Automated echocardiography view classification and quality assessment with recognition of unknown views.

Purpose: Interpreting echocardiographic exams requires substantial manual interaction as videos lack scan-plane information and have inconsistent image quality, ranging from clinically relevant to unrecognizable. Thus, a manual prerequisite step for analysis is to select the appropriate views that showcase both the target anatomy and optimal image quality. To automate this selection process, we present a method for automatic classification of routine views, recognition of unknown views, and quality assessment of detected views. Approach: We train a neural network for view classification and employ the logit activations from the neural network for unknown view recognition. Subsequently, we train a linear regression algorithm that uses feature embeddings from the neural network to predict view quality scores. We evaluate the method on a clinical test set of 2466 echocardiography videos with expert-annotated view labels and a subset of 438 videos with expert-rated view quality scores. A second observer annotated a subset of 894 videos, including all quality-rated videos. Results: The proposed method achieved an accuracy of 84.9 % ± 0.67 for the joint objective of routine view classification and unknown view recognition, whereas a second observer reached an accuracy of 87.6%. For view quality assessment, the method achieved a Spearman's rank correlation coefficient of 0.71, whereas a second observer reached a correlation coefficient of 0.62. Conclusion: The proposed method approaches expert-level performance, enabling fully automatic selection of the most appropriate views for manual or automatic downstream analysis.

Multi-centre real-world validation of automated treatment planning for breast radiotherapy.

PURPOSE: To present the results of the first multi-centre real-world validation of autoplanning for whole breast irradiation after breast-sparing surgery, encompassing high complexity cases (e.g. with a boost or regional lymph nodes) and a wide range of clinical practices. METHODS: The 24 participating centers each included 10 IMRT/VMAT/Tomotherapy patients, previously treated with a manually generated plan ('manplan'). There were no restrictions regarding case complexity, planning aims, plan evaluation parameters and criteria, fractionation, treatment planning system or treatment machine/technique. In addition to dosimetric comparisons of autoplans with manplans, blinded plan scoring/ranking was conducted by a clinician from the treating center. Autoplanning was performed using a single configuration for all patients in all centres. Deliverability was verified through measurements at delivery units. RESULTS: Target dosimetry showed comparability, while reductions in OAR dose parameters were 21.4 % for heart Dmean, 16.7 % for ipsilateral lung Dmean, and 101.9 %, 45.5 %, and 35.7 % for contralateral breast D0.03cc, D5% and Dmean, respectively (all p < 0.001). Among the 240 patients included, the clinicians preferred the autoplan for 119 patients, with manplans preferred for 96 cases (p = 0.01). Per centre there were on average 5.0 ± 2.9 (1SD) patients with a preferred autoplan (range [0-10]), compared to 4.0 ± 2.7 with a preferred manplan ([0,9]). No differences were observed regarding deliverability. CONCLUSION: The automation significantly reduced the hands-on planning workload compared to manual planning, while also achieving an overall superiority. However, fine-tuning of the autoplanning configuration prior to clinical implementation may be necessary in some centres to enhance clinicians' satisfaction with the generated autoplans.

Stochastic co-teaching for training neural networks with unknown levels of label noise.

Label noise hampers supervised training of neural networks. However, data without label noise is often infeasible to attain, especially for medical tasks. Attaining high-quality medical labels would require a pool of experts and their consensus reading, which would be extremely costly. Several methods have been proposed to mitigate the adverse effects of label noise during training. State-of-the-art methods use multiple networks that exploit different decision boundaries to identify label noise. Among the best performing methods is co-teaching. However, co-teaching comes with the requirement of knowing label noise a priori. Hence, we propose a co-teaching method that does not require any prior knowledge about the level of label noise. We introduce stochasticity to select or reject training instances. We have extensively evaluated the method on synthetic experiments with extreme label noise levels and applied it to real-world medical problems of ECG classification and cardiac MRI segmentation. Results show that the approach is robust to its hyperparameter choice and applies to various classification tasks with unknown levels of label noise.

Are intraoral customized stents still necessary in the era of Highly Conformal Radiotherapy for Head & Neck cancer? Case series and literature review.

AIM: To evaluate the dose sparing efficacy of intraoral customized stents in combination with IGRT/VMAT in Head & Neck cancer patients. BACKGROUND: Despite advances in high-dose conformal radiotherapy (RT) techniques, adverse effects (such as oral mucositis) during and after RT often require temporary suspension of treatment and affect the quality of life in survivors. Intraoral customized stents can decrease radiation doses in healthy tissues and minimize damage from radiations. At the best of our knowledge the clinical impact of such devices in combination with VMAT (volumetric modulated arc therapy) is not reported in the literature. CASES DESCRIPTION: Three Head & Neck cancer patients were submitted to image guided (IG) RT/VMAT in their treatment protocol. Dose distribution with and without the use of an intraoral stent was compared in each patient. Mean radiation doses proved to be lower in all patients, especially in the subsite: oral cavity. CONCLUSIONS: There are several reports on the efficacy of IS during RT for Head & Neck cancer. Despite technological advances, the combination between high conformal RT and intraoral stents could still play a role in the management of this kind of patients. This strengthens the usefulness of the individualization of treatments and multidisciplinary approach.

Comparison of mathematical models for red marrow and blood absorbed dose estimation in the radioiodine treatment of advanced differentiated thyroid carcinoma.

Metastatic and recurrent differentiated thyroid carcinoma is preferably treated with (131)I, whose administered activity is limited by red marrow (RM) toxicity, originally correlated by Benua to a blood absorbed dose higher than 2 Gy. Afterward a variety of dosimetric approaches has been proposed. The aim of this work is to compare the results of the Benua formula with the ones of other three blood and RM absorbed dose formulae. Materials and methods have been borrowed by the dosimetric protocol of the Italian Internal Dosimetry group and adapted to the routine of our centre. Wilcoxon t-tests and percentage differences have been applied for comparison purposes. Results are significantly different (p < 0.05) from each other, with an average percentage difference between Benua versus other results of -22%. The dosimetric formula applied to determine blood or RM absorbed dose may contribute significantly to increase heterogeneity in absorbed dose and dose-response results. Standardization should be a major objective.