver function, two prescription regimens were adopted: 48–75 Gy in three fractions or 36–60 Gy in six fractions. SBRT was delivered using the volumetric modulated arc therapy technique with flattening filter-free photon beams. The primary end points of this study were in-field local control (LC) and toxicity. Secondary end points were overall survival (OS) and progression-free survival (PFS).
Results
Forty-three patients with 63 HCC lesions were irradiated. All patients had Child–Turcotte–Pugh class A or B disease. Thirty lesions (48 %) were treated with 48–75 Gy in three consecutive fractions, and 33 (52 %) received 36–60 Gy in six fractions. Median follow-up was 8 months (range 3–43 months). Actuarial LC at 6, 12 and 24 months was 94.2 ± 3.3, 85.8 ± 5.5 and 64.4 ± 11.5 %, respectively. A biological equivalent dose (BED) >100 Gy and GTV size were significant prognostic factors for LC in univariate analysis (
p
< 0.001 and
p
< 0.02). Median OS was 18.0 ± 5.8 months. Actuarial OS at 6, 12 and 24 months was 91.1 ± 4.9, 77.9 ± 8.2 and 45.3 ± 14.0 %, respectively. Univariate analysis showed that OS is correlated with LC (
p
< 0.04), BED >100 (
p
< 0.05) and cumulative gross tumor volume GTV <5 cm (
p
< 0.04). Median PFS was 8 months, with a 1-year PFS rate of 41 %. A significant (≥grade 3) toxicity was observed in seven patients (16 %) 2–6 months after the completion of the treatment. No classic radiation-induced liver disease was observed.
Conclusion
Stereotactic body radiation therapy is a safe and effective therapeutic option for HCC lesions unsuitable to standard loco-regional therapies, with acceptable local control rates and low treatment-related toxicity. The significant correlation between LC and higher doses and between LC and OS supports the clinical value of SBRT in these patients.