Deep Learning for Automated Measurement of Total Cardiac Volume for Heart Transplantation Size Matching.

Total Cardiac Volume (TCV)-based size matching using Computed Tomography (CT) is a novel technique to compare donor and recipient heart size in pediatric heart transplant that may increase overall utilization of available grafts. TCV requires manual segmentation, which limits its widespread use due to time and specialized software and training needed for segmentation. This study aims to determine the accuracy of a Deep Learning (DL) approach using 3-dimensional Convolutional Neural Networks (3D-CNN) to calculate TCV, with the clinical aim of enabling fast and accurate TCV use at all transplant centers. Ground truth TCV was segmented on CT scans of subjects aged 0-30 years, identified retrospectively. Ground truth segmentation masks were used to train and test a custom 3D-CNN model consisting of a DenseNet architecture in combination with residual blocks of ResNet architecture. The model was trained on a cohort of 270 subjects and a validation cohort of 44 subjects (36 normal, 8 heart disease retained for model testing). The average Dice similarity coefficient of the validation cohort was 0.94 ± 0.03 (range 0.84-0.97). The mean absolute percent error of TCV estimation was 5.5%. There is no significant association between model accuracy and subject age, weight, or height. DL-TCV was on average more accurate for normal hearts than those listed for transplant (mean absolute percent error 4.5 ± 3.9 vs. 10.5 ± 8.5, p = 0.08). A deep learning-based 3D-CNN model can provide accurate automatic measurement of TCV from CT images. This initial study is limited as a single-center study, though future multicenter studies may enable generalizable and more accurate TCV measurement by inclusion of more diverse cardiac pathology and increasing the training data.

Deep Learning for Automated Measurement of Total Cardiac Volume for Heart Transplantation Size Matching.

Background: Total Cardiac Volume (TCV) based size matching using Computed Tomography (CT) is a novel technique to compare donor and recipient heart size in pediatric heart transplant that may increase overall utilization of available grafts. TCV requires manual segmentation, which limits its widespread use due to time and specialized software and training needed for segmentation. Objective: This study aims to determine the accuracy of a Deep Learning (DL) approach using 3-dimensional Convolutional Neural Networks (3D-CNN) to calculate TCV, with the clinical aim of enabling fast and accurate TCV use at all transplant centers. Materials and Methods: Ground truth TCV was segmented on CT scans of subjects aged 0-30 years, identified retrospectively. Ground truth segmentation masks were used to train and test a custom 3D-CNN model consisting of a Dense-Net architecture in combination with residual blocks of ResNet architecture. Results: The model was trained on a cohort of 270 subjects and a validation cohort of 44 subjects (36 normal, 8 heart disease retained for model testing). The average Dice similarity coefficient of the validation cohort was 0.94 ± 0.03 (range 0.84-0.97). The mean absolute percent error of TCV estimation was 5.5%. There is no significant association between model accuracy and subject age, weight, or height. DL-TCV was on average more accurate for normal hearts than those listed for transplant (mean absolute percent error 4.5 ± 3.9 vs. 10.5 ± 8.5, p = 0.08). Conclusion: A deep learning based 3D-CNN model can provide accurate automatic measurement of TCV from CT images.

Energetics of peptide (pHLIP) binding to and folding across a lipid bilayer membrane.

The pH low-insertion peptide (pHLIP) serves as a model system for peptide insertion and folding across a lipid bilayer. It has three general states: (I) soluble in water or (II) bound to the surface of a lipid bilayer as an unstructured monomer, and (III) inserted across the bilayer as a monomeric alpha-helix. We used fluorescence spectroscopy and isothermal titration calorimetry to study the interactions of pHLIP with a palmitoyloleoylphosphatidylcholine (POPC) lipid bilayer and to calculate the transition energies between states. We found that the Gibbs free energy of binding to a POPC surface at low pHLIP concentration (state I-state II transition) at 37 degrees C is approximately -7 kcal/mol near neutral pH and that the free energy of insertion and folding across a lipid bilayer at low pH (state II-state III transition) is nearly -2 kcal/mol. We discuss a number of related thermodynamic parameters from our measurements. Besides its fundamental interest as a model system for the study of membrane protein folding, pHLIP has utility as an agent to target diseased tissues and translocate molecules through the membrane into the cytoplasm of cells in environments with elevated levels of extracellular acidity, as in cancer and inflammation. The results give the amount of energy that might be used to move cargo molecules across a membrane.

A monomeric membrane peptide that lives in three worlds: in solution, attached to, and inserted across lipid bilayers.

The membrane peptide pH (low) insertion peptide (pHLIP) lives in three worlds, being soluble in aqueous solution at pH 7.4, binding to the surface of lipid bilayers, and inserting as a transbilayer helix at low pH. With low pH driving the process, pHLIP can translocate cargo molecules attached to its C-terminus via a disulfide and release them in the cytoplasm of a cell. Here we examine a key aspect of the mechanism, showing that pHLIP is monomeric in each of its three major states: soluble in water near neutral pH (state I), bound to the surface of a membrane near neutral pH (state II), and inserted across the membrane as an alpha-helix at low pH (state III). The peptide does not induce fusion or membrane leakage. The unique properties of pHLIP made it attractive for the biophysical investigation of membrane protein folding in vitro and for the development of a novel class of delivery peptides for the transport of therapeutic and diagnostic agents to acidic tissue sites associated with various pathological processes in vivo.

Mechanism and uses of a membrane peptide that targets tumors and other acidic tissues in vivo.

The pH-selective insertion and folding of a membrane peptide, pHLIP [pH (low) insertion peptide], can be used to target acidic tissue in vivo, including acidic foci in tumors, kidneys, and inflammatory sites. In a mouse breast adenocarcinoma model, fluorescently labeled pHLIP finds solid acidic tumors with high accuracy and accumulates in them even at a very early stage of tumor development. The fluorescence signal is stable for >4 days and is approximately five times higher in tumors than in healthy counterpart tissue. In a rat antigen-induced arthritis model, pHLIP preferentially accumulates in inflammatory foci. pHLIP also maps the renal cortical interstitium; however, kidney accumulation can be reduced significantly by providing mice with bicarbonate-containing drinking water. The peptide has three states: soluble in water, bound to the surface of a membrane, and inserted across the membrane as an alpha-helix. At physiological pH, the equilibrium is toward water, which explains its low affinity for cells in healthy tissue; at acidic pH, titration of Asp residues shifts the equilibrium toward membrane insertion and tissue accumulation. The replacement of two key Asp residues located in the transmembrane part of pHLIP by Lys or Asn led to the loss of pH-sensitive insertion into membranes of liposomes, red blood cells, and cancer cells in vivo, as well as to the loss of specific accumulation in tumors. pHLIP nanotechnology introduces a new method of detecting, targeting, and possibly treating acidic diseased tissue by using the selective insertion and folding of membrane peptides.