Medlea activity tracker

AI learning

​

Learning AI

Started learning bits of AI

Studying papers on AI usage for spirometry

Studied papers on AI usage for spirometry, seems recent studies have been able to do it

Analysis of lobe ratio with FEV

Did analysis of lobe volume ratio to preop FEV for Gemelli patients

Paper study on empirical FEV

Studied an interesting paper about empirical postop FEV from preop FEV

Weresp demo meeting

Weresp demo meeting with Satish

More tests with water as fluid

Tested for higher pressure difference by simulating with water. Same problem of error in BC

Literature search for empirical models of spirometry prediction

Found a paper where the POST op FEV is predicted as an empirical value of Lung expansion and other parameters. Had an idea to create something similar, where FVC predicted as function of lobe size and flow ratio

Proper sized geometry test for BC

Simulation and analysis of proper sized geometry.There is definitely some bug in the BC condition, have to check for larger pressure difference

Meeting with Simone

Demonstrate to SImone about bug in BC in Moebius

Documenting issue with pressure pressure BC

Tried to document the issue of pressure pressure BC both on a sample geometry and on the airway T1- T3 cases

Issue in pressure output in Meobius

Tried to find out the cause for pressure error. Moebius outputs an unphyiscal pressure at the end cells, which messes up the average. So now use the centerline pressure

Simulation for T2

Had huge problem in creating geometries for T2. Satisfied all conditions of watertightness, but due to a small fragment, it was giving problem in meshing. Finally was able to get the meshing done

Meeting with Simone

Meeting with Simone. Discussed about approach of Spanish group. Decided to use the inlet velocity as a calibration to get the pressure and then use it to get FEV etc.

Analysis of pressure vs flow simulation

Modified some points in the report as there was a bug in the scripts.

Airway segmentation using MONAI module

Simulating effects of grid, pressure difference

Analysis of POST results and pressure variation

Creating and modifying analysis scripts

Improved and corrected errors in the analysis scripts for analysing the flow rates, pressure and other parameters

Comparison of previous and new model

Created a report comparing the airflow distribution between previous airway+parenchymna model and current only airway model. Airway distribution shows signifcant changes from the normal CT volumetric based prediction

Framework for analyis fluxes from the simulations

1. Spent lot of time writing scripts. 2. Main issues in postprocessing - inlet/outlet faces are not aligned with axes, have to determine the proper normals and centroid from the VTP, then a cut plane a few mesh points inside the domain by moving along the surface normal. 2. Had to clip data in a spherical region and then take slice, otherwise extra regions get included in the slice

Simulations for airway model

1. Simulated T1 PRE and POST after modifying the STLs. 2. Crashing due to CFL happening, had to test many parameters and finally settled on Mach 0.01 with constant inflow BC

Setting up framework for creating simulations

Reading PhD thesis on spirometry simulation

PhD thesis of Fernandez compiles the papers on the lung spirometry simulation. Although the approach seemed a bit odd, have to think more about it!

Creating VTPs for simulation wiith Blender

Setting up simulation for new model

Meeting with Simone

Meeting to discuss about CFL crashing in pressure - pressure BC

Using Blender for VTP creation for simulation

Created the framework for tagging faces with Blender, then exporting to VTP for final configuration. Also created scripts to check watertight STLs, scaling operations and final FaceID scalar creation as per Moebius rule

Exploring Blender as an alternative for STL face tagging

Blender learning continued, steep learning curve, but very powerful and flexible

Exploring Blender as an alternative for STL face tagging

Learnt the basics of Blender from Youtube tutorials and materials, seemed most efficient in terms of STL face selection

Meeting

Meeting with Simone to discuss and setup Gmoebius on my system

Creating simulation setup from STL

Explored Gmoebius and tried to create the VTP for simulation. Lot of problems faced - 1. The selected traingles for tagging donot appear correctly (I think it is a display issue!) 2. Sometimes the substract mesh operation was not working properly. Overall, realised neither Meshlab nor Gmoebius will be good candidate for creating VTPs from STL, as the most critical and difficult part was the tagging of faces and both of them had below par selection interface. So need to find a software where STL manipulation/selection is most convenient.

Install gmoebius

Setup of airway simulation with pressure BCs

STL modification for inlet outlet modification

Spent time and figured out the complete steps to truncate STL, close holes, paint inlets/outlets. Finally export the STL with ‘color’ functions defined which can be used in Paraview to define FaceID scalar (necessary for Moebius meshing)

Airway segmentation for Gemelli POST

T26 took a lot of time to remove kinks on the lobe surfaces. Completed segmentation for T26 to T29. Spent lot of time on T30, but could not do it, as CT quality seemed to be poor, as well as orientation was tilted.

STL modification for inlet outlet modification

Explored different ways to label outlets from the airway STL. Thinking of Meshlab as an option, will have to study more. Maybe start with a truncated airway with only two bronchus ans see if it can tagged and readied for simulation.

Airway segmentation for Gemelli POST

Segmented PRE cases from T21 to T25. Needed time to run Lungmask on my system for all the cases (does not have GPU and no GUI access to remote machine for Slicer access on it) and then modify segmentations. Specially, case T23 and T25 took lot of time to remove kinks on the lobes surface.

Papers on spirometry through CFD

Collected multile cited papers and also PhD thesis of Fernandez, created summary in Onenote about spirometry, model and applied BCs.

Airway segmentation for Gemelli POST

Segmented cases from T26 to T29, had to redo T21 due to error. Tried T30, but lobe demarcation couldnot be identified, so failed to do it!

Group meeting with John (Demcon) and clinicians from Leiden

Detailed meeting on ventilators and possibilty of prediction of ventilator performance through CFD simulations. Lot of focus on tissue effects due to flow and also appopriate ways to model patient degeneration etc.

New model for lungs

Collected multiple papers on airflow and CFD simulation