Machine learning classification of mediastinal lymph node metastasis in NSCLC: a multicentre study in a Western European patient population.

BACKGROUND: [18F] FDG PET-CT has an important role in the initial staging of lung cancer; however, accurate differentiation between activity in malignant and benign intrathoracic lymph nodes on PET-CT scans can be challenging. The purpose of the current study was to investigate the effect of incorporating primary tumour data and clinical features to differentiate between [18F] FDG-avid malignant and benign intrathoracic lymph nodes. METHODS: We retrospectively selected lung cancer patients who underwent PET-CT for initial staging in two centres in the Netherlands. The primary tumour and suspected lymph node metastases were annotated and cross-referenced with pathology results. Lymph nodes were classified as malignant or benign. From the image data, we extracted radiomic features and trained the classifier model using the extreme gradient boost (XGB) algorithm. Various scenarios were defined by selecting different combinations of data input and clinical features. Data from centre 1 were used for training and validation of the models using the XGB algorithm. To determine the performance of the model in a different hospital, the XGB model was tested using data from centre 2. RESULTS: Adding primary tumour data resulted in a significant gain in the performance of the trained classifier model. Adding the clinical information about distant metastases did not lead to significant improvement. The performance of the model in the test set (centre 2) was slightly but statistically significantly lower than in the validation set (centre 1). CONCLUSIONS: Using the XGB algorithm potentially leads to an improved model for the classification of intrathoracic lymph nodes. The inclusion of primary tumour data improved the performance of the model, while additional knowledge of distant metastases did not. In patients in whom metastases are limited to lymph nodes in the thorax, this may reduce costly and invasive procedures such as endobronchial ultrasound or mediastinoscopy procedures.

Noninvasive mapping of left ventricular electromechanical asynchrony by three-dimensional echocardiography and semi-automatic contour detection.

A system for analyzing left ventricular (LV) electromechanical asynchrony based on transesophageal 3-dimensional echocardiography (3-DE) and semi-automatic endocardial contour detection is described. Eighteen consecutive patients underwent 3-DE. Using TomTec 4DLV software, a 3-dimensional endocardial surface was reconstructed throughout the cardiac cycle. Matlab software generated color-coded polar maps, displaying regional LV displacement and its timing. At the segmental level, Bland-Altmann assessment showed intraobserver variability of LV displacement of 0.1 +/- 3.0 mm and timing of -5.6 +/- 160 ms (bias +/- 2 SD) for all segments and -1.6 +/- 94 ms for the nonapical segments. The combination of 3-DE and semi-automatic contour detection is feasible and provides unique information for assessing regional LV endocardial displacement and electromechanical asynchrony.

Early identification of left ventricular remodelling after myocardial infarction, assessed by transthoracic 3D echocardiography.

AIMS: The usefulness of 3D echocardiography (3DE) for accurate evaluation of left ventricular (LV) remodelling after acute myocardial infarction (AMI), and early identification of remodelling in the subacute phase, was assessed. METHODS AND RESULTS: Thirty-three AMI patients (21 anterior AMIs) underwent 3DE prospectively at baseline (6+/-4 days) and at 3, 6, and 12 months post-AMI. Remodelling was defined as >20% increase in end-diastolic volume (EDV) at 6 or 12 months in relation to baseline. In patients with remodelling (n = 13) at baseline, EDV and end-systolic volume (ESV), but not ejection fraction (EF), were significantly increased compared to patients without subsequent remodelling (n = 20). At 12 months, EDV and ESV increased further and significantly, and EF was unchanged in patients with remodelling, whilst LV volumes were unchanged and EF slightly increased in patients without remodelling. Clinical, electrocardiographic, and echocardiographic variables were analysed for the early identification of LV remodelling. Of these, at baseline the 3D sphericity index (EDV divided by the volume of a sphere, the diameter of which is the LV major end-diastolic long axis) was, by far, the most predictive variable with a sensitivity, specificity, and positive and negative predictive value for a cutoff value of >0.25 of 100%, 90%, 87% and 100%, respectively. CONCLUSIONS: Three-dimensional echocardiography can differentiate patients with and without subsequent development of LV remodelling accurately and early on the basis of the 3D sphericity index, a new and highly predictive variable.

Quantification of left ventricular volumes and ejection fraction using freehand transthoracic three-dimensional echocardiography: comparison with magnetic resonance imaging.

OBJECTIVES: Our aim was to validate 3-dimensional echocardiography (3DE) for assessment of left ventricular (LV) end-diastolic volume, end-systolic volume (ESV), stroke volume, and ejection fraction (EF) using the freehand-acquisition method. Furthermore, LV volumes by breath hold-versus free breathing-3DE acquisition were assessed and compared with magnetic resonance imaging (MRI). METHODS: From the apical position, a fan-like 3DE image was acquired during free breathing and another, thereafter, during breath hold. In 27 patients, 28 breath hold- and 24 free breathing-3DE images were acquired. A total of 17 patients underwent both MRI and 3DE. MRI contours were traced along the outer endocardial contour, including trabeculae, and along the inner endocardial contour, excluding trabeculae, from the LV volume. RESULTS: All 28 (100%) breath hold- and 86% of free breathing-3DE acquisitions could be analyzed. Intraobserver variation (percentual bias +/- 2 SD) of end-diastolic volume, ESV, stroke volume, and EF for breath-hold 3DE was, respectively, 0.3 +/- 10.2%, 0.3 +/- 14.6%, 0.1 +/- 18.4%, and -0.1 +/- 5.8%. For free-breathing 3DE, findings were similar. A significantly better interobserver variability, however, was observed for breath-hold 3DE for ESV and EF. Comparison of breath-hold 3DE with MRI inner contour showed for end-diastolic volume, ESV, stroke volume, and EF, a percentual bias (+/- 2 SD) of, respectively, -13.5 +/- 26.9%, -17.7 +/- 47.8%, -10.6 +/- 43.6%, and -1.8 +/- 11.6%. Compared with the MRI outer contour, a significantly greater difference was observed, except for EF. CONCLUSIONS: 3DE using the freehand method is fast and highly reproducible for (serial) LV volume and EF measurement, and, hence, ideally suited for clinical decision making and trials. Breath-hold 3DE is superior to free-breathing 3DE regarding image quality and reproducibility. Compared with MRI, 3DE underestimates LV volumes, but not EF, which is mainly explained by differences in endocardial contour tracing by MRI (outer contour) and 3DE (inner contour) of the trabecularized endocardium. Underestimation is reduced when breath-hold 3DE is compared with inner contour analysis of the MRI dataset.

Importance of transducer displacement and tilting on three-dimensional echocardiographic volume assessment using apical or off-axis rotational acquisition: an in vitro study.

OBJECTIVE: The goal of this study was to assess effects of translation (horizontal displacement) and angulation (transducer tilting) on 3-dimensional (3D) echocardiographic volumes of both balloons and human left ventricles after autopsy. METHODS: Six water-filled (non-) aneurysmatic balloons of 150, 250, and 350 mL and 3 hearts of different sizes and shapes were suspended upright in a water bath. Angulation and/or translation was performed respectively by tilting the transducer with a mechanical arm in a vertical plane relative to the balloon tip or true apex of the hearts and by shifting the water bath in the same vertical plane. For balloon and left ventricular (LV) volume assessment, a 3D conical data set was obtained by TomTec rotational acquisition in combination with a HP Sonos 5500 ultrasound machine. RESULTS: For the 6 balloons, translation from 1 to 4 cm yielded volumes of up to 74% of the optimal volume (100%); angulation of 10 degrees or 20 degrees, volumes of up to 80% and 34%. Translation with 10-degree angulation yielded volumes up to 64%; for 20-degree angulation and translation, there was no volume loss. Results were similar for the left ventricles. CONCLUSIONS: Even minor angulation or translation of the transducer yields substantial underestimation of the true volume. Off-axis para-apical views, however, defined as angulation of 20 degrees and greater than 0.5 cm translation in this in vitro model, obviate volume underestimation. Such views in patients, if obtainable, may be an attractive alternative for conventional apical 3D acquisition, especially in dilated and aneurysmatic hearts.