In Gastrointestinal endoscopy ; h5-index 72.0

BACKGROUND AND AIMS : Colorectal cancer is a leading cause of death. Colonoscopy is the criterion standard for detection and removal of precancerous lesions and has been shown to reduce mortality. The polyp miss rate during colonoscopies is 22% to 28%. DEEP DEtection of Elusive Polyps (DEEP²) is a new polyp detection system based on deep learning, which alerts the operator in real-time to the presence and location of polyps. The primary outcome was the performance of DEEP² on the detection of elusive polyps.

METHODS : The DEEP² system was trained on 3,611 hours of colonoscopy videos derived from 2 sources, and was validated on a set comprising 1,393 hours, from a third unrelated source. The ground truth labeling was provided by offline gastroenterologist annotators, who were able to watch the video in slow-motion and pause/rewind as required. To assess the applicability, stability, user experience and in order to obtain some preliminary data on performance in a real-life scenario, a preliminary prospective clinical validation study was performed, comprising 100 procedures (clinicaltrial.gov ID: NCT04693078).

RESULTS : DEEP² achieved a sensitivity of 97.1% at 4.6 false alarms per video for all polyps, 88.5% and 84.9% for polyps that are in the field of view for less than 5 and 2 seconds, respectively. DEEP² was able to detect polyps, not seen by live real-time endoscopists or offline annotators in an average of 0.22 polyps per sequence. In the clinical validation study the system detected an average of 0.89 additional polyps per procedure. No adverse events occurred.

CONCLUSIONS : DEEP² has a high sensitivity for polyp detection and was effective in increasing the detection of polyps both in colonoscopy videos and in real procedures with a low number of false alarms.

Livovsky Dan M, Veikherman Danny, Golany Tomer, Aides Amit, Dashinsky Valentin, Rabani Nadav, Ben Shimol David, Blau Yochai, Katzir Liran, Shimshoni Ilan, Liu Yun, Segol Ori, Goldin Eran, Corrado Greg, Lachter Jesse, Matias Yossi, Rivlin Ehud, Freedman Daniel

2021-Jun-30

In International journal of computer assisted radiology and surgery

PURPOSE : Deep Brain Stimulation (DBS) is a proven therapy for Parkinson's Disease (PD), frequently resulting in an enhancement of motor function. Nonetheless, several undesirable side effects can occur after DBS, which can worsen the quality of life of the patient. Thus, the clinical team has to carefully select patients on whom to perform DBS. Over the past decade, there have been some attempts to relate pre-operative data and DBS clinical outcomes, with most focused on the motor symptomatology. In this paper, we propose a machine learning-based method able to predict a large number of DBS clinical outcomes for PD.

METHODS : We propose a multimodal pipeline, referred to as PassFlow, which predicts 84 clinical post-operative clinical scores. PassFlow is composed of an artificial neural network to compress clinical information, an image processing method from the state-of-the-art to extract morphological biomarkers our of T1 imaging, and an SVM to perform the regressions. We validated PassFlow on 196 PD patients who undergone a DBS.

RESULTS : PassFlow showed correlation coefficients as high as 0.71 and were able to significantly predict 63 out of the 84 scores, outperforming a comparative linear method. The number of metrics that are predicted with this pre-operative information was also found to be correlated with the number of patients with this information available, indicating that the PassFlow method is still actively learning.

CONCLUSION : We presented a novel, machine learning-based pipeline to predict a variety of post-operative clinical outcomes of DBS for PD patients. PassFlow took into account various bio-markers, arising from different data modalities, showing high correlation coefficients for some scores from pre-operative data only. It indicates that many clinical outcomes of DBS can be predicted agnostic to the specific simulation parameters, as PassFlow has been validated without such stimulation-related information.

Peralta Maxime, Haegelen Claire, Jannin Pierre, Baxter John S H

2021-Jul-03

Clinical prediction, Deep brain stimulation, Machine learning, Parkinson’s disease

In NeuroImage. Clinical

Chen Ji, Wensing Tobias, Hoffstaedter Felix, Cieslik Edna C, Müller Veronika I, Patil Kaustubh R, Aleman André, Derntl Birgit, Gruber Oliver, Jardri Renaud, Kogler Lydia, Sommer Iris E, Eickhoff Simon B, Nickl-Jockschat Thomas

2021

Formal thought disorder, Machine learning, Neuroimaging

In Drug discovery today ; h5-index 68.0

Liu Zhichao, Roberts Ruth A, Lal-Nag Madhu, Chen Xi, Huang Ruili, Tong Weida

2021-Jun-30

Artificial intelligence, COVID-19, Drug development, Drug discovery, Language models, Natural language processing

In Carbon balance and management

BACKGROUND : Forests provide the largest terrestrial sink of carbon (C). However, these C stocks are threatened by forest land conversion. Land use change has global impacts and is a critical component when studying C fluxes, but it is not always fully considered in C accounting despite being a major contributor to emissions. An urgent need exists among decision-makers to identify the likelihood of forest conversion to other land uses and factors affecting C loss. To help address this issue, we conducted our research in California, Colorado, Georgia, New York, Texas, and Wisconsin. The objectives were to (1) model the probability of forest conversion and C stocks dynamics using USDA Forest Service Forest Inventory and Analysis (FIA) data and (2) create wall-to-wall maps showing estimates of the risk of areas to convert from forest to non-forest. We used two modeling approaches: a machine learning algorithm (random forest) and generalized mixed-effects models. Explanatory variables for the models included ecological attributes, topography, census data, forest disturbances, and forest conditions. Model predictions and Landsat spectral information were used to produce wall-to-wall probability maps of forest change using Google Earth Engine.

RESULTS : During the study period (2000-2017), 3.4% of the analyzed FIA plots transitioned from forest to mixed or non-forested conditions. Results indicate that the change in land use from forests is more likely with increasing human population and housing growth rates. Furthermore, non-public forests showed a higher probability of forest change compared to public forests. Areas closer to cities and coastal areas showed a higher risk of transition to non-forests. Out of the six states analyzed, Colorado had the highest risk of conversion and the largest amount of aboveground C lost. Natural forest disturbances were not a major predictor of land use change.

CONCLUSIONS : Land use change is accelerating globally, causing a large increase in C emissions. Our results will help policy-makers prioritize forest management activities and land use planning by providing a quantitative framework that can enhance forest health and productivity. This work will also inform climate change mitigation strategies by understanding the role that land use change plays in C emissions.

Fitts Lucia A, Russell Matthew B, Domke Grant M, Knight Joseph K

2021-Jul-03

Carbon dynamics, Ecosystem services, Forest inventory, Forest loss drivers, Remote sensing, USDA Forest Inventory and Analysis (FIA) data

In European spine journal : official publication of the European Spine Society, the European Spinal Deformity Society, and the European Section of the Cervical Spine Research Society

PURPOSE : We developed a software program that automatically extracts a three-dimensional (3D) lumbar nerve root image from magnetic resonance imaging (MRI) lumbar nerve volume data using artificial intelligence. The aim of this study is to evaluate the morphology of Kambin's triangle in three dimensions based on an actual endoscopic transforaminal surgical approach using three-dimensional (3D) computed tomography (CT)/ magnetic resonance imaging (MRI) fusion images of the lumbar spine and nerve tissue.

METHODS : Three-dimensional lumbar spine/nerve images of 100 patients (31 males and 69 females; mean age, 66.8 years) were used to evaluate the relationship between the superior articular process (SAP), exiting nerve root (ENR), and dural canal at the L2/3, L3/4, and L4/5 levels at 45° and 60° approach angles.

RESULTS : The SAP-ENR distance at 60° was the greatest at L4/5 and was significantly greater at L2/3 and L4/5 than at L3/4 (P < 0.01, P < 0.01, respectively). The SAP-ENR distance at 45° was the greatest at L2/3, and it was larger in L2/3 and L4/5 than in L3/4 (P < 0.01, P < 0.01, respectively). The SAP-ENR distances at L4/5 were significantly greater at 60° than at 45° (P < 0.01). The dural canal was located within Kambin's triangle on the plane of the upper endplate of the lower vertebra at L2/3 in 41.5% of the cases and at L3/4 in 14% of the cases at 60° but not at L4/5.

CONCLUSION : The 3D lumbar spine/nerve image enabled a combined assessment of the positional relationship between the SAP, ENR, and dural canal to quantify the safety zone of practical endoscopic spinal surgery using a transforaminal approach. Three-dimensional lumbar spine/nerve images could be useful for examining parameters, including bones and nerves, to ensure the safety of surgery.

Yamada Katsuhisa, Nagahama Ken, Abe Yuichiro, Hyugaji Yoshinori, Takahata Masahiko, Iwasaki Norimasa

2021-Jul-03

Artificial intelligence, Endoscopic spinal surgery, Kambin’s triangle, Lumbar spine, Working zone