Receive a weekly summary and discussion of the top papers of the week by leading researchers in the field.

General General

Clinical classification of scoliosis patients using machine learning and markerless 3D surface trunk data.

In Medical & biological engineering & computing ; h5-index 32.0

Markerless 3D surface topography for scoliosis diagnosis and brace treatment can avoid repeated radiation known from standard X-ray analysis and possible side effects. Combined with the method of torso asymmetry analysis, curve severity and progression can be evaluated with high reliability. In the current study, a machine learning approach was utilised to classify scoliosis patients based on their trunk surface asymmetry pattern. Frontal X-ray and 3D scanning analysis with a clinical classification based on Cobb angle and spinal curve pattern were performed with 50 patients. Similar as in a previous study, each patient's trunk 3D reconstruction was used for an elastic registration of a reference surface mesh with fixed number of vertices. Subsequently, an asymmetry distance map between original and reflected torso was calculated. A fully connected neural network was then utilised to classify patients regarding their Cobb angle (mild, moderate, severe) and an Augmented Lehnert-Schroth (ALS) classification based on their full torso asymmetry distance map. The results reveal a classification success rate of 90% (SE: 80%, SP: 100%) regarding the curve severity (mild vs moderate-severe) and 50-72% regarding the ALS group. Identifying patient curve severity and treatment group was reasonably possible allowing for a decision support during diagnosis and treatment planning. Graphical abstract.

Rothstock Stephan, Weiss Hans-Rudolf, Krueger Daniel, Paul Lothar


3D surface scan, Asymmetry distance map, Classification, Machine learning, Scoliosis

Radiology Radiology

Prediction of breast cancer molecular subtypes on DCE-MRI using convolutional neural network with transfer learning between two centers.

In European radiology ; h5-index 62.0

OBJECTIVES : To apply deep learning algorithms using a conventional convolutional neural network (CNN) and a recurrent CNN to differentiate three breast cancer molecular subtypes on MRI.

METHODS : A total of 244 patients were analyzed, 99 in training dataset scanned at 1.5 T and 83 in testing-1 and 62 in testing-2 scanned at 3 T. Patients were classified into 3 subtypes based on hormonal receptor (HR) and HER2 receptor: (HR+/HER2-), HER2+, and triple negative (TN). Only images acquired in the DCE sequence were used in the analysis. The smallest bounding box covering tumor ROI was used as the input for deep learning to develop the model in the training dataset, by using a conventional CNN and the convolutional long short-term memory (CLSTM). Then, transfer learning was applied to re-tune the model using testing-1(2) and evaluated in testing-2(1).

RESULTS : In the training dataset, the mean accuracy evaluated using tenfold cross-validation was higher by using CLSTM (0.91) than by using CNN (0.79). When the developed model was applied to the independent testing datasets, the accuracy was 0.4-0.5. With transfer learning by re-tuning parameters in testing-1, the mean accuracy reached 0.91 by CNN and 0.83 by CLSTM, and improved accuracy in testing-2 from 0.47 to 0.78 by CNN and from 0.39 to 0.74 by CLSTM. Overall, transfer learning could improve the classification accuracy by greater than 30%.

CONCLUSIONS : The recurrent network using CLSTM could track changes in signal intensity during DCE acquisition, and achieved a higher accuracy compared with conventional CNN during training. For datasets acquired using different settings, transfer learning can be applied to re-tune the model and improve accuracy.

KEY POINTS : • Deep learning can be applied to differentiate breast cancer molecular subtypes. • The recurrent neural network using CLSTM could track the change of signal intensity in DCE images, and achieved a higher accuracy compared with conventional CNN during training. • For datasets acquired using different scanners with different imaging protocols, transfer learning provided an efficient method to re-tune the classification model and improve accuracy.

Zhang Yang, Chen Jeon-Hor, Lin Yezhi, Chan Siwa, Zhou Jiejie, Chow Daniel, Chang Peter, Kwong Tiffany, Yeh Dah-Cherng, Wang Xinxin, Parajuli Ritesh, Mehta Rita S, Wang Meihao, Su Min-Ying


Breast neoplasms, Machine learning, Magnetic resonance imaging, Receptor, ErbB-2, Receptors, estrogen

Radiology Radiology

Opportunistic osteoporosis screening in multi-detector CT images using deep convolutional neural networks.

In European radiology ; h5-index 62.0

OBJECTIVE : To explore the application of deep learning in patients with primary osteoporosis, and to develop a fully automatic method based on deep convolutional neural network (DCNN) for vertebral body segmentation and bone mineral density (BMD) calculation in CT images.

MATERIALS AND METHODS : A total of 1449 patients were used for experiments and analysis in this retrospective study, who underwent spinal or abdominal CT scans for other indications between March 2018 and May 2020. All data was gathered from three different CT vendors. Among them, 586 cases were used for training, and other 863 cases were used for testing. A fully convolutional neural network, called U-Net, was employed for automated vertebral body segmentation. The manually sketched region of vertebral body was used as the ground truth for comparison. A convolutional neural network, called DenseNet-121, was applied for BMD calculation. The values post-processed by quantitative computed tomography (QCT) were identified as the standards for analysis.

RESULTS : Based on the diversity of CT vendors, all testing cases were split into three testing cohorts: Test set 1 (n = 463), test set 2 (n = 200), and test set 3 (n = 200). Automated segmentation correlated well with manual segmentation regarding four lumbar vertebral bodies (L1-L4): the minimum average dice coefficients for three testing sets were 0.823, 0.786, and 0.782, respectively. For testing sets from different vendors, the average BMDs calculated by automated regression showed high correlation (r > 0.98) and agreement with those derived from QCT.

CONCLUSIONS : A deep learning-based method could achieve fully automatic identification of osteoporosis, osteopenia, and normal bone mineral density in CT images.

KEY POINTS : • Deep learning can perform accurate fully automated segmentation of lumbar vertebral body in CT images. • The average BMDs obtained by deep learning highly correlates with ones derived from QCT. • The deep learning-based method could be helpful for clinicians in opportunistic osteoporosis screening in spinal or abdominal CT scans.

Fang Yijie, Li Wei, Chen Xiaojun, Chen Keming, Kang Han, Yu Pengxin, Zhang Rongguo, Liao Jianwei, Hong Guobin, Li Shaolin


Bone density, Deep learning, Osteoporosis, Spine, Tomography, X-ray computed

Public Health Public Health

Low-Pass Whole Genome Bisulfite Sequencing of Neonatal Dried Blood Spots Identifies a Role for RUNX1 in Down Syndrome DNA Methylation Profiles.

In Human molecular genetics ; h5-index 81.0

Neonatal dried blood spots (NDBS) are a widely banked sample source that enable retrospective investigation into early-life molecular events. Here, we performed low-pass whole genome bisulfite sequencing (WGBS) of 86 NDBS DNA to examine early-life Down syndrome (DS) DNA methylation profiles. DS represents an example of genetics shaping epigenetics, as multiple array-based studies have demonstrated that trisomy 21 is characterized by genome-wide alterations to DNA methylation. By assaying over 24 million CpG sites, thousands of genome-wide significant (q < 0.05) DMRs that distinguished DS from typical development (TD) and idiopathic developmental delay (DD) were identified. Machine learning feature selection refined these DMRs to 22 loci. The DS DMRs mapped to genes involved in neurodevelopment, metabolism, and transcriptional regulation. Based on comparisons to previous DS methylation studies and reference epigenomes, the hypermethylated DS DMRs were significantly (q < 0.05) enriched across tissues while the hypomethylated DS DMRs were significantly (q < 0.05) enriched for blood-specific chromatin states. A ~ 28 kb block of hypermethylation was observed on chromosome 21 in the RUNX1 locus, which encodes a hematopoietic transcription factor whose binding motif was the most significantly enriched (q < 0.05) overall and specifically within the hypomethylated DMRs. Finally, we also identified DMRs that distinguished DS NDBS based on the presence or absence of congenital heart disease (CHD). Together, these results not only demonstrate the utility of low-pass WGBS on NDBS samples for epigenome-wide association studies, but also provide new insights into the early-life mechanisms of epigenomic dysregulation resulting from trisomy 21.

Laufer Benjamin I, Hwang Hyeyeon, Jianu Julia M, Mordaunt Charles E, Korf Ian F, Hertz-Picciotto Irva, LaSalle Janine M


Cardiology Cardiology

Artificial intelligence and cardiovascular imaging: A win-win combination.

In Anatolian journal of cardiology

Rapid development of artificial intelligence (AI) is gaining grounds in medicine. Its huge impact and inevitable necessity are also reflected in cardiovascular imaging. Although AI would probably never replace doctors, it can significantly support and improve their productivity and diagnostic performance. Many algorithms have already proven useful at all stages of the cardiac imaging chain. Their crucial practical applications include classification, automatic quantification, notification, diagnosis, and risk prediction. Consequently, more reproducible and repeatable studies are obtained, and personalized reports may be available to any patient. Utilization of AI also increases patient safety and decreases healthcare costs. Furthermore, AI is particularly useful for beginners in the field of cardiac imaging as it provides anatomic guidance and interpretation of complex imaging results. In contrast, lack of interpretability and explainability in AI carries a risk of harmful recommendations. This review was aimed at summarizing AI principles, essential execution requirements, and challenges as well as its recent applications in cardiovascular imaging.

Badano Luigi P, Keller Daria M, Muraru Denisa, Torlasco Camilla, Parati Gianfranco


General General

Vicarious reward unblocks associative learning about novel cues in male rats.

In eLife

Many species, including rats, are sensitive to social signals and their valuation is important in social learning. Here we introduce a task that investigates if mutual reward delivery in male rats can drive associative learning. We found that when actor rats have fully learned a stimulus-self-reward association, adding a cue that predicted additional reward to a partner unblocked associative learning about this cue. By contrast, additional cues that did not predict partner reward remained blocked from acquiring positive associative value. Importantly, this social unblocking effect was still present when controlling for secondary reinforcement but absent when social information exchange was impeded, when mutual reward outcomes were disadvantageously unequal to the actor or when the added cue predicted reward delivery to an empty chamber. Taken together, these results suggest that mutual rewards can drive associative learning in rats and is dependent on vicariously experienced social and food-related cues.

van Gurp Sander, Hoog Jochen, Kalenscher Tobias, van Wingerden Marijn


associative learning, neuroscience, rat, unblocking, vicarious reward