import numpy
from scipy import ndimage

from radiomics.firstorder import RadiomicsFirstOrder
from .mybase import MyRadiomicsFeaturesBase


class MyRadiomicsFirstOrder(MyRadiomicsFeaturesBase):
    def __init__(self, inputImage, inputMask, **kwargs):
        super().__init__(inputImage, inputMask, **kwargs)

        self.pixelSpacing = inputImage.GetSpacing()
        self.voxelArrayShift = kwargs.get('voxelArrayShift', 0)

    def _initCalculation(self):
        return super()._initSegmentBasedCalculation()

        self.targetVoxelArray = self.imageArray[self.labelledVoxelCoordinates].astype(
            'float'
        )
        self.discretizedTargetVoxelArray = None  # Lazy instantiation

        self.logger.debug('First order feature class initialized')

    def _moment(self, a, moment=1, axis=0):
        return RadiomicsFirstOrder._moment(self, a, moment, axis)

    def _getDiscretizedTargetVoxelArray(self):
        return RadiomicsFirstOrder._getDiscretizedTargetVoxelArray(self)

    def getSuvMaxFeatureValue(self):
        r"""
        **20. SUV max**

        Implemented as one of the intensity features extracted by Vallieres et al.
        Maximum SUV of the tumour region. 

        .. note::
        Extracted from PET scans and not used in the CT feature set.
        """
        ROIPet = self.imageArray
        mask = self.maskArray
        ROIPet[~mask] = numpy.nan

        return numpy.max(ROIPet[~numpy.isnan(ROIPet)])

    def getSuvMeanFeatureValue(self):
        r"""
        **21. SUV mean**

        Implemented as one of the intensity features extracted by valieres et al.
        Average SUV of the tumour region. 

        .. note::
        Extracted from PET scans and not used in the CT feature set.
        """
        ROIPet = self.imageArray
        mask = self.maskArray
        ROIPet[~mask] = numpy.nan

        return numpy.mean(ROIPet[~numpy.isnan(ROIPet)])

    def getTLGFeatureValue(self):
        r"""
        **22. TLG**

        Total lesion glycolysis. 
        Defined as 
        
        .. math::
        \textit{SUVMean}\times \textit{total volume of the tumour region}
    
        .. note::
        Extracted from PET scans and not used in the CT feature set.
        """
        z, y, x = self.pixelSpacing
        Np = len(self.labelledVoxelCoordinates[0])
        volume = Np * (z * x * y)

        ROIPet = self.imageArray
        mask = self.maskArray
        ROIPet[~mask] = numpy.nan

        return numpy.mean(ROIPet[~numpy.isnan(ROIPet)]) * volume

    def getAUCCSHFeatureValue(self):
        r"""
        **23. Inactive volume **

        Area under the curve of the cumulative SUV-volume histogram describing the percentage 
        of total tumour volume above a percentage threshold of maximum SUV, as defined by van Velden et al.

        .. note::
        Extracted from PET scans and not used in the CT feature set.
        """
        nBins = 1000  # By default
        ROIPet = self.imageArray
        mask = self.maskArray

        ROIPet[~mask] = numpy.nan

        outliers = numpy.where(
            ROIPet
            > (
                numpy.mean(ROIPet[~numpy.isnan(ROIPet)])
                + 3 * numpy.std(ROIPet[~numpy.isnan(ROIPet)])
            )
        )[0]

        good_voxels = numpy.where(
            ROIPet
            <= (
                numpy.mean(ROIPet[~numpy.isnan(ROIPet)])
                + 3 * numpy.std(ROIPet[~numpy.isnan(ROIPet)])
            )
        )[0]

        ROIPet[outliers] = numpy.mean(ROIPet[good_voxels])
        ROIPet = ROIPet - numpy.min(ROIPet[~numpy.isnan(ROIPet)])
        ROIPet = ROIPet / numpy.max(ROIPet[~numpy.isnan(ROIPet)])
        volume = ROIPet[~numpy.isnan(ROIPet)]
        nVoxel = len(volume)

        bins, _ = numpy.histogram(volume, nBins)

        csh = numpy.flipud(numpy.cumsum(numpy.flipud(bins)) / nVoxel)
        return numpy.sum(csh / nBins)

    def getgETUFeatureValue(self):
        r"""
        **24. gETU**

        Generalized effective total uptake, with parameter a = 0.25 as defined by Rahim et al.

        - input: 3D array representing the PET volume in SUV format, with 
                voxels outside the ROI set to NaNs.

        .. note::
        Extracted from PET scans and not used in the CT feature set.
        """
        ROIPet = self.imageArray
        mask = self.maskArray
        ROIPet[~mask] = numpy.nan

        a = 0.25
        n_voxels = numpy.sum(ROIPet[~numpy.isnan(ROIPet)])
        ROIPet = ROIPet ** a

        return numpy.sum(ROIPet[~numpy.isnan(ROIPet)] / n_voxels) ** (1 / a)