%% Create an artificial set of QRS detections, based on multilead QRS detections created with wavedet algorithm.
% This is an auxiliar function for the ECGtask_QRS_detections_post_process.
% This function creates an artificial set of QRS detections, based on single
% lead detections, as described in [add reference].
% 
% 
%   all_detections = wavedet_QRS_detection_mix(struct_in, ECG_header, start_end_this_segment )
% 
% Arguments:
% 
%      + struct_in: the structure which results from loading the result of
%                   the QRS detection task used for invoking wavedet:
% 
%                   cached_filenames = ECG_w.GetCahchedFileName('QRS_detection');
%                   struct_in = load(cached_filenames{1});
% 
%      +ECG_header: [struct] OPTIONAL. 
% 
%             Description of the ECG typically available in the
%             ECG_header. Structure with fields:
% 
%               -freq: Sampling rate in Hz. (1)
% 
%               -nsig: Number of ECG leads. (size(ECG,2))
% 
%               -nsamp: Number of ECG samples. (size(ECG,1))
% 
%      + start_end_this_segment: an array with the first and last sample
%                                indexes.
% 
% Output:
% 
%      + all_detections: struct with the artificial detections named with
%                        prefix "wavedetMix_ECGmix". 
% 
% Example
% 
%     ECG_w.ECGtaskHandle = 'QRS_detections_post_process';
%     % Mix QRS detections strategy function
%     ECG_w.ECGtaskHandle.post_proc_func = 'wavedet_QRS_detection_mix';
%     ECG_w.ECGtaskHandle.payload = load(cached_filenames{1});
%     ECG_w.ECGtaskHandle.CalculatePerformance = true;
%     ECG_w.Run;
% 
% See also ECGtask_QRS_detections_post_process, calculate_artificial_QRS_detections
% 
% Author: Mariano Llamedo Soria llamedom@electron.frba.utn.edu.ar
% Version: 0.1 beta
% Birthdate: 01/01/2014
% Last update: 14/07/2015
% Copyright 2008-2015
% 
function all_detections = wavedet_QRS_detection_mix(struct_in, ECG_header, start_end_this_segment )

    all_detections = [];
    
    % attemp to build a better detection from single-lead detections.

    AnnNames = struct_in.series_quality.AnnNames(:,1);
    detector_name = cellfun( @(a)( strtok(a, '_')), AnnNames, 'UniformOutput', false);
    
    aux_idx = find(strcmpi(detector_name, 'wavedet'));
    
    all_annotations = {};
    for ii = rowvec(aux_idx)
        all_annotations = [all_annotations; {struct_in.(struct_in.series_quality.AnnNames{ii,1}).time}];
    end

    [ ratios, estimated_labs ] = CalcRRserieRatio(all_annotations, ECG_header, start_end_this_segment);

    [~, best_detections_idx] = sort(ratios, 'descend');

    % generate artificial annotations combining K best annotations
    aux_idx = best_detections_idx(1:min(10, length(best_detections_idx)) );
    artificial_annotations = combine_anns(all_annotations(aux_idx), estimated_labs(aux_idx), ECG_header );

    for ii = 1:length(artificial_annotations)
        aux_str = ['wavedetMix_ECGmix' num2str(ii)];
        all_detections.(aux_str) = artificial_annotations(ii);
    end