Real-time automatic tumor segmentation for ultrasound-guided breast-conserving surgery navigation

PURPOSE

Ultrasound-based navigation is a promising method in breast-conserving surgery, but tumor contouring often requires a radiologist at the time of surgery. Our goal is to develop a real-time automatic neural network-based tumor contouring process for intraoperative guidance. Segmentation accuracy is evaluated by both pixel-based metrics and expert visual rating.

METHODS

This retrospective study includes 7318 intraoperative ultrasound images acquired from 33 breast cancer patients, randomly split between 80:20 for training and testing. We implement a u-net architecture to label each pixel on ultrasound images as either tumor or healthy breast tissue. Quantitative metrics are calculated to evaluate the model's accuracy. Contour quality and usability are also assessed by fellowship-trained breast radiologists and surgical oncologists. Additionally, the viability of using our u-net model in an existing surgical navigation system is evaluated by measuring the segmentation frame rate.

RESULTS

The mean dice similarity coefficient of our u-net model is 0.78, with an area under the receiver-operating characteristics curve of 0.94, sensitivity of 0.95, and specificity of 0.67. Expert visual ratings are positive, with 93% of responses rating tumor contour quality at or above 7/10, and 75% of responses rating contour quality at or above 8/10. Real-time tumor segmentation achieved a frame rate of 16 frames-per-second, sufficient for clinical use.

CONCLUSION

Neural networks trained with intraoperative ultrasound images provide consistent tumor segmentations that are well received by clinicians. These findings suggest that neural networks are a promising adjunct to alleviate radiologist workload as well as improving efficiency in breast-conserving surgery navigation systems.

Participant selection for lung cancer screening by risk modelling (the Pan-Canadian Early Detection of Lung Cancer [PanCan] study): a single-arm, prospective study

BACKGROUND

Results from retrospective studies indicate that selecting individuals for low-dose CT lung cancer screening on the basis of a highly predictive risk model is superior to using criteria similar to those used in the National Lung Screening Trial (NLST; age, pack-year, and smoking quit-time). We designed the Pan-Canadian Early Detection of Lung Cancer (PanCan) study to assess the efficacy of a risk prediction model to select candidates for lung cancer screening, with the aim of determining whether this approach could better detect patients with early, potentially curable, lung cancer.

METHODS

We did this single-arm, prospective study in eight centres across Canada. We recruited participants aged 50-75 years, who had smoked at some point in their life (ever-smokers), and who did not have a self-reported history of lung cancer. Participants had at least a 2% 6-year risk of lung cancer as estimated by the PanCan model, a precursor to the validated PLCOm2012 model. Risk variables in the model were age, smoking duration, pack-years, family history of lung cancer, education level, body-mass index, chest x-ray in the past 3 years, and history of chronic obstructive pulmonary disease. Individuals were screened with low-dose CT at baseline (T0), and at 1 (T1) and 4 (T4) years post-baseline. The primary outcome of the study was incidence of lung cancer. This study is registered with ClinicalTrials.gov, number NCT00751660.

FINDINGS

7059 queries came into the study coordinating centre and were screened for PanCan risk. 15 were duplicates, so 7044 participants were considered for enrolment. Between Sept 24, 2008, and Dec 17, 2010, we recruited and enrolled 2537 eligible ever-smokers. After a median follow-up of 5·5 years (IQR 3·2-6·1), 172 lung cancers were diagnosed in 164 individuals (cumulative incidence 0·065 [95% CI 0·055-0·075], incidence rate 138·1 per 10 000 person-years [117·8-160·9]). There were ten interval lung cancers (6% of lung cancers and 6% of individuals with cancer): one diagnosed between T0 and T1, and nine between T1 and T4. Cumulative incidence was significantly higher than that observed in NLST (4·0%; p<0·0001). Compared with 593 (57%) of 1040 lung cancers observed in NLST, 133 (77%) of 172 lung cancers in the PanCan Study were early stage (I or II; p<0·0001).

INTERPRETATION

The PanCan model was effective in identifying individuals who were subsequently diagnosed with early, potentially curable, lung cancer. The incidence of cancers detected and the proportion of early stage cancers in the screened population was higher than observed in previous studies. This approach should be considered for adoption in lung cancer screening programmes.

FUNDING

Terry Fox Research Institute and Canadian Partnership Against Cancer.

Airway wall thickening and emphysema show independent familial aggregation in chronic obstructive pulmonary disease

RATIONALE

It is unclear whether airway wall thickening and emphysema make independent contributions to airflow limitation in chronic obstructive pulmonary disease (COPD) and whether these phenotypes cluster within families.

OBJECTIVES

To determine whether airway wall thickening and emphysema (1) make independent contributions to the severity of COPD and (2) show independent aggregation in families of individuals with COPD.

METHODS

Index cases with COPD and their smoking siblings underwent spirometry and were offered high-resolution computed tomography scans of the thorax to assess the severity of airway wall thickening and emphysema.

MEASUREMENTS AND MAIN RESULTS

A total of 3,096 individuals were recruited to the study, of whom 1,159 (519 probands and 640 siblings) had technically adequate high-resolution computed tomography scans without significant non-COPD-related thoracic disease. Airway wall thickness correlated with pack-years smoked (P < or = 0.001) and symptoms of chronic bronchitis (P < 0.001). FEV(1) (expressed as % predicted) was independently associated with airway wall thickness at a lumen perimeter of 10 mm (P = 0.0001) and 20 mm (P = 0.0013) and emphysema at -950 Hounsfield units (P < 0.0001). There was independent familial aggregation of both the emphysema (adjusted odds ratio, 2.1; 95% confidence interval, 1.1-4.0; P < or = 0.02) and airway disease phenotypes (P < 0.0001) of COPD.

CONCLUSIONS

Airway wall thickening and emphysema make independent contributions to airflow obstruction in COPD. These phenotypes show independent aggregation within families of individuals with COPD, suggesting that different genetic factors influence these disease processes.