Insulin-like growth factor-I (IGF-I) is associated with muscle hypertrophy, and circulating IGF-I levels are correlated with fitness. To test the hypothesis that IGF-I increases with increased physical activity in adolescent males, 38 subjects (16 ± 0.7 yr old) were randomized to control (n = 18) or increased physical activity groups for 5 wk. Before and after the intervention, we measured thigh muscle volume using magnetic resonance imaging and serum levels of mean growth hormone (GH) by overnight multiple sampling, GH binding protein (GHBP), IGF-I, and IGFBPs 1-5 by standard assays. Energy expenditure was assessed with the doubly labeled water technique toward the end of the study. In the training subjects there was 1) a significant increase in thigh muscle volume (+3.6 ± 1%), 2) 15.5 ± 3.3% greater energy expenditure than in controls, and 3) no evidence of weight loss (+1.44 ± 0.4%). In contrast to our hypothesis, but similar to our recent observations in adolescent females, training decreased IGF-I (-12 ± 4%, P < 0.005). Moreover, training substantially reduced GHBP (-21 ± 4%, P < 0.00002) and increased IGFBP-2 (+40 ± 16%, P < 0.008). Brief training increased muscle volume in weight-stable adolescent males and, surprisingly, influenced not only IGF-I but GHBP and IGFBP-2 as well in a manner typically found in energy-deficient states.

An acute insulin-like growth factor I (IGF-I) response to 10 min of above-lactate threshold cycle ergometer exercise was studied in 10 subjects (age 22-35 yr). Each subject exercised on three separate mornings after ingesting one of two isocaloric isovolemic liquid meals high in either fat or glucose or an isovolemic noncaloric placebo. The high-fat meal attenuated the growth hormone (GH) response (Cappon et al., J. Clin. Endocrinol. Metab. 76: 1418-1422, 1993). In contrast, IGF-I increased equally for all protocols [e.g., after the placebo meal IGF-I increased from 21,716 (SE) ng/ml preexercise to 25,316 ng/ml at 10 min of exercise; P < 0.05]. IGF-I peaked by the 10th min of exercise, like GH, and remained significantly elevated for only 20 min of recovery. We tested for possible GH-dependent mechanisms in which circulating IGF-I would increase 12-24 h after exercise. Ten subjects (age 23-32 yr) performed 10 min of above-lactate threshold exercise at 9, 10, and 11 A.M. GH was elevated after the first exercise bout (peak GH 6.05 +/- 1.45 ng/ml; P < 0.001) but was significantly reduced for the second and third bouts (peak GH 2.52 +/- 0.76 and 1.50 +/- 0.40 ng/ml, respectively). No increase in IGF-I was observed by 8 A.M. on the following day. Heavy ergometer exercise led to brief and small increases in circulating IGF-I that were independent of circulating GH.(ABSTRACT TRUNCATED AT 250 WORDS)

Background and purpose

Most brain lesions are characterized by hyperintense signal on FLAIR. We sought to develop an automated deep learning-based method for segmentation of abnormalities on FLAIR and volumetric quantification on clinical brain MRIs across many pathologic entities and scanning parameters. We evaluated the performance of the algorithm compared with manual segmentation and existing automated methods.

Materials and methods

We adapted a U-Net convolutional neural network architecture for brain MRIs using 3D volumes. This network was retrospectively trained on 295 brain MRIs to perform automated FLAIR lesion segmentation. Performance was evaluated on 92 validation cases using Dice scores and voxelwise sensitivity and specificity, compared with radiologists' manual segmentations. The algorithm was also evaluated on measuring total lesion volume.

Results

Our model demonstrated accurate FLAIR lesion segmentation performance (median Dice score, 0.79) on the validation dataset across a large range of lesion characteristics. Across 19 neurologic diseases, performance was significantly higher than existing methods (Dice, 0.56 and 0.41) and approached human performance (Dice, 0.81). There was a strong correlation between the predictions of lesion volume of the algorithm compared with true lesion volume (ρ = 0.99). Lesion segmentations were accurate across a large range of image-acquisition parameters on >30 different MR imaging scanners.

Conclusions

A 3D convolutional neural network adapted from a U-Net architecture can achieve high automated FLAIR segmentation performance on clinical brain MR imaging across a variety of underlying pathologies and image acquisition parameters. The method provides accurate volumetric lesion data that can be incorporated into assessments of disease burden or into radiologic reports.