Reproducibility in Radiomics: A Comparison of Feature Extraction Methods and Two Independent Datasets.

Radiomics involves the extraction of information from medical images that are not visible to the human eye. There is evidence that these features can be used for treatment stratification and outcome prediction. However, there is much discussion about the reproducibility of results between different studies. This paper studies the reproducibility of CT texture features used in radiomics, comparing two feature extraction implementations, namely the MATLAB toolkit and Pyradiomics, when applied to independent datasets of CT scans of patients: (i) the open access RIDER dataset containing a set of repeat CT scans taken 15 min apart for 31 patients (RIDER Scan 1 and Scan 2, respectively) treated for lung cancer; and (ii) the open access HN1 dataset containing 137 patients treated for head and neck cancer. Gross tumor volume (GTV), manually outlined by an experienced observer available on both datasets, was used. The 43 common radiomics features available in MATLAB and Pyradiomics were calculated using two intensity-level quantization methods with and without an intensity threshold. Cases were ranked for each feature for all combinations of quantization parameters, and the Spearman's rank coefficient, rs, calculated. Reproducibility was defined when a highly correlated feature in the RIDER dataset also correlated highly in the HN1 dataset, and vice versa. A total of 29 out of the 43 reported stable features were found to be highly reproducible between MATLAB and Pyradiomics implementations, having a consistently high correlation in rank ordering for RIDER Scan 1 and RIDER Scan 2 (rs > 0.8). 18/43 reported features were common in the RIDER and HN1 datasets, suggesting they may be agnostic to disease site. Useful radiomics features should be selected based on reproducibility. This study identified a set of features that meet this requirement and validated the methodology for evaluating reproducibility between datasets.

Multi-centre radiomics for prediction of recurrence following radical radiotherapy for head and neck cancers: Consequences of feature selection, machine learning classifiers and batch-effect harmonization.

Background and purpose: Radiomics models trained with limited single institution data are often not reproducible and generalisable. We developed radiomics models that predict loco-regional recurrence within two years of radiotherapy with private and public datasets and their combinations, to simulate small and multi-institutional studies and study the responsiveness of the models to feature selection, machine learning algorithms, centre-effect harmonization and increased dataset sizes. Materials and methods: 562 patients histologically confirmed and treated for locally advanced head-and-neck cancer (LA-HNC) from two public and two private datasets; one private dataset exclusively reserved for validation. Clinical contours of primary tumours were not recontoured and were used for Pyradiomics based feature extraction. ComBat harmonization was applied, and LASSO-Logistic Regression (LR) and Support Vector Machine (SVM) models were built. 95% confidence interval (CI) of 1000 bootstrapped area-under-the-Receiver-operating-curves (AUC) provided predictive performance. Responsiveness of the models' performance to the choice of feature selection methods, ComBat harmonization, machine learning classifier, single and pooled data was evaluated. Results: LASSO and SelectKBest selected 14 and 16 features, respectively; three were overlapping. Without ComBat, the LR and SVM models for three institutional data showed AUCs (CI) of 0.513 (0.481-0.559) and 0.632 (0.586-0.665), respectively. Performances following ComBat revealed AUCs of 0.559 (0.536-0.590) and 0.662 (0.606-0.690), respectively. Compared to single cohort AUCs (0.562-0.629), SVM models from pooled data performed significantly better at AUC = 0.680. Conclusions: Multi-institutional retrospective data accentuates the existing variabilities that affect radiomics. Carefully designed prospective, multi-institutional studies and data sharing are necessary for clinically relevant head-and-neck cancer prognostication models.

Determining the Occurrence of Hypothyroidism Following Treatment With Radiation Therapy in Head and Neck Carcinoma Patients and the Associated Role of Risk Factors and Dose-Volume Histograms: A Prospective Study.

Background Head and neck carcinomas are one of the most common malignancies in developing countries including India. Most patients are treated with radiotherapy. Although post-radiotherapy hypothyroidism is a known complication, data regarding its incidence and factors influencing it are scarce. This study aimed to determine the incidence of post-radiotherapy hypothyroidism in head and neck carcinoma patients treated with radiotherapy and the factors influencing it. Methodology Patients with head and neck carcinomas treated with radiotherapy as one of the modalities were included in this study. Thyroid function tests were done, and quality of life questionnaires were completed before treatment and during follow-up. Dose-volume histogram (DVH), demographic data, and disease-related parameters were compared. Results Out of the 95 patients screened, 14 were found to be hypothyroid prior to the commencement of radiotherapy and were excluded. With a median follow-up duration of 34 weeks, 29.6% developed hypothyroidism, with 19% developing it in the first year. On univariate and multivariate analysis of the DVH of the thyroid gland, volume receiving 50 Gy (V50), dose received to 50% volume (D50), and the mean dose (more than 50 Gy) were found to be significantly associated with hypothyroidism. Conclusions Hypothyroidism is a significant comorbid factor in Indian patients with head and neck carcinomas. The incidence of post-radiotherapy hypothyroidism is significant and occurs early compared to the western population leading to significant deterioration in the quality of life. Parameters such as the volume of the thyroid gland, V50, D50, and mean dose to the thyroid gland influence the incidence of hypothyroidism. The use of appropriate constraints can significantly prevent radiotherapy-induced hypothyroidism.

Framework for Machine Learning of CT and PET Radiomics to Predict Local Failure after Radiotherapy in Locally Advanced Head and Neck Cancers.

CONTEXT: Cancer Radiomics is an emerging field in medical imaging and refers to the process of converting routine radiological images that are typically qualitatively interpreted to quantifiable descriptions of the tumor phenotypes and when combined with statistical analytics can improve the accuracy of clinical outcome prediction models. However, to understand the radiomic features and their correlation to molecular changes in the tumor, first, there is a need for the development of robust image analysis methods, software tools and statistical prediction models which is often limited in low- and middle-income countries (LMIC). AIMS: The aim is to build a framework for machine learning of radiomic features of planning computed tomography (CT) and positron emission tomography (PET) using open source radiomics and data analytics platforms to make it widely accessible to clinical groups. The framework is tested in a small cohort to predict local disease failure following radiation treatment for head-and-neck cancer (HNC). The predictors were also compared with the existing Aerts HNC radiomics signature. SETTINGS AND DESIGN: Retrospective analysis of patients with locally advanced HNC between 2017 and 2018 and 31 patients with both pre- and post-radiation CT and evaluation PET were selected. SUBJECTS AND METHODS: Tumor volumes were delineated on baseline PET using the semi-automatic adaptive-threshold algorithm and propagated to CT; PyRadiomics features (total of 110 under shape/intensity/texture classes) were extracted. Two feature-selection methods were tested for model stability. Models were built based on least absolute shrinkage and selection operator-logistic and Ridge regression of the top pretreatment radiomic features and compared to Aerts' HNC-signature. Average model performance across all internal validation test folds was summarized by the area under the receiver operator curve (ROC). RESULTS: Both feature selection methods selected CT features MCC (GLCM), SumEntropy (GLCM) and Sphericity (Shape) that could predict the binary failure status in the cross-validated group and achieved an AUC >0.7. However, models using Aerts' signature features (Energy, Compactness, GLRLM-GrayLevelNonUniformity and GrayLevelNonUniformity-HLH wavelet) could not achieve a clear separation between outcomes (AUC = 0.51-0.54). CONCLUSIONS: Radiomics pipeline included open-source workflows which makes it adoptable in LMIC countries. Additional independent validation of data is crucial for the implementation of radiomic models for clinical risk stratification.

Patterns of Recurrence in Locally Advanced Resectable Oesophageal Carcinoma: Retrospective Review from a Tertiary Cancer Centre in South India.

PURPOSE: The study aims to analyse patterns of recurrence following neoadjuvant treatment and surgery in carcinoma oesophagus with an intent to postulate optimal nodal radiation. METHODOLOGY: A retrospective review of patients who presented to our centre within a 5-year period (2014-2018), with recurrence following sequential neoadjuvant treatment and radical surgery, was conducted in this single-institution study. The patterns of recurrence and duration of disease-free survival were analysed. RESULTS: Twenty-one patients (14 men, 7 women) presented with recurrence, of which 13, 7, and 1 patient(s) had received NACT, NACTRT, or both, respectively. Six patients who did not receive neoadjuvant radiotherapy received adjuvant RT. Among the 10 patients who had nodal recurrence after RT (either neoadjuvant or adjuvant), 6 and 4 patients had in-field and out-of-field nodal recurrences, respectively-the latter were equally distributed within 5 cm and outside 5 cm of the PTV margin. CONCLUSION: Among the patients who presented with recurrence, more than half had not received neoadjuvant RT (treated in the 'pre-CROSS era' or due to long-segment disease), reasserting the therapeutic superiority of NACTRT. Increased regularity of recurrences in the draining nodal region was not noted in this study, but large-scale, prospective, randomised head-to-head comparative trials to determine optimal nodal irradiation in carcinoma oesophagus are required.

Effects of radiotherapy on olfaction and nasal function in head and neck cancer patients.

PURPOSE: Radiotherapy (RT) is a major component of treatment in head and neck malignancies and often the radiation field includes the nasal cavity and olfactory cleft region. We aimed to assess olfaction, mucociliary clearance time and quality of life (QOL) before RT and during the course of radiotherapy. METHODS: This prospective, observational, cohort study was conducted over a period of 1 year. The olfactory function, mucociliary clearance and QOL of patients with primary head and neck cancers undergoing radiation therapy as part of treatment were assessed prior to radiotherapy and followed up serially up to 3 months after radiotherapy. A total of 21 patients were enrolled. Assessment was done using noninvasive tests for better compliance and ease of examination. RESULTS: Among the 21 patients recruited, 18 completed radiotherapy and 13 were assessed 3 months post radiotherapy. Mean olfactory scores (including olfactory threshold and odor identification), using Connecticut Chemosensory Clinical Research Center (CCCRC) test, deteriorated significantly at the end of radiotherapy (p < 0.001) as compared to scores before irradiation. Subjective assessment of olfaction by Appetite, Hunger and Sensory perception (AHSP) questionnaire did not demonstrate significant impairment in nasal function (p < 0.319) although overall QOL significantly deteriorated (p 0.004). The mucociliary clearance time was prolonged in 72% of the patients at the end of radiotherapy. CONCLUSION: Deterioration in olfactory function was found to occur during the course of radiotherapy with gradual improvement after 3 months. However, patients did not notice olfactory dysfunction subjectively. Mucociliary dysfunction persisted even after 3 months following radiation.

Management of Locally Advanced Carcinoma Oesophagus with Radiation/Chemoradiation: Single Institute Experience.

PURPOSE: Oesophageal malignancy is the fourth commonest cancer which has a very poor outcome. In the management of oesophageal malignancy, radiation therapy is underutilized by many physicians due to feared complications though its role is well proven in literature. This study brings the role of radiation therapy in terms of loco-regional and distant relapse. This study characterizes demographic characteristics, treatment modality and the impact of these on recurrence-free survival of locally advanced carcinoma oesophagus. METHOD: In a retrospective analysis over a period of 24 months, 28 patients diagnosed to have locally advanced oesophageal malignancy were included. Neoadjuvant chemotherapy was carried out in 4 patients and concurrent chemoradiotherapy in 18 patients. Among the 28 patients, 7 patients who were operable and fit underwent surgery 6-8 weeks after completion of the planned neoadjuvant therapy. All patients received concurrent chemoradiation for a period of 5 weeks with platinum and/or taxane-based chemotherapy. RESULTS: With a median follow-up of 12 months, loco-regional recurrence and distant relapse were substantially less with 14 and 10 %, respectively. The Kaplan-Meier recurrence-free survival was 60 % at 24 months. All patients who had surgery as a part of the multimodality management were disease free at the last follow-up. CONCLUSION: In locally advanced oesophageal malignancy, radiation therapy plays an important role in downsizing the tumour for operability or can also be utilized as the sole modality of management.

Role of conventional and diffusion weighted MRI in predicting treatment response after low dose radiation and chemotherapy in locally advanced carcinoma cervix.

BACKGROUND AND PURPOSE: To assess the diagnostic performance of conventional and diffusion weighted (DWI) magnetic resonance imaging (MRI) in predicting response in locally advanced cervical cancer. MATERIALS AND METHODS: Total 24 patients with stage IIB-IIIB squamous cell carcinoma cervix were treated with initial two cycles of paclitaxel and carboplatin and concurrent low dose radiotherapy prior to standard chemoradiation. Response was assessed clinically and radiologically after 3 weeks of initial treatment. Volumetric and functional parameters derived from conventional and diffusion weighted MRI, due to treatment were measured. RESULTS: Significant reduction of GTV was noted in MRI (54 cm(3) vs. 11 cm(3), p < 0.01) and DWI (44 cm(3) vs. 6 cm(3), p < 0.01, ΔADC = 0.49 × 10(-3)mm(2)/sec, p < 0.01) after treatment. Tumor volume reduction rate (TVRR) in DWI was significantly higher in pathological good responders (p = 0.03). In this group both mean post treatment apparent diffusion coefficient (ADC) value and ΔADC were significantly higher (p = 0.01 and p = 0.03). ADC was a good predictor for pathological response (area under receiver operating characteristic curve (ROC) 0.814). CONCLUSION: TVRR (DWI) and ΔADC can be used as a predictor of early pathological response. Complete response based on DWI, could be a useful predictor of long term disease control.

Radiation Pneumonitis After Conventional Radiotherapy For Breast Cancer: A Prospective Study.

BACKGROUND: Loco-regional radiotherapy is an important treatment modality in breast cancer and radiation pneumonitis (RP) is one of the early toxicities. AIM: To study the occurrence, correlation of RP with patient and radiotherapy related factors and the effects on pulmonary function following conventional radiotherapy in breast cancer. SETTINGS AND DESIGN: Prospective study, from a tertiary hospital in a developing country. MATERIALS AND METHODS: Prospective analysis of clinical symptoms, pulmonary function and radiologic changes was done prior to and 12 weeks after adjuvant radiotherapy (n=46). Statistical analysis was done using SPSS version 10 software. RESULTS: Radiological and clinical RP was seen in 45.65% (n=21) and 19.56% (n=9) respectively. RP was significantly higher with age >50 years (OR 4.4), chest wall irradiation with electrons, (electrons 83.3% vs cobalt60 32.4%, p=0.02) and supraclavicular field treatment with 6 MV photons (p= 0.011). There was significant relationship between Inferior Lung Distance (ILD) and RP (p=0.013). The fall in Total Lung Capacity (TLC) was significantly more in those with RP (p=0.02). CONCLUSION: Clinical RP occurs in almost one-fifth of breast cancer patients treated with conventional radiotherapy. Chest wall irradiation with electrons, supraclavicular field irradiation with 6 MV photons, higher ILD and age >50 years was associated with increased RP. The pulmonary function parameter most affected was TLC. The factors associated with increased RP should be considered when adjuvant radiotherapy is planned to minimize its likelihood and intervene appropriately.

Low-dose fractionated radiation and chemotherapy prior to definitive chemoradiation in locally advanced carcinoma of the uterine cervix: Results of a prospective phase II clinical trial.

BACKGROUND: We investigated the feasibility of neoadjuvant low-dose radiation and chemotherapy with paclitaxel and carboplatin (LDCRT) before radical chemoradiation (CRT) and assessed the feasibility, efficacy, and response rate to such a regimen. METHODS: This is a single-arm phase II trial of 24 patients, with locally advanced squamous cell carcinoma of the cervix (stage IIB-IIIB). Patients received low-dose fractionated radiotherapy, carboplatin (AUC×5) and paclitaxel (175 mg/m(2)), three weekly for two cycles followed by CRT. The primary end point was overall and disease-free survival. RESULTS: Mean age of the patients at diagnosis was 50 years; Radiological complete or partial response rate was 40% and 60%, respectively, post-LDCRT. The median follow-up was 30 months (24-36 months). Both overall and progression-free survivals at 2.5 years were 84%. Grade 3/4 toxicities were 24% hematological toxicity during LDCRT and 46% during CRT (hematological: 42%, non-hematological: 4%). CONCLUSION: A good response rate is achieved by low-dose radiation and chemotherapy with carboplatin and paclitaxel followed by radical CRT. This treatment regimen is feasible and effective as evidenced by the acceptable toxicity and 84% local control at 2.5 years.

Conventional margins not sufficient for post-prostatectomy prostate bed coverage: an analysis of 477 cone-beam computed tomography scans.

PURPOSE: To study prostate bed deformation, and compare coverage by 5 mm and 10mm posterior expansion PTV margins. METHOD: Fifty patients who completed post-prostatectomy radiotherapy had two expansion margins applied to the planning CT CTV: PTV10 (10 mm isometrically) and PTV5 (5 mm posteriorly, 10 mm all other directions). The CTV was then contoured on 477 pre-treatment CBCTs, and PTV5 and PTV10 coverage of each CBCT-CTVs was assessed. The maximum distance from the planning CT CTV to the combined CTV of all CBCTs including the planning CT CTV was measured for the superior part of the prostate bed, and the inferior part of the prostate bed, for every patient. RESULTS: The mean difference between largest and smallest CBCT-CTVs per patient was 18.7 cm(3) (range 6.3-34.2 cm(3)). Out of 477 CBCTs, there were 43 anterior geometric geographical misses for either PTV with a mean volume of 2.25 cm(3) (range 0.01-18.88 cm(3)). For PTV10, there were 26 posterior geometric geographical misses with a mean volume of 1.37 cm(3) (0.01-11.02 cm(3)). For PTV5, there were 46 posterior geometric geographical misses with a mean volume of 3.22 cm(3) (0.01-19.82 cm(3)). The maximum edge-to-edge distance for the superior prostate bed was anterior 19 mm, posterior 16 mm, left and right 7 mm. The maximum edge-to-edge distance for the inferior prostate bed was anterior 4mm, posterior 12 mm, left and right 7 mm. CONCLUSION: This study supports differential margins for the superior and inferior portions of the prostate bed. Because of the large deformation of CTV volume seen, adaptive radiotherapy solutions should be investigated further.

Hypofractionated Palliative Radiotherapy in Locally Advanced Inoperable Head and Neck Cancer: CMC Vellore Experience.

BACKGROUND: A novel, short duration, palliative radiotherapy schedule for inoperable head and neck cancer was evaluated in terms of palliation of cancer-related symptoms and acute toxicities. MATERIALS AND METHODS: Thirty-six patients with inoperable head and neck cancer were included in the study (2010-2012). All patients received 40 Gy in 10 fractions (equivalent dose: 49.8 Gy in conventional fractionation) with 2 fractions per week. Treatment-related toxicity was assessed using Radiation Therapy Oncology Group criteria. Functional Assessment of Cancer Therapy (Head and Neck, FACT H and N) quality of life (QOL) tool was administered before starting and at the completion of radiotherapy. Mean value before and after treatment was compared (paired t-test, P = 0.05, two-tailed for significance). RESULTS: Thirty-three patients (male: 29, female: 4, mean age: 57.8 ± 9.7 years) were included in the analysis (three patients discontinued treatment due to socioeconomic reasons). All patients had advanced inoperable head and neck cancers (27% IVA, 61% IVB, 9% IVC, TNM stage and 3% recurrent disease). Distressing pain at primary site (42%), dysphagia (18%), neck swelling (30%), and hoarseness (10%) were common presentations. Incidence of grade III mucositis and dermatitis and pain was 18%, 3%, and 24%, respectively. Planned radiotherapy without any interruptions was completed by 73% patients. QOL assessment showed improvement in social well-being (17.4 vs. 20.01, P = 0.03), but no significant change was observed in head and neck specific score (25.1 vs. 25.0, P = NS) after treatment. Reduction of pain was observed in 88% patients and 60% patients had improvement of performance status. Median overall survival of the cohort was 7 months. CONCLUSIONS: The study shows that this short duration palliative radiotherapy schedule is a clinically viable option for advanced inoperable head and neck cancer to achieve significant palliation of the main presenting symptoms like pain, dysphagia, and throat pain.

Nanotechnology in oncology: Characterization and in vitro release kinetics of cisplatin-loaded albumin nanoparticles: Implications in anticancer drug delivery.

CONTEXT: Nanotechnology is an empowering technology that holds promise in cancer therapeutics by increasing the ratio of tumor control probability to normal tissue complication probability. It can increase the bioavailability of the drug at the target site, reduce the frequency of administration and reach otherwise lesser-accessible sites. The present study shows the feasibility of the cisplatin-loaded albumin nanoparticle as a sustained delivery system. AIMS: Cisplatin is one of the most widely used chemotherapeutic agents for the treatment of malignant disorders. Conventional cisplatin formulation given as intravenous infusion has low bioavailability to the target organ in addition to significant side-effects, like ototoxicity and nephrotoxicity. The aim of this study was to develop a protein-based nanoparticulate system for sustained release of cisplatin. MATERIALS AND METHODS: Nanoparticles were prepared by the coacervaton method of microcapsulation and chemical cross-linking with glutaraldehyde. Particle size was characterized by dynamic light scattering and transmission electron microscopy. RESULTS AND CONCLUSIONS: Using the coacervation method, nanoparticles of less than 70 nm diameter were produced. Drug encapsulation measured by ultraviolet spectroscopy varied from 30% to 80% for different ratios of cisplatin and protein. In vitro release kinetics shows that the nanoparticle-based formulation has biphasic release kinetics and is capable of sustained release compared with the free drug (80% release in 45 h). The study proves the feasibility of the albumin-based cisplatin nanoparticle formulation as a sustained release vehicle of cisplatin.

Accuracy of relocation, evaluation of geometric uncertainties and clinical target volume (CTV) to planning target volume (PTV) margin in fractionated stereotactic radiotherapy for intracranial tumors using relocatable Gill-Thomas-Cosman (GTC) frame.

The present study is aimed at determination of accuracy of relocation of Gill-Thomas-Cosman frame during fractionated stereotactic radiotherapy. The study aims to quantitatively determine the magnitudes of error in anteroposterior, mediolateral and craniocaudal directions, and determine the margin between clinical target volume to planning target volume based on systematic and random errors. Daily relocation error was measured using depth helmet and measuring probe. Based on the measurements, translational displacements in anteroposterior (z), mediolateral (x), and craniocaudal (y) directions were calculated. Based on the displacements in x, y and z directions, systematic and random error were calculated and three-dimensional radial displacement vector was determined. Systematic and random errors were used to derive CTV to PTV margin. The errors were within ± 2 mm in 99.2% cases in anteroposterior direction (AP), in 99.6% cases in mediolateral direction (ML), and in 97.6% cases in craniocaudal direction (CC). In AP, ML and CC directions, systematic errors were 0.56, 0.38, 0.42 mm and random errors were 1.86, 1.36 and 0.73 mm, respectively. Mean radial displacement was 1.03 mm ± 0.34. CTV to PTV margins calculated by ICRU formula were 1.86, 1.45 and 0.93 mm; by Stroom's formula they were 2.42, 1.74 and 1.35 mm; by van Herk's formula they were 2.7, 1.93 and 1.56 mm (AP, ML and CC directions). Depth helmet with measuring probe provides a clinically viable way for assessing the relocation accuracy of GTC frame. The errors were within ± 2 mm in all directions. Systematic and random errors were more along the anteroposterior axes. According to the ICRU formula, a margin of 2 mm around the tumor seems to be adequate.