Add files via upload

Author: Zhentao-Liu

ct/leap_ct.py

@@ -0,0 +1,75 @@
+from leapctype import *
+import numpy as np
+import time
+from leaptorch import Projector
+import copy
+from tqdm import tqdm
+import math
+import torch
+
+class leap_ct():
+    def __init__(self, cam_infos, recon_args, use_device='cuda:0'):
+        
+        self.use_device = use_device
+        self.device = torch.device(use_device)
+        self.projector = Projector(forward_project=True, use_static=True, use_gpu=True, gpu_device=self.device)
+        # self.projector.leapct.set_projector('VD')
+
+        # set up the projection geometry
+        numAngles = len(cam_infos)
+        numRows = cam_infos[0].height 
+        numCols = cam_infos[0].width
+        pixelHeight = cam_infos[0].sy
+        pixelWidth = cam_infos[0].sx
+        sod = cam_infos[0].sad
+        sdd = cam_infos[0].sid
+        centerRow = 0.5*(numRows-1)
+        centerCol = 0.5*(numCols-1)
+
+        projs = []
+        PrimaryAngles = []
+        for cam in cam_infos:
+            projs.append(cam.image.squeeze())
+            PrimaryAngles.append(cam.PrimaryAngle)
+        projs = np.stack(projs, axis=0)
+        PrimaryAngles = np.stack(PrimaryAngles, axis=0)
+        self.projs = np.ascontiguousarray(projs, dtype=np.float32)
+        PrimaryAngles = np.rad2deg(PrimaryAngles)
+        PrimaryAngles += 90.0   # align DSA imaging system with LEAP-toolbox convention
+        self.PrimaryAngles = np.ascontiguousarray(PrimaryAngles, dtype=np.float32)
+
+        self.projector.leapct.set_conebeam(numAngles, numRows, numCols, pixelHeight, pixelWidth, 
+                                           centerRow, centerCol, self.PrimaryAngles, sod, sdd)
+            
+        numX, numY, numZ = recon_args["volume_resolution"]
+        voxelWidth, voxelHeight = recon_args["volume_spacing"][0], recon_args["volume_spacing"][-1]
+
+        self.projector.leapct.set_volume(numX, numY, numZ, voxelWidth, voxelHeight)
+
+        self.projector.allocate_batch_data()   # necessary for torch version
+
+    def fdk(self, projs_, PrimaryAngles_, VERSE=False):
+        if VERSE:
+            print('Reconstruction with LEAP-toolbox fdk...')
+
+        startTime = time.time()
+
+        projs = copy.deepcopy(projs_)   # do not let inner operation influence input
+        projs = np.flip(projs, axis=1)
+        g = np.ascontiguousarray(projs, dtype=np.float32)   # shape is numAngles, numRows, numCols
+
+        PrimaryAngles = copy.deepcopy(PrimaryAngles_)
+        self.projector.leapct.set_phis(PrimaryAngles)
+        self.projector.allocate_batch_data()
+
+        f = self.projector.leapct.allocate_volume() # shape is numZ, numY, numX
+        f[:] = 0.0 # initialize the volume to zero
+        self.projector.leapct.FBP(g, f)
+
+        if VERSE:
+            print('fdk Reconstruction Elapsed Time: ' + str(time.time()-startTime))
+
+        return f
+
+
+