Migrated from GitHub

data/.mdai/config.yaml

@@ -0,0 +1,2 @@
+base_image: py37
+device_type: gpu

data/.mdai/helper.py

@@ -0,0 +1,67 @@
+import numpy as np
+from skimage.measure import find_contours
+
+
+def get_values_r231(outmask, ds):
+    contours = find_contours(outmask, 0)
+    if contours:
+        outputs = []
+        lcounter, rcounter = 0, 1
+        for contour in contours:
+            data = {"vertices": [[(v[1]), (v[0])] for v in contour.tolist()]}
+            if contour[0][1] <= 250:
+                output = {
+                    "type": "ANNOTATION",
+                    "study_uid": str(ds.StudyInstanceUID),
+                    "series_uid": str(ds.SeriesInstanceUID),
+                    "instance_uid": str(ds.SOPInstanceUID),
+                    "class_index": rcounter,
+                    "data": data,
+                }
+                rcounter = 2
+            else:
+                output = {
+                    "type": "ANNOTATION",
+                    "study_uid": str(ds.StudyInstanceUID),
+                    "series_uid": str(ds.SeriesInstanceUID),
+                    "instance_uid": str(ds.SOPInstanceUID),
+                    "class_index": lcounter,
+                    "data": data,
+                }
+                lcounter = 2
+            outputs.append(output)
+    else:
+        outputs = [
+            {
+                "type": "NONE",
+                "study_uid": str(ds.StudyInstanceUID),
+                "series_uid": str(ds.SeriesInstanceUID),
+                "instance_uid": str(ds.SOPInstanceUID),
+            }
+        ]
+    return outputs
+
+
+def get_values_ltrclobes(masks, ds):
+    if masks:
+        preds = []
+        for submask, label in masks:
+            output = {
+                "type": "ANNOTATION",
+                "study_uid": str(ds.StudyInstanceUID),
+                "series_uid": str(ds.SeriesInstanceUID),
+                "instance_uid": str(ds.SOPInstanceUID),
+                "class_index": int(label),
+                "data": {"mask": submask.tolist()},
+            }
+            preds.append(output)
+    else:
+        preds = [
+            {
+                "type": "NONE",
+                "study_uid": str(ds.StudyInstanceUID),
+                "series_uid": str(ds.SeriesInstanceUID),
+                "instance_uid": str(ds.SOPInstanceUID),
+            }
+        ]
+    return preds

data/.mdai/mask.py

@@ -0,0 +1,100 @@
+import os
+import numpy as np
+import torch
+import warnings
+import sys
+import logging
+
+from resunet import UNet
+from utils import preprocess, postrocessing, reshape_mask
+
+logging.basicConfig(stream=sys.stdout, level=logging.INFO)
+warnings.filterwarnings("ignore", category=UserWarning)
+
+# stores urls and number of classes of the models
+model_urls = {
+    ("unet", "r231"): ("unet_r231-d5d2fc3d.pth", 3),
+    ("unet", "ltrclobes"): ("unet_ltrclobes-3a07043d.pth", 6),
+    ("unet", "r231covidweb"): ("unet_r231covid-0de78a7e.pth", 3),
+}
+
+
+def apply(image, model, device, volume_postprocessing=True):
+    tvolslices, xnew_box = preprocess(image, resolution=[256, 256])
+    tvolslices[tvolslices > 600] = 600
+    tvolslices = np.divide((tvolslices + 1024), 1624)
+    timage_res = np.empty((np.append(0, tvolslices[0].shape)), dtype=np.uint8)
+
+    with torch.no_grad():
+        X = torch.Tensor(tvolslices).unsqueeze(0).to(device)
+        prediction = model(X)
+        pls = torch.max(prediction, 1)[1].detach().cpu().numpy().astype(np.uint8)
+        timage_res = np.vstack((timage_res, pls))
+
+    # postprocessing includes removal of small connected components, hole filling and mapping of small components to
+    # neighbors
+    if volume_postprocessing:
+        outmask = postrocessing(timage_res)
+    else:
+        outmask = timage_res
+
+    outmask = np.asarray(
+        [
+            reshape_mask(outmask[i], xnew_box[i], image.shape[1:])
+            for i in range(outmask.shape[0])
+        ],
+        dtype=np.uint8,
+    )
+
+    return outmask.astype(np.uint8)
+
+
+def get_model(modeltype, modelname, modelpath, device):
+    model_url, n_classes = model_urls[(modeltype, modelname)]
+    model_file = os.path.join(modelpath, model_url)
+    if device.type == "cpu":
+        state_dict = torch.load(model_file, map_location=torch.device("cpu"))
+    else:
+        state_dict = torch.load(model_file)
+
+    if modeltype == "unet":
+        model = UNet(
+            n_classes=n_classes,
+            padding=True,
+            depth=5,
+            up_mode="upsample",
+            batch_norm=True,
+            residual=False,
+        )
+    elif modeltype == "resunet":
+        model = UNet(
+            n_classes=n_classes,
+            padding=True,
+            depth=5,
+            up_mode="upsample",
+            batch_norm=True,
+            residual=True,
+        )
+    else:
+        logging.exception(f"Model {modelname} not known")
+    model.load_state_dict(state_dict)
+    model.eval()
+    return model
+
+
+"""
+def apply_fused(image, basemodel = 'LTRCLobes', fillmodel = 'R231', volume_postprocessing=True):
+    # Will apply basemodel and use fillmodel to mitiage false negatives
+    mdl_r = get_model('unet',fillmodel)
+    mdl_l = get_model('unet',basemodel)
+    logging.info("Apply: %s" % basemodel)
+    res_l = apply(image, mdl_l, force_cpu=force_cpu, batch_size=batch_size,  volume_postprocessing=volume_postprocessing, noHU=noHU)
+    logging.info("Apply: %s" % fillmodel)
+    res_r = apply(image, mdl_r, force_cpu=force_cpu, batch_size=batch_size,  volume_postprocessing=volume_postprocessing, noHU=noHU)
+    spare_value = res_l.max()+1
+    res_l[np.logical_and(res_l==0, res_r>0)] = spare_value
+    res_l[res_r==0] = 0
+    logging.info("Fusing results... this may take up to several minutes!")
+    return utils.postrocessing(res_l, spare=[spare_value])
+"""
+

data/.mdai/mdai_deploy.py

@@ -0,0 +1,72 @@
+import os
+from io import BytesIO
+import numpy as np
+import torch
+import pydicom
+
+from mask import get_model, apply
+from helper import get_values_r231, get_values_ltrclobes
+
+# Currently supported model names - R231, LTRCLobes, R231CovidWeb
+args = {
+    "model_type": "unet",
+    "model_name": "R231",
+    "postprocess": True,
+}
+
+
+class MDAIModel:
+    def __init__(self):
+        self.model_type = args.get("model_type", "unet").lower()
+        self.model_name = args.get("model_name").lower()
+        self.postprocessing = args.get("postprocess", True)
+
+        root_path = os.path.dirname(os.path.dirname(__file__))
+        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+        self.model = get_model(self.model_type, self.model_name, root_path, self.device)
+        self.model.to(self.device)
+
+    def predict(self, data):
+        input_files = data["files"]
+        input_annotations = data["annotations"]
+        input_args = data["args"]
+
+        outputs = []
+
+        for file in input_files:
+            if file["content_type"] != "application/dicom":
+                continue
+
+            ds = pydicom.dcmread(BytesIO(file["content"]))
+            image = ds.pixel_array
+
+            if image.max() > 4095 or image.min() < -2000:
+                image = np.int16(
+                    (image - image.min())
+                    * ((4095 + 1024) / (image.max() - image.min()))
+                    - 1024
+                )
+            else:
+                image = np.int16(image) - 1024
+
+            if len(image.shape) == 2:
+                image = np.expand_dims(image, 0)
+
+            outmask = apply(
+                image,
+                self.model,
+                self.device,
+                volume_postprocessing=self.postprocessing,
+            )
+            outmask = np.uint8(outmask.squeeze(0))
+
+            if self.model_name == "ltrclobes":
+                vals = set(np.unique(outmask))
+                masks = [
+                    (np.uint8(outmask) == i, i - 1) for i in range(1, 6) if i in vals
+                ]
+                outputs += get_values_ltrclobes(masks, ds)
+            else:
+                outputs += get_values_r231(outmask, ds)
+        return outputs

data/.mdai/requirements.txt

@@ -0,0 +1,5 @@
+scipy==1.5.2
+scikit-image==0.16.2
+torch==1.6.0
+pillow==7.2.0
+fill_voids

data/.mdai/resunet.py

@@ -0,0 +1,152 @@
+import torch
+from torch import nn
+import torch.nn.functional as F
+
+class UNet(nn.Module):
+    def __init__(self, in_channels=1, n_classes=2, depth=5, wf=6, padding=False,
+                 batch_norm=False, up_mode='upconv', residual=False):
+        """
+        Implementation of
+        U-Net: Convolutional Networks for Biomedical Image Segmentation
+        (Ronneberger et al., 2015)
+        https://arxiv.org/abs/1505.04597
+        Using the default arguments will yield the exact version used
+        in the original paper
+        Args:
+            in_channels (int): number of input channels
+            n_classes (int): number of output channels
+            depth (int): depth of the network
+            wf (int): number of filters in the first layer is 2**wf
+            padding (bool): if True, apply padding such that the input shape
+                            is the same as the output.
+                            This may introduce artifacts
+            batch_norm (bool): Use BatchNorm after layers with an
+                               activation function
+            up_mode (str): one of 'upconv' or 'upsample'.
+                           'upconv' will use transposed convolutions for
+                           learned upsampling.
+                           'upsample' will use bilinear upsampling.
+            residual: if True, residual connections will be added
+        """
+        super(UNet, self).__init__()
+        assert up_mode in ('upconv', 'upsample')
+        self.padding = padding
+        self.depth = depth
+        prev_channels = in_channels
+        self.down_path = nn.ModuleList()
+        for i in range(depth):
+            if i == 0 and residual:
+                self.down_path.append(UNetConvBlock(prev_channels, 2 ** (wf + i),
+                                                    padding, batch_norm, residual, first=True))
+            else:
+                self.down_path.append(UNetConvBlock(prev_channels, 2 ** (wf + i),
+                                                    padding, batch_norm, residual))
+            prev_channels = 2 ** (wf + i)
+
+        self.up_path = nn.ModuleList()
+        for i in reversed(range(depth - 1)):
+            self.up_path.append(UNetUpBlock(prev_channels, 2 ** (wf + i), up_mode,
+                                            padding, batch_norm, residual))
+            prev_channels = 2 ** (wf + i)
+
+        self.last = nn.Conv2d(prev_channels, n_classes, kernel_size=1)
+        self.softmax = nn.LogSoftmax(dim=1)
+
+    def forward(self, x):
+        blocks = []
+        for i, down in enumerate(self.down_path):
+            x = down(x)
+            if i != len(self.down_path) - 1:
+                blocks.append(x)
+                x = F.avg_pool2d(x, 2)
+
+        for i, up in enumerate(self.up_path):
+            x = up(x, blocks[-i - 1])
+
+        res = self.last(x)
+        return self.softmax(res)
+
+
+class UNetConvBlock(nn.Module):
+    def __init__(self, in_size, out_size, padding, batch_norm, residual=False, first=False):
+        super(UNetConvBlock, self).__init__()
+        self.residual = residual
+        self.out_size = out_size
+        self.in_size = in_size
+        self.batch_norm = batch_norm
+        self.first = first
+        self.residual_input_conv = nn.Conv2d(self.in_size, self.out_size, kernel_size=1)
+        self.residual_batchnorm = nn.BatchNorm2d(self.out_size)
+
+        if residual:
+            padding = 1
+        block = []
+
+        if residual and not first:
+            block.append(nn.ReLU())
+            if batch_norm:
+                block.append(nn.BatchNorm2d(in_size))
+
+        block.append(nn.Conv2d(in_size, out_size, kernel_size=3,
+                               padding=int(padding)))
+        block.append(nn.ReLU())
+        if batch_norm:
+            block.append(nn.BatchNorm2d(out_size))
+
+        block.append(nn.Conv2d(out_size, out_size, kernel_size=3,
+                               padding=int(padding)))
+
+        if not residual:
+            block.append(nn.ReLU())
+            if batch_norm:
+                block.append(nn.BatchNorm2d(out_size))
+        self.block = nn.Sequential(*block)
+
+    def forward(self, x):
+        out = self.block(x)
+        if self.residual:
+            if self.in_size != self.out_size:
+                x = self.residual_input_conv(x)
+                x = self.residual_batchnorm(x)
+            out = out + x
+
+        return out
+
+
+class UNetUpBlock(nn.Module):
+    def __init__(self, in_size, out_size, up_mode, padding, batch_norm, residual=False):
+        super(UNetUpBlock, self).__init__()
+        self.residual = residual
+        self.in_size = in_size
+        self.out_size = out_size
+        self.residual_input_conv = nn.Conv2d(self.in_size, self.out_size, kernel_size=1)
+        self.residual_batchnorm = nn.BatchNorm2d(self.out_size)
+
+        if up_mode == 'upconv':
+            self.up = nn.ConvTranspose2d(in_size, out_size, kernel_size=2,
+                                         stride=2)
+        elif up_mode == 'upsample':
+            self.up = nn.Sequential(nn.Upsample(mode='bilinear', scale_factor=2),
+                                    nn.Conv2d(in_size, out_size, kernel_size=1))
+
+        self.conv_block = UNetConvBlock(in_size, out_size, padding, batch_norm)
+
+    @staticmethod
+    def center_crop(layer, target_size):
+        _, _, layer_height, layer_width = layer.size()
+        diff_y = (layer_height - target_size[0]) // 2
+        diff_x = (layer_width - target_size[1]) // 2
+        return layer[:, :, diff_y:(diff_y + target_size[0]), diff_x:(diff_x + target_size[1])]
+
+    def forward(self, x, bridge):
+        up = self.up(x)
+        crop1 = self.center_crop(bridge, up.shape[2:])
+        out_orig = torch.cat([up, crop1], 1)
+        out = self.conv_block(out_orig)
+        if self.residual:
+            if self.in_size != self.out_size:
+                out_orig = self.residual_input_conv(out_orig)
+                out_orig = self.residual_batchnorm(out_orig)
+            out = out + out_orig
+
+        return out

data/.mdai/utils.py

@@ -0,0 +1,162 @@
+import scipy.ndimage as ndimage
+import skimage.measure
+import numpy as np
+import fill_voids
+import skimage.morphology
+
+
+def preprocess(img, label=None, resolution=[192, 192]):
+    imgmtx = np.copy(img)
+    lblsmtx = np.copy(label)
+
+    imgmtx[imgmtx < -1024] = -1024
+    imgmtx[imgmtx > 600] = 600
+    cip_xnew = []
+    cip_box = []
+    cip_mask = []
+    for i in range(imgmtx.shape[0]):
+        if label is None:
+            (im, m, box) = crop_and_resize(imgmtx[i, :, :], width=resolution[0], height=resolution[1])
+        else:
+            (im, m, box) = crop_and_resize(imgmtx[i, :, :], mask=lblsmtx[i, :, :], width=resolution[0],
+                                           height=resolution[1])
+            cip_mask.append(m)
+        cip_xnew.append(im)
+        cip_box.append(box)
+    if label is None:
+        return np.asarray(cip_xnew), cip_box
+    else:
+        return np.asarray(cip_xnew), cip_box, np.asarray(cip_mask)
+
+
+def simple_bodymask(img):
+    maskthreshold = -500
+    oshape = img.shape
+    img = ndimage.zoom(img, 128/np.asarray(img.shape), order=0)
+    bodymask = img > maskthreshold
+    bodymask = ndimage.binary_closing(bodymask)
+    bodymask = ndimage.binary_fill_holes(bodymask, structure=np.ones((3, 3))).astype(int)
+    bodymask = ndimage.binary_erosion(bodymask, iterations=2)
+    bodymask = skimage.measure.label(bodymask.astype(int), connectivity=1)
+    regions = skimage.measure.regionprops(bodymask.astype(int))
+    if len(regions) > 0:
+        max_region = np.argmax(list(map(lambda x: x.area, regions))) + 1
+        bodymask = bodymask == max_region
+        bodymask = ndimage.binary_dilation(bodymask, iterations=2)
+    real_scaling = np.asarray(oshape)/128
+    return ndimage.zoom(bodymask, real_scaling, order=0)
+
+
+def crop_and_resize(img, mask=None, width=192, height=192):
+    bmask = simple_bodymask(img)
+    # img[bmask==0] = -1024 # this line removes background outside of the lung.
+    # However, it has been shown problematic with narrow circular field of views that touch the lung.
+    # Possibly doing more harm than help
+    reg = skimage.measure.regionprops(skimage.measure.label(bmask))
+    if len(reg) > 0:
+        bbox = np.asarray(reg[0].bbox)
+    else:
+        bbox = (0, 0, bmask.shape[0], bmask.shape[1])
+    img = img[bbox[0]:bbox[2], bbox[1]:bbox[3]]
+    img = ndimage.zoom(img, np.asarray([width, height]) / np.asarray(img.shape), order=1)
+    if not mask is None:
+        mask = mask[bbox[0]:bbox[2], bbox[1]:bbox[3]]
+        mask = ndimage.zoom(mask, np.asarray([width, height]) / np.asarray(mask.shape), order=0)
+        # mask = ndimage.binary_closing(mask,iterations=5)
+    return img, mask, bbox
+
+
+def reshape_mask(mask, tbox, origsize):
+    res = np.ones(origsize) * 0
+    resize = [tbox[2] - tbox[0], tbox[3] - tbox[1]]
+    imgres = ndimage.zoom(mask, resize / np.asarray(mask.shape), order=0)
+    res[tbox[0]:tbox[2], tbox[1]:tbox[3]] = imgres
+    return res
+
+
+def postrocessing(label_image, spare=[]):
+    '''some post-processing mapping small label patches to the neighbout whith which they share the
+        largest border. All connected components smaller than min_area will be removed
+    '''
+
+    # merge small components to neighbours
+    regionmask = skimage.measure.label(label_image)
+    origlabels = np.unique(label_image)
+    origlabels_maxsub = np.zeros((max(origlabels) + 1,), dtype=np.uint32)  # will hold the largest component for a label
+    regions = skimage.measure.regionprops(regionmask, label_image)
+    regions.sort(key=lambda x: x.area)
+    regionlabels = [x.label for x in regions]
+
+    # will hold mapping from regionlabels to original labels
+    region_to_lobemap = np.zeros((len(regionlabels) + 1,), dtype=np.uint8)
+    for r in regions:
+        if r.area > origlabels_maxsub[r.max_intensity]:
+            origlabels_maxsub[r.max_intensity] = r.area
+            region_to_lobemap[r.label] = r.max_intensity
+
+    for r in regions:
+        if (r.area < origlabels_maxsub[r.max_intensity] or r.max_intensity in spare) and r.area>2: # area>2 improves runtime because small areas 1 and 2 voxel will be ignored
+            bb = bbox_3D(regionmask == r.label)
+            sub = regionmask[bb[0]:bb[1], bb[2]:bb[3], bb[4]:bb[5]]
+            dil = ndimage.binary_dilation(sub == r.label)
+            neighbours, counts = np.unique(sub[dil], return_counts=True)
+            mapto = r.label
+            maxmap = 0
+            myarea = 0
+            for ix, n in enumerate(neighbours):
+                if n != 0 and n != r.label and counts[ix] > maxmap and n != spare:
+                    maxmap = counts[ix]
+                    mapto = n
+                    myarea = r.area
+            regionmask[regionmask == r.label] = mapto
+            # print(str(region_to_lobemap[r.label]) + ' -> ' + str(region_to_lobemap[mapto])) # for debugging
+            if regions[regionlabels.index(mapto)].area == origlabels_maxsub[
+                regions[regionlabels.index(mapto)].max_intensity]:
+                origlabels_maxsub[regions[regionlabels.index(mapto)].max_intensity] += myarea
+            regions[regionlabels.index(mapto)].__dict__['_cache']['area'] += myarea
+
+    outmask_mapped = region_to_lobemap[regionmask]
+    outmask_mapped[outmask_mapped==spare] = 0 
+
+    if outmask_mapped.shape[0] == 1:
+        # holefiller = lambda x: ndimage.morphology.binary_fill_holes(x[0])[None, :, :] # This is bad for slices that show the liver
+        holefiller = lambda x: skimage.morphology.area_closing(x[0].astype(int), area_threshold=64)[None, :, :] == 1
+    else:
+        holefiller = fill_voids.fill
+
+    outmask = np.zeros(outmask_mapped.shape, dtype=np.uint8)
+    for i in np.unique(outmask_mapped)[1:]:
+        outmask[holefiller(keep_largest_connected_component(outmask_mapped == i))] = i
+
+    return outmask
+
+
+def bbox_3D(labelmap, margin=2):
+    shape = labelmap.shape
+    r = np.any(labelmap, axis=(1, 2))
+    c = np.any(labelmap, axis=(0, 2))
+    z = np.any(labelmap, axis=(0, 1))
+
+    rmin, rmax = np.where(r)[0][[0, -1]]
+    rmin -= margin if rmin >= margin else rmin
+    rmax += margin if rmax <= shape[0] - margin else rmax
+    cmin, cmax = np.where(c)[0][[0, -1]]
+    cmin -= margin if cmin >= margin else cmin
+    cmax += margin if cmax <= shape[1] - margin else cmax
+    zmin, zmax = np.where(z)[0][[0, -1]]
+    zmin -= margin if zmin >= margin else zmin
+    zmax += margin if zmax <= shape[2] - margin else zmax
+    
+    if rmax-rmin == 0:
+        rmax = rmin+1
+
+    return np.asarray([rmin, rmax, cmin, cmax, zmin, zmax])
+
+
+def keep_largest_connected_component(mask):
+    mask = skimage.measure.label(mask)
+    regions = skimage.measure.regionprops(mask)
+    resizes = np.asarray([x.area for x in regions])
+    max_region = np.argsort(resizes)[-1] + 1
+    mask = mask == max_region
+    return mask

data/LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

data/lungmask/__main__.py

@@ -0,0 +1,65 @@
+import sys
+import argparse
+import logging
+from lungmask import mask
+from lungmask import utils
+import os
+import SimpleITK as sitk
+import pkg_resources
+import numpy as np
+
+
+def path(string):
+    if os.path.exists(string):
+        return string
+    else:
+        sys.exit(f'File not found: {string}')
+
+
+def main():
+    version = pkg_resources.require("lungmask")[0].version
+    
+    parser = argparse.ArgumentParser()
+    parser.add_argument('input', metavar='input', type=path, help='Path to the input image, can be a folder for dicoms')
+    parser.add_argument('output', metavar='output', type=str, help='Filepath for output lungmask')
+    parser.add_argument('--modeltype', help='Default: unet', type=str, choices=['unet'], default='unet')
+    parser.add_argument('--modelname', help="spcifies the trained model, Default: R231", type=str, choices=['R231','LTRCLobes','LTRCLobes_R231','R231CovidWeb'], default='R231')
+    parser.add_argument('--cpu', help="Force using the CPU even when a GPU is available, will override batchsize to 1", action='store_true')
+    parser.add_argument('--nopostprocess', help="Deactivates postprocessing (removal of unconnected components and hole filling", action='store_true')
+    parser.add_argument('--noHU', help="For processing of images that are not encoded in hounsfield units (HU). E.g. png or jpg images from the web. Be aware, results may be substantially worse on these images", action='store_true')
+    parser.add_argument('--batchsize', type=int, help="Number of slices processed simultaneously. Lower number requires less memory but may be slower.", default=20)
+    parser.add_argument('--version', help="Shows the current version of lungmask", action='version', version=version)
+
+    argsin = sys.argv[1:]
+    args = parser.parse_args(argsin)
+    
+    batchsize = args.batchsize
+    if args.cpu:
+        batchsize = 1
+
+    logging.info(f'Load model')
+    
+    input_image = utils.get_input_image(args.input)
+    logging.info(f'Infer lungmask')
+    if args.modelname == 'LTRCLobes_R231':
+        result = mask.apply_fused(input_image, force_cpu=args.cpu, batch_size=batchsize, volume_postprocessing=not(args.nopostprocess), noHU=args.noHU)
+    else:
+        model = mask.get_model(args.modeltype, args.modelname)
+        result = mask.apply(input_image, model, force_cpu=args.cpu, batch_size=batchsize, volume_postprocessing=not(args.nopostprocess), noHU=args.noHU)
+        
+    if args.noHU:
+        file_ending = args.output.split('.')[-1]
+        print(file_ending)
+        if file_ending in ['jpg','jpeg','png']:
+            result = (result/(result.max())*255).astype(np.uint8)
+        result = result[0]
+             
+    result_out= sitk.GetImageFromArray(result)
+    result_out.CopyInformation(input_image)
+    logging.info(f'Save result to: {args.output}')
+    sys.exit(sitk.WriteImage(result_out, args.output))
+
+
+if __name__ == "__main__":
+    print('called as script')
+    main()

data/lungmask/mask.py

@@ -0,0 +1,116 @@
+import numpy as np
+import torch
+from lungmask import utils
+import SimpleITK as sitk
+from .resunet import UNet
+import warnings
+import sys
+from tqdm import tqdm
+import skimage
+import logging
+
+logging.basicConfig(stream=sys.stdout, level=logging.INFO)
+warnings.filterwarnings("ignore", category=UserWarning)
+
+# stores urls and number of classes of the models
+model_urls = {('unet', 'R231'): ('https://github.com/JoHof/lungmask/releases/download/v0.0/unet_r231-d5d2fc3d.pth', 3),
+              ('unet', 'LTRCLobes'): (
+                  'https://github.com/JoHof/lungmask/releases/download/v0.0/unet_ltrclobes-3a07043d.pth', 6),
+              ('unet', 'R231CovidWeb'): (
+                  'https://github.com/JoHof/lungmask/releases/download/v0.0/unet_r231covid-0de78a7e.pth', 3)}
+
+
+def apply(image, model=None, force_cpu=False, batch_size=20, volume_postprocessing=True, noHU=False):
+    if model is None:
+        model = get_model('unet', 'R231')
+
+    inimg_raw = sitk.GetArrayFromImage(image)
+    directions = np.asarray(image.GetDirection())
+    if len(directions) == 9:
+        inimg_raw = np.flip(inimg_raw, np.where(directions[[0,4,8]][::-1]<0)[0])
+    del image
+
+    if force_cpu:
+        device = torch.device('cpu')
+    else:
+        if torch.cuda.is_available():
+            device = torch.device('cuda')
+        else:
+            logging.info("No GPU support available, will use CPU. Note, that this is significantly slower!")
+            batch_size = 1
+            device = torch.device('cpu')
+    model.to(device)
+
+    
+    if not noHU:
+        tvolslices, xnew_box = utils.preprocess(inimg_raw, resolution=[256, 256])
+        tvolslices[tvolslices > 600] = 600
+        tvolslices = np.divide((tvolslices + 1024), 1624)
+    else:
+        # support for non HU images. This is just a hack. The models were not trained with this in mind
+        tvolslices = skimage.color.rgb2gray(inimg_raw)
+        tvolslices = skimage.transform.resize(tvolslices, [256, 256])
+        tvolslices = np.asarray([tvolslices*x for x in np.linspace(0.3,2,20)])
+        tvolslices[tvolslices>1] = 1
+        sanity = [(tvolslices[x]>0.6).sum()>25000 for x in range(len(tvolslices))]
+        tvolslices = tvolslices[sanity]
+    torch_ds_val = utils.LungLabelsDS_inf(tvolslices)
+    dataloader_val = torch.utils.data.DataLoader(torch_ds_val, batch_size=batch_size, shuffle=False, num_workers=1,
+                                                 pin_memory=False)
+
+    timage_res = np.empty((np.append(0, tvolslices[0].shape)), dtype=np.uint8)
+
+    with torch.no_grad():
+        for X in tqdm(dataloader_val):
+            X = X.float().to(device)
+            prediction = model(X)
+            pls = torch.max(prediction, 1)[1].detach().cpu().numpy().astype(np.uint8)
+            timage_res = np.vstack((timage_res, pls))
+
+    # postprocessing includes removal of small connected components, hole filling and mapping of small components to
+    # neighbors
+    if volume_postprocessing:
+        outmask = utils.postrocessing(timage_res)
+    else:
+        outmask = timage_res
+
+    if noHU:
+        outmask = skimage.transform.resize(outmask[np.argmax((outmask==1).sum(axis=(1,2)))], inimg_raw.shape[:2], order=0, anti_aliasing=False, preserve_range=True)[None,:,:]
+    else:
+         outmask = np.asarray(
+            [utils.reshape_mask(outmask[i], xnew_box[i], inimg_raw.shape[1:]) for i in range(outmask.shape[0])],
+            dtype=np.uint8)
+        
+    if len(directions) == 9:
+        outmask = np.flip(outmask, np.where(directions[[0,4,8]][::-1]<0)[0])
+
+    return outmask.astype(np.uint8)
+
+
+def get_model(modeltype, modelname):
+    model_url, n_classes = model_urls[(modeltype, modelname)]
+    state_dict = torch.hub.load_state_dict_from_url(model_url, progress=True, map_location=torch.device('cpu'))
+    if modeltype == 'unet':
+        model = UNet(n_classes=n_classes, padding=True, depth=5, up_mode='upsample', batch_norm=True, residual=False)
+    elif modeltype == 'resunet':
+        model = UNet(n_classes=n_classes, padding=True, depth=5, up_mode='upsample', batch_norm=True, residual=True)
+    else:
+        logging.exception(f"Model {modelname} not known")
+    model.load_state_dict(state_dict)
+    model.eval()
+    return model
+
+
+def apply_fused(image, basemodel = 'LTRCLobes', fillmodel = 'R231', force_cpu=False, batch_size=20, volume_postprocessing=True, noHU=False):
+    '''Will apply basemodel and use fillmodel to mitiage false negatives'''
+    mdl_r = get_model('unet',fillmodel)
+    mdl_l = get_model('unet',basemodel)
+    logging.info("Apply: %s" % basemodel)
+    res_l = apply(image, mdl_l, force_cpu=force_cpu, batch_size=batch_size,  volume_postprocessing=volume_postprocessing, noHU=noHU)
+    logging.info("Apply: %s" % fillmodel)
+    res_r = apply(image, mdl_r, force_cpu=force_cpu, batch_size=batch_size,  volume_postprocessing=volume_postprocessing, noHU=noHU)
+    spare_value = res_l.max()+1
+    res_l[np.logical_and(res_l==0, res_r>0)] = spare_value
+    res_l[res_r==0] = 0
+    logging.info("Fusing results... this may take up to several minutes!")
+    return utils.postrocessing(res_l, spare=[spare_value])

data/lungmask/resunet.py

@@ -0,0 +1,155 @@
+# Adapted from https://discuss.pytorch.org/t/unet-implementation/426
+
+import torch
+from torch import nn
+import torch.nn.functional as F
+
+
+class UNet(nn.Module):
+    def __init__(self, in_channels=1, n_classes=2, depth=5, wf=6, padding=False,
+                 batch_norm=False, up_mode='upconv', residual=False):
+        """
+        Implementation of
+        U-Net: Convolutional Networks for Biomedical Image Segmentation
+        (Ronneberger et al., 2015)
+        https://arxiv.org/abs/1505.04597
+        Using the default arguments will yield the exact version used
+        in the original paper
+        Args:
+            in_channels (int): number of input channels
+            n_classes (int): number of output channels
+            depth (int): depth of the network
+            wf (int): number of filters in the first layer is 2**wf
+            padding (bool): if True, apply padding such that the input shape
+                            is the same as the output.
+                            This may introduce artifacts
+            batch_norm (bool): Use BatchNorm after layers with an
+                               activation function
+            up_mode (str): one of 'upconv' or 'upsample'.
+                           'upconv' will use transposed convolutions for
+                           learned upsampling.
+                           'upsample' will use bilinear upsampling.
+            residual: if True, residual connections will be added
+        """
+        super(UNet, self).__init__()
+        assert up_mode in ('upconv', 'upsample')
+        self.padding = padding
+        self.depth = depth
+        prev_channels = in_channels
+        self.down_path = nn.ModuleList()
+        for i in range(depth):
+            if i == 0 and residual:
+                self.down_path.append(UNetConvBlock(prev_channels, 2 ** (wf + i),
+                                                    padding, batch_norm, residual, first=True))
+            else:
+                self.down_path.append(UNetConvBlock(prev_channels, 2 ** (wf + i),
+                                                    padding, batch_norm, residual))
+            prev_channels = 2 ** (wf + i)
+
+        self.up_path = nn.ModuleList()
+        for i in reversed(range(depth - 1)):
+            self.up_path.append(UNetUpBlock(prev_channels, 2 ** (wf + i), up_mode,
+                                            padding, batch_norm, residual))
+            prev_channels = 2 ** (wf + i)
+
+        self.last = nn.Conv2d(prev_channels, n_classes, kernel_size=1)
+        self.softmax = nn.LogSoftmax(dim=1)
+
+    def forward(self, x):
+        blocks = []
+        for i, down in enumerate(self.down_path):
+            x = down(x)
+            if i != len(self.down_path) - 1:
+                blocks.append(x)
+                x = F.avg_pool2d(x, 2)
+
+        for i, up in enumerate(self.up_path):
+            x = up(x, blocks[-i - 1])
+
+        res = self.last(x)
+        return self.softmax(res)
+
+
+class UNetConvBlock(nn.Module):
+    def __init__(self, in_size, out_size, padding, batch_norm, residual=False, first=False):
+        super(UNetConvBlock, self).__init__()
+        self.residual = residual
+        self.out_size = out_size
+        self.in_size = in_size
+        self.batch_norm = batch_norm
+        self.first = first
+        self.residual_input_conv = nn.Conv2d(self.in_size, self.out_size, kernel_size=1)
+        self.residual_batchnorm = nn.BatchNorm2d(self.out_size)
+
+        if residual:
+            padding = 1
+        block = []
+
+        if residual and not first:
+            block.append(nn.ReLU())
+            if batch_norm:
+                block.append(nn.BatchNorm2d(in_size))
+
+        block.append(nn.Conv2d(in_size, out_size, kernel_size=3,
+                               padding=int(padding)))
+        block.append(nn.ReLU())
+        if batch_norm:
+            block.append(nn.BatchNorm2d(out_size))
+
+        block.append(nn.Conv2d(out_size, out_size, kernel_size=3,
+                               padding=int(padding)))
+
+        if not residual:
+            block.append(nn.ReLU())
+            if batch_norm:
+                block.append(nn.BatchNorm2d(out_size))
+        self.block = nn.Sequential(*block)
+
+    def forward(self, x):
+        out = self.block(x)
+        if self.residual:
+            if self.in_size != self.out_size:
+                x = self.residual_input_conv(x)
+                x = self.residual_batchnorm(x)
+            out = out + x
+
+        return out
+
+
+class UNetUpBlock(nn.Module):
+    def __init__(self, in_size, out_size, up_mode, padding, batch_norm, residual=False):
+        super(UNetUpBlock, self).__init__()
+        self.residual = residual
+        self.in_size = in_size
+        self.out_size = out_size
+        self.residual_input_conv = nn.Conv2d(self.in_size, self.out_size, kernel_size=1)
+        self.residual_batchnorm = nn.BatchNorm2d(self.out_size)
+
+        if up_mode == 'upconv':
+            self.up = nn.ConvTranspose2d(in_size, out_size, kernel_size=2,
+                                         stride=2)
+        elif up_mode == 'upsample':
+            self.up = nn.Sequential(nn.Upsample(mode='bilinear', scale_factor=2),
+                                    nn.Conv2d(in_size, out_size, kernel_size=1))
+
+        self.conv_block = UNetConvBlock(in_size, out_size, padding, batch_norm)
+
+    @staticmethod
+    def center_crop(layer, target_size):
+        _, _, layer_height, layer_width = layer.size()
+        diff_y = (layer_height - target_size[0]) // 2
+        diff_x = (layer_width - target_size[1]) // 2
+        return layer[:, :, diff_y:(diff_y + target_size[0]), diff_x:(diff_x + target_size[1])]
+
+    def forward(self, x, bridge):
+        up = self.up(x)
+        crop1 = self.center_crop(bridge, up.shape[2:])
+        out_orig = torch.cat([up, crop1], 1)
+        out = self.conv_block(out_orig)
+        if self.residual:
+            if self.in_size != self.out_size:
+                out_orig = self.residual_input_conv(out_orig)
+                out_orig = self.residual_batchnorm(out_orig)
+            out = out + out_orig
+
+        return out

data/lungmask/utils.py

@@ -0,0 +1,279 @@
+import scipy.ndimage as ndimage
+import skimage.measure
+import numpy as np
+from torch.utils.data import Dataset
+import os
+import sys
+import SimpleITK as sitk
+import pydicom as pyd
+import logging
+from tqdm import tqdm
+import fill_voids
+import skimage.morphology
+
+
+def preprocess(img, label=None, resolution=[192, 192]):
+    imgmtx = np.copy(img)
+    lblsmtx = np.copy(label)
+
+    imgmtx[imgmtx < -1024] = -1024
+    imgmtx[imgmtx > 600] = 600
+    cip_xnew = []
+    cip_box = []
+    cip_mask = []
+    for i in range(imgmtx.shape[0]):
+        if label is None:
+            (im, m, box) = crop_and_resize(imgmtx[i, :, :], width=resolution[0], height=resolution[1])
+        else:
+            (im, m, box) = crop_and_resize(imgmtx[i, :, :], mask=lblsmtx[i, :, :], width=resolution[0],
+                                           height=resolution[1])
+            cip_mask.append(m)
+        cip_xnew.append(im)
+        cip_box.append(box)
+    if label is None:
+        return np.asarray(cip_xnew), cip_box
+    else:
+        return np.asarray(cip_xnew), cip_box, np.asarray(cip_mask)
+
+
+def simple_bodymask(img):
+    maskthreshold = -500
+    oshape = img.shape
+    img = ndimage.zoom(img, 128/np.asarray(img.shape), order=0)
+    bodymask = img > maskthreshold
+    bodymask = ndimage.binary_closing(bodymask)
+    bodymask = ndimage.binary_fill_holes(bodymask, structure=np.ones((3, 3))).astype(int)
+    bodymask = ndimage.binary_erosion(bodymask, iterations=2)
+    bodymask = skimage.measure.label(bodymask.astype(int), connectivity=1)
+    regions = skimage.measure.regionprops(bodymask.astype(int))
+    if len(regions) > 0:
+        max_region = np.argmax(list(map(lambda x: x.area, regions))) + 1
+        bodymask = bodymask == max_region
+        bodymask = ndimage.binary_dilation(bodymask, iterations=2)
+    real_scaling = np.asarray(oshape)/128
+    return ndimage.zoom(bodymask, real_scaling, order=0)
+
+
+def crop_and_resize(img, mask=None, width=192, height=192):
+    bmask = simple_bodymask(img)
+    # img[bmask==0] = -1024 # this line removes background outside of the lung.
+    # However, it has been shown problematic with narrow circular field of views that touch the lung.
+    # Possibly doing more harm than help
+    reg = skimage.measure.regionprops(skimage.measure.label(bmask))
+    if len(reg) > 0:
+        bbox = np.asarray(reg[0].bbox)
+    else:
+        bbox = (0, 0, bmask.shape[0], bmask.shape[1])
+    img = img[bbox[0]:bbox[2], bbox[1]:bbox[3]]
+    img = ndimage.zoom(img, np.asarray([width, height]) / np.asarray(img.shape), order=1)
+    if not mask is None:
+        mask = mask[bbox[0]:bbox[2], bbox[1]:bbox[3]]
+        mask = ndimage.zoom(mask, np.asarray([width, height]) / np.asarray(mask.shape), order=0)
+        # mask = ndimage.binary_closing(mask,iterations=5)
+    return img, mask, bbox
+
+
+## For some reasons skimage.transform leads to edgy mask borders compared to ndimage.zoom
+# def reshape_mask(mask, tbox, origsize):
+#     res = np.ones(origsize) * 0
+#     resize = [tbox[2] - tbox[0], tbox[3] - tbox[1]]
+#     imgres = skimage.transform.resize(mask, resize, order=0, mode='constant', cval=0, anti_aliasing=False, preserve_range=True)
+#     res[tbox[0]:tbox[2], tbox[1]:tbox[3]] = imgres
+#     return res
+
+
+def reshape_mask(mask, tbox, origsize):
+    res = np.ones(origsize) * 0
+    resize = [tbox[2] - tbox[0], tbox[3] - tbox[1]]
+    imgres = ndimage.zoom(mask, resize / np.asarray(mask.shape), order=0)
+    res[tbox[0]:tbox[2], tbox[1]:tbox[3]] = imgres
+    return res
+
+
+class LungLabelsDS_inf(Dataset):
+    def __init__(self, ds):
+        self.dataset = ds
+
+    def __len__(self):
+        return len(self.dataset)
+
+    def __getitem__(self, idx):
+        return self.dataset[idx, None, :, :].astype(np.float)
+
+
+def read_dicoms(path, primary=True, original=True):
+    allfnames = []
+    for dir, _, fnames in os.walk(path):
+        [allfnames.append(os.path.join(dir, fname)) for fname in fnames]
+
+    dcm_header_info = []
+    dcm_parameters = []
+    unique_set = []  # need this because too often there are duplicates of dicom files with different names
+    i = 0
+    for fname in tqdm(allfnames):
+        filename_ = os.path.splitext(os.path.split(fname)[1])
+        i += 1
+        if filename_[0] != 'DICOMDIR':
+            try:
+                dicom_header = pyd.dcmread(fname, defer_size=100, stop_before_pixels=True, force=True)
+                if dicom_header is not None:
+                    if 'ImageType' in dicom_header:
+                        if primary:
+                            is_primary = all([x in dicom_header.ImageType for x in ['PRIMARY']])
+                        else:
+                            is_primary = True
+
+                        if original:
+                            is_original = all([x in dicom_header.ImageType for x in ['ORIGINAL']])
+                        else:
+                            is_original = True
+
+                        # if 'ConvolutionKernel' in dicom_header:
+                        #     ck = dicom_header.ConvolutionKernel
+                        # else:
+                        #     ck = 'unknown'
+                        if is_primary and is_original and 'LOCALIZER' not in dicom_header.ImageType:
+                            h_info_wo_name = [dicom_header.StudyInstanceUID, dicom_header.SeriesInstanceUID,
+                                              dicom_header.ImagePositionPatient]
+                            h_info = [dicom_header.StudyInstanceUID, dicom_header.SeriesInstanceUID, fname,
+                                      dicom_header.ImagePositionPatient]
+                            if h_info_wo_name not in unique_set:
+                                unique_set.append(h_info_wo_name)
+                                dcm_header_info.append(h_info)
+                                # kvp = None
+                                # if 'KVP' in dicom_header:
+                                #     kvp = dicom_header.KVP
+                                # dcm_parameters.append([ck, kvp,dicom_header.SliceThickness])
+            except:
+                logging.error("Unexpected error:", sys.exc_info()[0])
+                logging.warning("Doesn't seem to be DICOM, will be skipped: ", fname)
+
+    conc = [x[1] for x in dcm_header_info]
+    sidx = np.argsort(conc)
+    conc = np.asarray(conc)[sidx]
+    dcm_header_info = np.asarray(dcm_header_info)[sidx]
+    # dcm_parameters = np.asarray(dcm_parameters)[sidx]
+    vol_unique = np.unique(conc, return_index=1, return_inverse=1)  # unique volumes
+    n_vol = len(vol_unique[1])
+    logging.info('There are ' + str(n_vol) + ' volumes in the study')
+
+    relevant_series = []
+    relevant_volumes = []
+
+    for i in range(len(vol_unique[1])):
+        curr_vol = i
+        info_idxs = np.where(vol_unique[2] == curr_vol)[0]
+        vol_files = dcm_header_info[info_idxs, 2]
+        positions = np.asarray([np.asarray(x[2]) for x in dcm_header_info[info_idxs, 3]])
+        slicesort_idx = np.argsort(positions)
+        vol_files = vol_files[slicesort_idx]
+        relevant_series.append(vol_files)
+        reader = sitk.ImageSeriesReader()
+        reader.SetFileNames(vol_files)
+        vol = reader.Execute()
+        relevant_volumes.append(vol)
+
+    return relevant_volumes
+
+
+def get_input_image(path):
+    if os.path.isfile(path):
+        logging.info(f'Read input: {path}')
+        input_image = sitk.ReadImage(path)
+    else:
+        logging.info(f'Looking for dicoms in {path}')
+        dicom_vols = read_dicoms(path, original=False, primary=False)
+        if len(dicom_vols) < 1:
+            sys.exit('No dicoms found!')
+        if len(dicom_vols) > 1:
+            logging.warning("There are more than one volume in the path, will take the largest one")
+        input_image = dicom_vols[np.argmax([np.prod(v.GetSize()) for v in dicom_vols], axis=0)]
+    return input_image
+
+
+def postrocessing(label_image, spare=[]):
+    '''some post-processing mapping small label patches to the neighbout whith which they share the
+        largest border. All connected components smaller than min_area will be removed
+    '''
+
+    # merge small components to neighbours
+    regionmask = skimage.measure.label(label_image)
+    origlabels = np.unique(label_image)
+    origlabels_maxsub = np.zeros((max(origlabels) + 1,), dtype=np.uint32)  # will hold the largest component for a label
+    regions = skimage.measure.regionprops(regionmask, label_image)
+    regions.sort(key=lambda x: x.area)
+    regionlabels = [x.label for x in regions]
+
+    # will hold mapping from regionlabels to original labels
+    region_to_lobemap = np.zeros((len(regionlabels) + 1,), dtype=np.uint8)
+    for r in regions:
+        if r.area > origlabels_maxsub[r.max_intensity]:
+            origlabels_maxsub[r.max_intensity] = r.area
+            region_to_lobemap[r.label] = r.max_intensity
+
+    for r in tqdm(regions):
+        if (r.area < origlabels_maxsub[r.max_intensity] or r.max_intensity in spare) and r.area>2: # area>2 improves runtime because small areas 1 and 2 voxel will be ignored
+            bb = bbox_3D(regionmask == r.label)
+            sub = regionmask[bb[0]:bb[1], bb[2]:bb[3], bb[4]:bb[5]]
+            dil = ndimage.binary_dilation(sub == r.label)
+            neighbours, counts = np.unique(sub[dil], return_counts=True)
+            mapto = r.label
+            maxmap = 0
+            myarea = 0
+            for ix, n in enumerate(neighbours):
+                if n != 0 and n != r.label and counts[ix] > maxmap and n != spare:
+                    maxmap = counts[ix]
+                    mapto = n
+                    myarea = r.area
+            regionmask[regionmask == r.label] = mapto
+            # print(str(region_to_lobemap[r.label]) + ' -> ' + str(region_to_lobemap[mapto])) # for debugging
+            if regions[regionlabels.index(mapto)].area == origlabels_maxsub[
+                regions[regionlabels.index(mapto)].max_intensity]:
+                origlabels_maxsub[regions[regionlabels.index(mapto)].max_intensity] += myarea
+            regions[regionlabels.index(mapto)].__dict__['_cache']['area'] += myarea
+
+    outmask_mapped = region_to_lobemap[regionmask]
+    outmask_mapped[outmask_mapped==spare] = 0 
+
+    if outmask_mapped.shape[0] == 1:
+        # holefiller = lambda x: ndimage.morphology.binary_fill_holes(x[0])[None, :, :] # This is bad for slices that show the liver
+        holefiller = lambda x: skimage.morphology.area_closing(x[0].astype(int), area_threshold=64)[None, :, :] == 1
+    else:
+        holefiller = fill_voids.fill
+
+    outmask = np.zeros(outmask_mapped.shape, dtype=np.uint8)
+    for i in np.unique(outmask_mapped)[1:]:
+        outmask[holefiller(keep_largest_connected_component(outmask_mapped == i))] = i
+
+    return outmask
+
+
+def bbox_3D(labelmap, margin=2):
+    shape = labelmap.shape
+    r = np.any(labelmap, axis=(1, 2))
+    c = np.any(labelmap, axis=(0, 2))
+    z = np.any(labelmap, axis=(0, 1))
+
+    rmin, rmax = np.where(r)[0][[0, -1]]
+    rmin -= margin if rmin >= margin else rmin
+    rmax += margin if rmax <= shape[0] - margin else rmax
+    cmin, cmax = np.where(c)[0][[0, -1]]
+    cmin -= margin if cmin >= margin else cmin
+    cmax += margin if cmax <= shape[1] - margin else cmax
+    zmin, zmax = np.where(z)[0][[0, -1]]
+    zmin -= margin if zmin >= margin else zmin
+    zmax += margin if zmax <= shape[2] - margin else zmax
+    
+    if rmax-rmin == 0:
+        rmax = rmin+1
+
+    return np.asarray([rmin, rmax, cmin, cmax, zmin, zmax])
+
+
+def keep_largest_connected_component(mask):
+    mask = skimage.measure.label(mask)
+    regions = skimage.measure.regionprops(mask)
+    resizes = np.asarray([x.area for x in regions])
+    max_region = np.argsort(resizes)[-1] + 1
+    mask = mask == max_region
+    return mask

data/requirements.txt

@@ -0,0 +1,9 @@
+pydicom==1.3.0
+numpy==1.17.2
+scikit_image==0.15.0
+torch==1.2.0
+torchvision==0.4.0a0+6b959ee
+scipy==1.3.1
+SimpleITK==1.2.4
+skimage==0.0
+fill_voids

data/setup.py

@@ -0,0 +1,37 @@
+import setuptools
+
+with open("README.md", "r", encoding='utf-8') as fh:
+    long_description = fh.read()
+
+setuptools.setup(
+    name="lungmask",
+    version="0.2.8",
+    author="Johannes Hofmanninger",
+    author_email="[email protected]",
+    description="Package for automated lung segmentation in CT",
+    long_description=long_description,
+    long_description_content_type="text/markdown",
+    url="https://github.com/JoHof/lungmask",
+    packages=setuptools.find_packages(),
+    entry_points={
+        'console_scripts': [
+            'lungmask = lungmask.__main__:main'
+        ]
+    },
+    install_requires=[
+        'pydicom',
+        'numpy',
+        'torch',
+        'scipy',
+        'SimpleITK',
+        'tqdm',
+        'scikit-image',
+        'fill_voids'
+    ],
+    classifiers=[
+        "Programming Language :: Python :: 3",
+        "License :: OSI Approved :: GPLv3",
+        "Operating System :: OS Independent"
+    ],
+    python_requires='>=3.6',
+)