Computer Vision News - February 2022

39 FMI and Deep Learning in Medical Imaging Dicom image info However sometimes it is better to get more targeted information from the metadata. get_image_info supplies image specific information such as the Modality, Photometric Interpretation, Pixel Spacing, SliceThickness, Bits Allocated, RescaleIntercept and RescaleSlope. This information is important when modelling the data . Note that in this dataset the dicom images do not have any Rescale Intercept or Rescale Slope get_image_info(p_items[0]) (0008, 0060) Modality CS: 'CR' (0028, 0004) Photometric Interpretation CS: 'MONOCHROME2' (0028, 0030) Pixel Spacing DS: [0.168, 0.168] No SliceThickness (0028, 0100) Bits Allocated US: 8 (0028, 0101) Bits Stored US: 8 (0028, 0103) Pixel Representation US: 0 No Pixel Padding Value No Rescale Intercept No Rescale Slope What about PII? PII or PersonallyIdentifiableInformation refers toinformationthatcanbeusedtodistinguish or trace an individual's identity, either alone or when combined with other personal or naming information that is linked or linkable to a specific individual. Understanding what identifiable information is contained within the metadata is important in creating a fair non-biased model . get_pii conveniently lists this information as well as looking for the Image Comments tag within the metadata that may hold useful information. It is also sometimes useful to know if there is any information about with de-identification information is available in the dicom meta-data. get_pii(p_items[0]) (0010, 0010) Patient's Name PN: '16d7f894-55d7-4d95-8957-d18987f0e981' (0010, 0030) Patient's Birth Date DA: '' (0010, 0040) Patient's Sex CS: 'M' (0010, 1010) Patient's Age AS: '62' No Image Comments No De-Identification method No De-identification Method Code Sequence DICOMS method from fastai fastai has a handy method from_dicoms that can convert dicom metadata into a dataframe. By default, from_dicoms returns more stats such as minimal pixel value, largest pixel value, mean pixel value, image standard deviation and pct_in_windowvalue depending on the specified window value. These stats are returned at the end of the dataframe.