Dr Ellen Donovan

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.

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.

Background Whole-breast radiotherapy (WBRT) is the standard treatment for breast cancer following breast-conserving surgery. Evidence shows that tumour recurrences occur near the original cancer: the tumour bed. New treatment developments include increasing dose to the tumour bed during WBRT (synchronous integrated boost) and irradiating only the region around the tumour bed, for patients at high and low risk of tumour recurrence, respectively. Currently, standard imaging uses bony anatomy to ensure accurate delivery of WBRT. It is debatable whether or not more targeted treatments such as synchronous integrated boost and partial-breast radiotherapy require image-guided radiotherapy (IGRT) focusing on implanted tumour bed clips (clip-based IGRT). Objectives Primary – to compare accuracy of patient set-up using standard imaging compared with clip-based IGRT. Secondary – comparison of imaging techniques using (1) tumour bed radiotherapy safety margins, (2) volume of breast tissue irradiated around tumour bed, (3) estimated breast toxicity following development of a normal tissue control probability model and (4) time taken. Design Multicentre observational study embedded within a national randomised trial: IMPORT-HIGH (Intensity Modulated and Partial Organ Radiotherapy – HIGHer-risk patient group) testing synchronous integrated boost and using clip-based IGRT. Setting Five radiotherapy departments, participating in IMPORT-HIGH. Participants Two-hundred and eighteen patients receiving breast radiotherapy within IMPORT-HIGH. Interventions There was no direct intervention in patients’ treatment. Experimental and control intervention were clip-based IGRT and standard imaging, respectively. IMPORT-HIGH patients received clip-based IGRT as routine; standard imaging data were obtained from clip-based IGRT images. Main outcome measures Difference in (1) set-up errors, (2) safety margins, (3) volume of breast tissue irradiated, (4) breast toxicity and (5) time, between clip-based IGRT and standard imaging. Results The primary outcome of overall mean difference in clip-based IGRT and standard imaging using daily set-up errors was 2–2.6 mm (p 

Aims: To measure cardiac tissue doses in left-sided breast cancer patients receiving supine tangential field radiotherapy with multileaf collimation (MLC) cardiac shielding of the heart and to assess the effect on target volume coverage. Materials and methods: Sixty-seven consecutive patients who underwent adjuvant radiotherapy to the left breast (n=48) or chest wall (n=19) in 2009/2010 were analysed. The heart, left anterior descending coronary artery (LAD), whole breast and partial breast clinical target volumes (WBCTV and PBCTV) were outlined retrospectively (the latter only in patients who had undergone breast-conserving surgery [BCS]). The mean heart and LAD NTD and maximum LAD doses (LAD) were calculated for all patients (NTD is a biologically weighted mean dose normalised to 2Gy fractions using a standard linear quadratic model). Coverage of WBCTV and PBCTV by the 95% isodose was assessed (BCS patients only). Results: The mean heart NTD (standard deviation) was 0.8 (0.3) Gy, the mean LAD NTD 6.7 (4.3) Gy and the mean LAD 40.3 (10.1) Gy. Coverage of the WBCTV by 95% isodose was

Aims: To determine the effect of image-guided radiotherapy on the dose distributions in breast boost treatments. Materials and methods: Computed tomography images from a cohort of 60 patients treated within the IMPORT HIGH trial (CRUK/06/003) were used to create sequential and concomitant boost treatment plans (30 cases each). Two treatment plans were created for each case using tumour bed planning target volume (PTV) margins of 5 mm (achieved with image-guided radiotherapy) and 8 mm (required for bony anatomy verification). Dose data were collected for breast, lung and heart; differences with margin size were tested for statistical significance. Results: A median decrease of 29 cm (range 11-193 cm) of breast tissue receiving 95% of the prescribed dose was observed where image-guided radiotherapy margins were used. Decreases in doses to lungs, contralateral breast and heart were modest, but statistically significant (P < 0.01). Plan quality was compromised with the 8 mm PTV margin in one in eight sequential boost plans and one third of concomitant boost plans. Tumour bed PTV coverage was 91%) of the prescribed dose in 12 cases; in addition, the required partial breast median dose was exceeded in nine concomitant boost cases by 0.5-3.7 Gy. Conclusions: The use of image guidance and, hence, a reduced tumour bed PTV margin, in breast boost radiotherapy resulted in a modest reduction in radiation dose to breast, lung and heart tissues. Reduced margins enabled by image guidance were necessary to discriminate between dose levels to multiple PTVs in the concomitant breast boost plans investigated.

Purpose To determine whether voluntary deep-inspiratory breath-hold (v-DIBH) and deep-inspiratory breath-hold with the active breathing coordinator™ (ABC-DIBH) in patients undergoing left breast radiotherapy are comparable in terms of normal-tissue sparing, positional reproducibility and feasibility of delivery. Methods Following surgery for early breast cancer, patients underwent planning-CT scans in v-DIBH and ABC-DIBH. Patients were randomised to receive one technique for fractions 1-7 and the second technique for fractions 8-15 (40 Gy/15 fractions total). Daily electronic portal imaging (EPI) was performed and matched to digitally-reconstructed radiographs. Cone-beam CT (CBCT) images were acquired for 6/15 fractions and matched to planning-CT data. Population systematic (Σ) and random errors (σ) were estimated. Heart, left-anterior-descending coronary artery, and lung doses were calculated. Patient comfort, radiographer satisfaction and scanning/treatment times were recorded. Within-patient comparisons between the two techniques used the paired t-test or Wilcoxon signed-rank test. Results Twenty-three patients were recruited. All completed treatment with both techniques. EPI-derived Σ were ≤1.8 mm (v-DIBH) and ≤2.0 mm (ABC-DIBH) and σ ≤2.5 mm (v-DIBH) and ≤2.2 mm (ABC-DIBH) (all p non-significant). CBCT-derived Σ were ≤3.9 mm (v-DIBH) and ≤4.9 mm (ABC-DIBH) and σ ≤ 4.1 mm (v-DIBH) and ≤ 3.8 mm (ABC-DIBH). There was no significant difference between techniques in terms of normal-tissue doses (all p non-significant). Patients and radiographers preferred v-DIBH (p = 0.007, p = 0.03, respectively). Scanning/treatment setup times were shorter for v-DIBH (p = 0.02, p = 0.04, respectively). Conclusions v-DIBH and ABC-DIBH are comparable in terms of positional reproducibility and normal tissue sparing. v-DIBH is preferred by patients and radiographers, takes less time to deliver, and is cheaper than ABC-DIBH. © 2013 Elsevier Ireland Ltd. All rights reserved.

This paper studies the sensitivity of a range of image texture parameters used in radiomics to: i) the number of intensity levels, ii) the method of quantisation to select the intensity levels and iii) the use of an intensity threshold. 43 commonly used texture features were studied for the gross target volume outlined on the CT component of PET/CT scans of 50 patients with non-small cell lung carcinoma (NSCLC). All cases were quantised for all values between 4 and 128 intensity levels using four commonly used quantisation methods. All results were analysed with and without a threshold range of -200 HU to 300 HU. Cases were ranked for each texture feature and for all quantisation methods with the Spearman's rank correlation coefficient determined to evaluate stability. Results showed large fluctuations in ranking, particularly for low numbers of levels, differences between quantisation methods and with the use of a threshold, with values Spearman's Rank Correlation for many parameters below 0.2. Our results demonstrated the sensitivity of radiomics features to the parameters used during analysis and highlight the risk of low reproducibility comparing studies with slightly different parameters. In terms of the lung cancer CT datasets, this study supports the use of 128 intensity levels, the same uniform quantiser applied to all scans and thresholding of the data. It also supports several of the features recommended in the literature for such studies such as skewness and kurtosis. A recommended framework is presented for curation of the data analysis process to ensure stability of results.

Purpose To compare mean heart and left anterior descending coronary artery (LAD) doses (NTDmean) and positional reproducibility in larger-breasted women receiving left breast radiotherapy using supine voluntary deep-inspiratory breath-hold (VBH) and free-breathing prone techniques. Materials and methods Following surgery for early breast cancer, patients with estimated breast volumes >750 cm3 underwent planning-CT scans in supine VBH and free-breathing prone positions. Radiotherapy treatment plans were prepared, and mean heart and LAD doses were calculated. Patients were randomised to receive one technique for fractions 1–7, before switching techniques for fractions 8–15 (40 Gy/15 fractions total). Daily electronic portal imaging and alternate-day cone-beam CT (CBCT) imaging were performed. The primary endpoint was the difference in mean LAD NTDmean between techniques. Population systematic (Σ) and random errors (σ) were estimated. Within-patient comparisons between techniques used Wilcoxon signed-rank tests. Results 34 patients were recruited, with complete dosimetric data available for 28. Mean heart and LAD NTDmean doses for VBH and prone treatments respectively were 0.4 and 0.7 (p < 0.001) and 2.9 and 7.8 (p < 0.001). Clip-based CBCT errors for VBH and prone respectively were ⩽3.0 mm and ⩽6.5 mm (Σ) and ⩽3.5 mm and ⩽5.4 mm (σ). Conclusions In larger-breasted women, supine VBH provided superior cardiac sparing and reproducibility than a free-breathing prone position.

Voluntary inspiration breath hold (VIBH) for left breast cancer patients has been shown to be a safe and effective method of reducing radiation dose to the heart. Currently, VIBH protocol compliance is monitored visually. In this work, we establish whether it is possible to gate the delivery of radiation from an Elekta linac using the Microsoft Kinect version 2 (Kinect v2) depth sensor to measure a patient breathing signal. This would allow contactless monitoring during VMAT treatment, as an alternative to equipment–assisted methods such as active breathing control (ABC). Breathing traces were acquired from six left breast radiotherapy patients during VIBH. We developed a gating interface to an Elekta linac, using the depth signal from a Kinect v2 to control radiation delivery to a programmable motion platform following patient breathing patterns. Radiation dose to a moving phantom with gating was verified using point dose measurements and a Delta4 verification phantom. 60 breathing traces were obtained with an acquisition success rate of 100%. Point dose measurements for gated deliveries to a moving phantom agreed to within 0.5% of ungated delivery to a static phantom using both a conventional and VMAT treatment plan. Dose measurements with the verification phantom showed that there was a median dose difference of better than 0.5% and a mean (3% 3 mm) gamma index of 92.6% for gated deliveries when using static phantom data as a reference. It is possible to use a Kinect v2 device to monitor voluntary breath hold protocol compliance in a cohort of left breast radiotherapy patients. Furthermore, it is possible to use the signal from a Kinect v2 to gate an Elekta linac to deliver radiation only during the peak inhale VIBH phase.

Aims To evaluate the feasibility and heart-sparing ability of the voluntary breath-hold (VBH) technique in a multicentre setting. Materials and methods Patients were recruited from 10 UK centres. Following surgery for early left breast cancer, patients with any heart inside the 50% isodose from a standard free-breathing tangential field treatment plan underwent a second planning computed tomography (CT) scan using the VBH technique. A separate treatment plan was prepared on the VBH CT scan and used for treatment. The mean heart, left anterior descending coronary artery (LAD) and lung doses were calculated. Daily electronic portal imaging (EPI) was carried out and scanning/treatment times were recorded. The primary end point was the percentage of patients achieving a reduction in mean heart dose with VBH. Population systematic (Σ) and random errors (σ) were estimated. Within-patient comparisons between techniques used Wilcoxon signed-rank tests. Results In total, 101 patients were recruited during 2014. Primary end point data were available for 93 patients, 88 (95%) of whom achieved a reduction in mean heart dose with VBH. Mean cardiac doses (Gy) for free-breathing and VBH techniques, respectively, were: heart 1.8 and 1.1, LAD 12.1 and 5.4, maximum LAD 35.4 and 24.1 (all P

Local cancer relapse risk after breast conservation surgery followed by radiotherapy has fallen sharply in many countries, and is influenced by patient age and clinicopathological factors. We hypothesise that partial-breast radiotherapy restricted to the vicinity of the original tumour in women at lower than average risk of local relapse will improve the balance of beneficial versus adverse effects compared with whole-breast radiotherapy. IMPORT LOW is a multicentre, randomised, controlled, phase 3, non-inferiority trial done in 30 radiotherapy centres in the UK. Women aged 50 years or older who had undergone breast-conserving surgery for unifocal invasive ductal adenocarcinoma of grade 1-3, with a tumour size of 3 cm or less (pT1-2), none to three positive axillary nodes (pN0-1), and minimum microscopic margins of non-cancerous tissue of 2 mm or more, were recruited. Patients were randomly assigned (1:1:1) to receive 40 Gy whole-breast radiotherapy (control), 36 Gy whole-breast radiotherapy and 40 Gy to the partial breast (reduced-dose group), or 40 Gy to the partial breast only (partial-breast group) in 15 daily treatment fractions. Computer-generated random permuted blocks (mixed sizes of six and nine) were used to assign patients to groups, stratifying patients by radiotherapy treatment centre. Patients and clinicians were not masked to treatment allocation. Field-in-field intensity-modulated radiotherapy was delivered using standard tangential beams that were simply reduced in length for the partial-breast group. The primary endpoint was ipsilateral local relapse (80% power to exclude a 2·5% increase [non-inferiority margin] at 5 years for each experimental group; non-inferiority was shown if the upper limit of the two-sided 95% CI for the local relapse hazard ratio [HR] was less than 2·03), analysed by intention to treat. Safety analyses were done in all patients for whom data was available (ie, a modified intention-to-treat population). This study is registered in the ISRCTN registry, number ISRCTN12852634. Between May 3, 2007, and Oct 5, 2010, 2018 women were recruited. Two women withdrew consent for use of their data in the analysis. 674 patients were analysed in the whole-breast radiotherapy (control) group, 673 in the reduced-dose group, and 669 in the partial-breast group. Median follow-up was 72·2 months (IQR 61·7-83·2), and 5-year estimates of local relapse cumulative incidence were 1·1% (95% CI 0·5-2·3) of patients in the control group, 0·2% (0·02-1·2) in the reduced-dose group, and 0·5% (0·2-1·4) in the partial-breast group. Estimated 5-year absolute differences in local relapse compared with the control group were -0·73% (-0·99 to 0·22) for the reduced-dose and -0·38% (-0·84 to 0·90) for the partial-breast groups. Non-inferiority can be claimed for both reduced-dose and partial-breast radiotherapy, and was confirmed by the test against the critical HR being more than 2·03 (p=0·003 for the reduced-dose group and p=0·016 for the partial-breast group, compared with the whole-breast radiotherapy group). Photographic, patient, and clinical assessments recorded similar adverse effects after reduced-dose or partial-breast radiotherapy, including two patient domains achieving statistically significantly lower adverse effects (change in breast appearance [p=0·007 for partial-breast] and breast harder or firmer [p=0·002 for reduced-dose and p