Qualitative Evaluation of Common Quantitative Metrics for Clinical Acceptance of Automatic Segmentation: a Case Study on Heart Contouring from CT Images by Deep Learning Algorithms.

Organs-at-risk contouring is time consuming and labour intensive. Automation by deep learning algorithms would decrease the workload of radiotherapists and technicians considerably. However, the variety of metrics used for the evaluation of deep learning algorithms make the results of many papers difficult to interpret and compare. In this paper, a qualitative evaluation is done on five established metrics to assess whether their values correlate with clinical usability. A total of 377 CT volumes with heart delineations were randomly selected for training and evaluation. A deep learning algorithm was used to predict the contours of the heart. A total of 101 CT slices from the validation set with the predicted contours were shown to three experienced radiologists. They examined each slice independently whether they would accept or adjust the prediction and if there were (small) mistakes. For each slice, the scores of this qualitative evaluation were then compared with the Sørensen-Dice coefficient (DC), the Hausdorff distance (HD), pixel-wise accuracy, sensitivity and precision. The statistical analysis of the qualitative evaluation and metrics showed a significant correlation. Of the slices with a DC over 0.96 (N = 20) or a 95% HD under 5 voxels (N = 25), no slices were rejected by the readers. Contours with lower DC or higher HD were seen in both rejected and accepted contours. Qualitative evaluation shows that it is difficult to use common quantification metrics as indicator for use in clinic. We might need to change the reporting of quantitative metrics to better reflect clinical acceptance.

Automatic Cardiac Structure Contouring for Small Datasets with Cascaded Deep Learning Models.

Cardiac structure contouring is a time consuming and tedious manual activity used for radiotherapeutic dose toxicity planning. We developed an automatic cardiac structure segmentation pipeline for use in low-dose non-contrast planning CT based on deep learning algorithms for small datasets. Fifty CT scans were retrospectively selected and the whole heart, ventricles and atria were contoured. A two stage deep learning pipeline was trained on 41 non contrast planning CTs, tuned with 3 CT scans and validated on 6 CT scans. In the first stage, An InceptionResNetV2 network was used to identify the slices that contained cardiac structures. The second stage consisted of three deep learning models trained on the images containing cardiac structures to segment the structures. The three deep learning models predicted the segmentations/contours on axial, coronal and sagittal images and are combined to create the final prediction. The final accuracy of the pipeline was quantified on 6 volumes by calculating the Dice similarity coefficient (DC), 95% Hausdorff distance (95% HD) and volume ratios between predicted and ground truth volumes. Median DC and 95% HD of 0.96, 0.88, 0.92, 0.80 and 0.82, and 1.86, 2.98, 2.02, 6.16 and 6.46 were achieved for the whole heart, right and left ventricle, and right and left atria respectively. The median differences in volume were -4, -1, + 5, -16 and -20% for the whole heart, right and left ventricle, and right and left atria respectively. The automatic contouring pipeline achieves good results for whole heart and ventricles. Robust automatic contouring with deep learning methods seems viable for local centers with small datasets.

Supraventricular cardiac conduction system exposure in breast cancer patients treated with radiotherapy and association with heart and cardiac chambers doses.

Purpose: To assess sinoatrial node (SAN) and atrioventricular node (AVN) doses for breast cancer (BC) patients treated with 3D-CRT and evaluate whether "large" cardiac structures (whole heart and four cardiac chambers) would be relevant surrogates. Material and methods: This single center study was based on 116 BCE patients (56 left-sided, 60 right-sided) treated with 3D-CRT without respiratory gating strategies and few IMN irradiations from 2009 to 2013. The heart, the left and right ventricles (LV, RV), the left and right atria (LA, RA) were contoured using multi-atlases for auto-segmentation. The SAN and the AVN were manually delineated using a specific atlas. Based on regression analysis, the coefficients of determination (R2) were estimated to evaluate whether "large" cardiac structures were relevant surrogates (R2 > 0.70) of SAN and AVN doses. Results: For left-sided BC, mean doses were: 3.60 ± 2.28 Gy for heart, 0.47 ± 0.24 Gy for SAN and 0.74 ± 0.29 Gy for AVN. For right-sided BC, mean heart dose was 0.60 ± 0.25 Gy, mean SAN dose was 1.57 ± 0.63 Gy (>85 % of patients with SAN doses > 1 Gy) and mean AVN dose was 0.51 ± 0.14 Gy. Among all "large" cardiac structures, RA appeared as the best surrogate for SAN doses (R2 > 0.80). Regarding AVN doses, the RA may also be an interesting surrogate for left-sided BC (R2 = 0.78), but none of "large" cardiac structures appeared as relevant surrogates among right-sided BC (all R2 < 0.70), except the LA for patients with IMN (R2 = 0.83). Conclusions: In BC patients treated 10 years ago with 3D-CRT, SAN and AVN exposure was moderate but could exceed 1 Gy to the SAN in many right-sided patients with no IMN-inclusion. The RA appeared as an interesting surrogate for SAN exposure. Specific conduction nodes delineation remains necessary by using modern radiotherapy techniques.

A comparison of cardiovascular and pulmonary morbidities and risk factors in breast cancer survivors compared to an age-matched female control group in the Lifelines prospective population cohort.

PURPOSE: To provide more insight into late treatment-related toxicities among breast cancer (BC) survivors by comparing morbidities and risk factors between BC survivors and age-matched controls. MATERIALS AND METHODS: All female participants diagnosed with BC before inclusion in Lifelines, a population-based cohort in the Netherlands, were selected and matched 1:4 to female controls without any oncological history on birth year. Baseline was defined as the age at BC diagnosis. Outcomes were obtained from questionnaires and functional analyses performed at entry to Lifelines (follow-up 1; FU1) and several years later (FU2). Cardiovascular and pulmonary events were defined as morbidities that were absent at baseline but present at FU1 or FU2. RESULTS: The study consisted of 1,325 BC survivors and 5,300 controls. The median period from baseline (i.e., BC treatment) to FU1 and FU2 was 7 and 10 years, respectively. Among BC survivors more events of heart failure (OR: 1.72 [1.10-2.68]) and less events of hypertension (OR: 0.79 [0.66-0.94]) were observed. At FU2, more electrocardiographic abnormalities were found among BC survivors compared to controls (4.1% vs. 2.7%, respectively; p = 0.027) and Framingham scores for the 10-year risk of coronary heart disease were lower (difference: 0.37%; 95% CI [-0.70 to -0.03%]). At FU2, BC survivors had more frequently a forced vital capacity below the lower limit of normal than controls (5.4% vs. 2.9%, respectively; p = 0.040). CONCLUSION: BC survivors are at risk of late treatment-related toxicities despite a more favourable cardiovascular risk profile compared to age-matched female controls.

Complications of continuous-wear soft contact lenses in a nonreferral population.

Serious corneal complications occurred in an otherwise successful experience with continuous-wear soft contact lenses (SCLs) for aphakic correction. One hundred twenty eyes were fit, and 92% attained visual acuity of 20/40 or better. Severe corneal complications were observed in 13 eyes, including bacterial ulcers (six), apical erosions (three), and severe superficial vascularization (four). Corneal ulcers occurred in nondiabetic as well as diabetic subjects. Continuous-wear SCLs are not innocuous; as for any other drug or device, continuous, long-term medical supervision is necessary to minimize potentially severe complications and visual loss.

Joint Committee on Aviation Pathology: XIII. Sarcoid and the heart--an aeromedical risk.

Sarcoid lesions of the heart are reported in 13% to 20% of sarcoidosis patients. Sudden death and ventricular arrhythmias occur in 40% to 50%. In view of this frequency and life-threatening significance, review of military aircrew with documented sarcoidosis for myocardial involvement was undertaken. Case records were available for 11 US Army, 19 US Air Force, and 6 US Navy patients. Four patients (11%) had significant cardiac abnormalities to include sustained ventricular arrhythmias, and acquired anteroseptal infarction pattern by ECG with nomal coronary angiography. ECG abnormalities were found in eight additional patients. Suspected involvement of the heart in 33% of this series warrants complete cardiovascular evaluation of the sarcoid patient prior to return to flying duties. A case report demonstrates the difficulty in evaluating ECG abnormalities in the antemortem diagnosis of sarcoid heart disease.

Occlusion therapy for exodeviations occurring in infants and young children.

The authors treated 38 patients with exodeviations by occlusion of the preferred eye for three to six hours daily for an average of 15 months. This treatment resulted in a significant improvement in the latency of the deviation as well as a decrease in the size of the strabismic angle. The majority of those patients responding to this method of therapy initially had good central and maintained fixation in each eye, an exodeviation of 20 prism diopters or less at both 6 and 1/3 m, and the presence of fusional vergence amplitudes.

Light-chain paraproteins with lupus anticoagulant activity.

A patient with newly diagnosed multiple myeloma manifested by urine kappa light-chain excretion and a small monoclonal spike (0.4 g/dl), presented with lower extremity deep venous thrombosis. A preheparin plasma-activated partial thromboplastin time (aPTT) was prolonged at 68 sec (normal control 26-42 sec). Additional studies confirmed the presence of lupus anticoagulant activity in the serum: the modified Russell Viper Venom Time (MRVVT) was 73 sec (normal control 24-42 sec) and with a 50:50 mix of the patient's plasma and pooled normal plasma, the MRVVT remained prolonged. Kappa light chains (LC) were isolated from the patient's urine and their purity confirmed by electrophoresis and immunofixation using specific immunoglobulin antisera. The patient's LC mixed with pooled normal plasma demonstrated LA activity by in vitro clotting tests (plasma-activated partial thromboplastin time 62 sec, with normal control of 45 sec), MRVVT of 44 sec with normal control of 35 sec. Purified urinary kappa light chains from a control patient with multiple myeloma and normal clotting studies, failed to prolong either the plasma-activated partial thromboplastin time or the MRVVT. We hypothesize that kappa LC in our patient demonstrated LA activity, which was unique to these LCs. Paraproteins with LA activity, to date, have included only intact immunoglobulins (Ig). While intact Ig paraproteins have been reported to possess LA activity, this is the first report to our knowledge of light-chain paraproteins possessing similar activity and resulting in clinically evident thrombosis. Light chain paraproteins could serve as useful models for further study of the mechanisms of activity of acquired LA inhibitors.