adding original files

changelog.txt

@@ -0,0 +1,5 @@
+[07/20/2018] First release of DeepLesion.
+[07/27/2018] We re-zipped the images and replaced the old zips in Images_png. The files in old and new zips are the same in total except that every new zip file contains several independent folders and can be unzipped alone. Readme and FAQ are updated. We also added a folder "Key_slice_examples" to help users understand the dataset at a glance.
+[07/30/2018] The link to our GitHub page is added in readme, which includes a project showing how to train a lesion detector using DeepLesion.
+[09/05/2018] batch_download_zips.py is added to download all 56 zip files automatically.
+[05/13/2019] Address of the semantic labels (tags) for the lesions is addeed in readme and FAQ.

scripts/DL_save_nifti.py

@@ -0,0 +1,122 @@
+#!/usr/bin/env python
+"""
+Ke Yan
+Imaging Biomarkers and Computer-Aided Diagnosis Laboratory
+National Institutes of Health Clinical Center
+May 2018
+THIS SOFTWARE IS PROVIDED BY THE AUTHOR(S) ``AS IS'' AND ANY EXPRESS OR
+IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+IN NO EVENT SHALL THE AUTHOR(S) BE LIABLE FOR ANY DIRECT, INDIRECT,
+INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+"""
+
+"""
+A simple demo to load 2D 16-bit slices from DeepLesion and save to 3D nifti volumes.
+The nifti volumes can be viewed in software such as 3D slicer and ITK-SNAP.
+"""
+
+
+import numpy as np
+import nibabel as nib
+import os
+import cv2
+import csv
+
+
+dir_in = 'Images_png'
+dir_out = 'Images_nifti'
+out_fmt = '%s_%03d-%03d.nii.gz'  # format of the nifti file name to output
+info_fn = 'DL_info.csv'  # file name of the information file
+
+
+def slices2nifti(ims, fn_out, spacing):
+    """save 2D slices to 3D nifti file considering the spacing"""
+    if len(ims) < 300:  # cv2.merge does not support too many channels
+        V = cv2.merge(ims)
+    else:
+        V = np.empty((ims[0].shape[0], ims[0].shape[1], len(ims)))
+        for i in range(len(ims)):
+            V[:, :, i] = ims[i]
+
+    # the transformation matrix suitable for 3D slicer and ITK-SNAP
+    T = np.array([[0, -spacing[1], 0, 0], [-spacing[0], 0, 0, 0], [0, 0, -spacing[2], 0], [0, 0, 0, 1]])
+    img = nib.Nifti1Image(V, T)
+    path_out = os.path.join(dir_out, fn_out)
+    nib.save(img, path_out)
+    print fn_out, 'saved'
+
+
+def load_slices(dir, slice_idxs):
+    """load slices from 16-bit png files"""
+    slice_idxs = np.array(slice_idxs)
+    assert np.all(slice_idxs[1:] - slice_idxs[:-1] == 1)
+    ims = []
+    for slice_idx in slice_idxs:
+        fn = '%03d.png' % slice_idx
+        path = os.path.join(dir_in, dir, fn)
+        im = cv2.imread(path, -1)  # -1 is needed for 16-bit image
+        assert im is not None, 'error reading %s' % path
+        print 'read', path
+
+        # the 16-bit png file has a intensity bias of 32768
+        ims.append((im.astype(np.int32) - 32768).astype(np.int16))
+    return ims
+
+
+def read_DL_info():
+    """read spacings and image indices in DeepLesion"""
+    spacings = []
+    idxs = []
+    with open(info_fn, 'rb') as csvfile:
+        reader = csv.reader(csvfile)
+        rownum = 0
+        for row in reader:
+            if rownum == 0:
+                header = row
+                rownum += 1
+            else:
+                idxs.append([int(d) for d in row[1:4]])
+                spacings.append([float(d) for d in row[12].split(',')])
+
+    idxs = np.array(idxs)
+    spacings = np.array(spacings)
+    return idxs, spacings
+
+
+if __name__ == '__main__':
+    idxs, spacings = read_DL_info()
+    if not os.path.exists(dir_out):
+        os.mkdir(dir_out)
+    img_dirs = os.listdir(dir_in)
+    img_dirs.sort()
+    for dir1 in img_dirs:
+        # find the image info according to the folder's name
+        idxs1 = np.array([int(d) for d in dir1.split('_')])
+        i1 = np.where(np.all(idxs == idxs1, axis=1))[0]
+        spacings1 = spacings[i1[0]]
+
+        fns = os.listdir(os.path.join(dir_in, dir1))
+        slices = [int(d[:-4]) for d in fns if d.endswith('.png')]
+        slices.sort()
+
+        # Each folder contains png slices from one series (volume)
+        # There may be several sub-volumes in each volume depending on the key slices
+        # We group the slices into sub-volumes according to continuity of the slice indices
+        groups = []
+        for slice_idx in slices:
+            if len(groups) != 0 and slice_idx == groups[-1][-1]+1:
+                groups[-1].append(slice_idx)
+            else:
+                groups.append([slice_idx])
+
+        for group in groups:
+            # group contains slices indices of a sub-volume
+            ims = load_slices(dir1, group)
+            fn_out = out_fmt % (dir1, group[0], group[-1])
+            slices2nifti(ims, fn_out, spacings1)

scripts/batch_download_zips.py

@@ -0,0 +1,75 @@
+# Download the 56 zip files in Images_png in batches
+import urllib
+
+# URLs for the zip files
+links = [
+    'https://nihcc.box.com/shared/static/sp5y2k799v4x1x77f7w1aqp26uyfq7qz.zip',
+    'https://nihcc.box.com/shared/static/l9e1ys5e48qq8s409ua3uv6uwuko0y5c.zip',
+    'https://nihcc.box.com/shared/static/48jotosvbrw0rlke4u88tzadmabcp72r.zip',
+    'https://nihcc.box.com/shared/static/xa3rjr6nzej6yfgzj9z6hf97ljpq1wkm.zip',
+    'https://nihcc.box.com/shared/static/58ix4lxaadjxvjzq4am5ehpzhdvzl7os.zip',
+    'https://nihcc.box.com/shared/static/cfouy1al16n0linxqt504n3macomhdj8.zip',
+    'https://nihcc.box.com/shared/static/z84jjstqfrhhlr7jikwsvcdutl7jnk78.zip',
+    'https://nihcc.box.com/shared/static/6viu9bqirhjjz34xhd1nttcqurez8654.zip',
+    'https://nihcc.box.com/shared/static/9ii2xb6z7869khz9xxrwcx1393a05610.zip',
+    'https://nihcc.box.com/shared/static/2c7y53eees3a3vdls5preayjaf0mc3bn.zip',
+
+    'https://nihcc.box.com/shared/static/2zsqpzru46wsp0f99eaag5yiad42iezz.zip',
+    'https://nihcc.box.com/shared/static/8v8kfhgyngceiu6cr4sq1o8yftu8162m.zip',
+    'https://nihcc.box.com/shared/static/jl8ic5cq84e1ijy6z8h52mhnzfqj36q6.zip',
+    'https://nihcc.box.com/shared/static/un990ghdh14hp0k7zm8m4qkqrbc0qfu5.zip',
+    'https://nihcc.box.com/shared/static/kxvbvri827o1ssl7l4ji1fngfe0pbt4p.zip',
+    'https://nihcc.box.com/shared/static/h1jhw1bee3c08pgk537j02q6ue2brxmb.zip',
+    'https://nihcc.box.com/shared/static/78hamrdfzjzevrxqfr95h1jqzdqndi19.zip',
+    'https://nihcc.box.com/shared/static/kca6qlkgejyxtsgjgvyoku3z745wbgkc.zip',
+    'https://nihcc.box.com/shared/static/e8yrtq31g0d8yhjrl6kjplffbsxoc5aw.zip',
+    'https://nihcc.box.com/shared/static/vomu8feie1qembrsfy2yaq36cimvymj8.zip',
+
+    'https://nihcc.box.com/shared/static/ecwyyx47p2jd621wt5c5tc92dselz9nx.zip',
+    'https://nihcc.box.com/shared/static/fbnafa8rj00y0b5tq05wld0vbgvxnbpe.zip',
+    'https://nihcc.box.com/shared/static/50v75duviqrhaj1h7a1v3gm6iv9d58en.zip',
+    'https://nihcc.box.com/shared/static/oylbi4bmcnr2o65id2v9rfnqp16l3hp0.zip',
+    'https://nihcc.box.com/shared/static/mw15sn09vriv3f1lrlnh3plz7pxt4hoo.zip',
+    'https://nihcc.box.com/shared/static/zi68hd5o6dajgimnw5fiu7sh63kah5sd.zip',
+    'https://nihcc.box.com/shared/static/3yiszde3vlklv4xoj1m7k0syqo3yy5ec.zip',
+    'https://nihcc.box.com/shared/static/w2v86eshepbix9u3813m70d8zqe735xq.zip',
+    'https://nihcc.box.com/shared/static/0cf5w11yvecfq34sd09qol5atzk1a4ql.zip',
+    'https://nihcc.box.com/shared/static/275en88yybbvzf7hhsbl6d7kghfxfshi.zip',
+
+    'https://nihcc.box.com/shared/static/l52tpmmkgjlfa065ow8czhivhu5vx27n.zip',
+    'https://nihcc.box.com/shared/static/p89awvi7nj0yov1l2o9hzi5l3q183lqe.zip',
+    'https://nihcc.box.com/shared/static/or9m7tqbrayvtuppsm4epwsl9rog94o8.zip',
+    'https://nihcc.box.com/shared/static/vuac680472w3r7i859b0ng7fcxf71wev.zip',
+    'https://nihcc.box.com/shared/static/pllix2czjvoykgbd8syzq9gq5wkofps6.zip',
+    'https://nihcc.box.com/shared/static/2dn2kipkkya5zuusll4jlyil3cqzboyk.zip',
+    'https://nihcc.box.com/shared/static/peva7rpx9lww6zgpd0n8olpo3b2n05ft.zip',
+    'https://nihcc.box.com/shared/static/2fda8akx3r3mhkts4v6mg3si7dipr7rg.zip',
+    'https://nihcc.box.com/shared/static/ijd3kwljgpgynfwj0vhj5j5aurzjpwxp.zip',
+    'https://nihcc.box.com/shared/static/nc6rwjixplkc5cx983mng9mwe99j8oa2.zip',
+
+    'https://nihcc.box.com/shared/static/rhnfkwctdcb6y92gn7u98pept6qjfaud.zip',
+    'https://nihcc.box.com/shared/static/7315e79xqm72osa4869oqkb2o0wayz6k.zip',
+    'https://nihcc.box.com/shared/static/4nbwf4j9ejhm2ozv8mz3x9jcji6knhhk.zip',
+    'https://nihcc.box.com/shared/static/1lhhx2uc7w14bt70de0bzcja199k62vn.zip',
+    'https://nihcc.box.com/shared/static/guho09wmfnlpmg64npz78m4jg5oxqnbo.zip',
+    'https://nihcc.box.com/shared/static/epu016ga5dh01s9ynlbioyjbi2dua02x.zip',
+    'https://nihcc.box.com/shared/static/b4ebv95vpr55jqghf6bthg92vktocdkg.zip',
+    'https://nihcc.box.com/shared/static/byl9pk2y727wpvk0pju4ls4oomz9du6t.zip',
+    'https://nihcc.box.com/shared/static/kisfbpualo24dhby243nuyfr8bszkqg1.zip',
+    'https://nihcc.box.com/shared/static/rs1s5ouk4l3icu1n6vyf63r2uhmnv6wz.zip',
+
+    'https://nihcc.box.com/shared/static/7tvrneuqt4eq4q1d7lj0fnafn15hu9oj.zip',
+    'https://nihcc.box.com/shared/static/gjo530t0dgeci3hizcfdvubr2n3mzmtu.zip',
+    'https://nihcc.box.com/shared/static/7x4pvrdu0lhazj83sdee7nr0zj0s1t0v.zip',
+    'https://nihcc.box.com/shared/static/z7s2zzdtxe696rlo16cqf5pxahpl8dup.zip',
+    'https://nihcc.box.com/shared/static/shr998yp51gf2y5jj7jqxz2ht8lcbril.zip',
+    'https://nihcc.box.com/shared/static/kqg4peb9j53ljhrxe3l3zrj4ac6xogif.zip'
+]
+
+md5_link = 'https://nihcc.box.com/shared/static/q0f8gy79q2spw96hs6o4jjjfsrg17t55.txt'
+urllib.urlretrieve(md5_link, "MD5_checksums.txt")  # download the MD5 checksum file
+for idx, link in enumerate(links):
+    fn = 'Images_png_%02d.zip' % (idx+1)
+    print 'downloading', fn, '...'
+    urllib.urlretrieve(link, fn)  # download the zip file
+print "Download complete. Please check the MD5 checksums"