# -*- coding: utf-8 -*- """ Plotting functions for visualizing experiment results. UPDATED: Publication-ready styling (Large fonts, Colorblind-friendly, High contrast). """ import matplotlib.pyplot as plt import seaborn as sns import numpy as np import pandas as pd import os import matplotlib.gridspec as gridspec import traceback import logging import matplotlib.patheffects as pe logger = logging.getLogger('DatnikExperiment') # Map variable names to readable labels VARIABLE_NAMES_MAP = { 'meanamplitude': 'Mean Amplitude', 'stdamplitude': 'SD of Amplitude', 'meanrmsvelocity': 'Mean RMS Velocity', 'stdrmsvelocity': 'SD of RMS Velocity', 'meanopeningspeed': 'Mean Opening Speed', 'stdopeningspeed': 'SD of Opening Speed', 'meanclosingspeed': 'Mean Closing Speed', 'stdclosingspeed': 'SD of Closing Speed', 'meancycleduration': 'Mean Cycle Duration', 'stdcycleduration': 'SD of Cycle Duration', 'rangecycleduration': 'Range of Cycle Duration', 'amplitudedecay': 'Amplitude Decay', 'velocitydecay': 'Velocity Decay', 'ratedecay': 'Rate Decay', 'cvamplitude': 'CV of Amplitude', 'cvcycleduration': 'CV of Cycle Duration', 'cvrmsvelocity': 'CV of RMS Velocity', 'cvopeningspeed': 'CV of Opening Speed', 'cvclosingspeed': 'CV of Closing Speed', 'rate': 'Frequency', 'meanspeed': 'Mean Speed', 'stdspeed': 'SD of Speed', 'cvspeed': 'CV of Speed' } def apply_pub_style(): """Sets consistent large fonts and styling for journal submission.""" sns.set_context("paper", font_scale=1.6) sns.set_style("whitegrid") plt.rcParams.update({ 'font.family': 'sans-serif', 'font.sans-serif': ['Arial', 'Helvetica', 'DejaVu Sans'], 'axes.labelsize': 14, 'axes.titlesize': 16, 'xtick.labelsize': 12, 'ytick.labelsize': 12, 'legend.fontsize': 12, 'lines.linewidth': 2.5, 'figure.titlesize': 20, 'figure.dpi': 300, 'axes.edgecolor': 'black', 'axes.linewidth': 1.2 }) def plot_figure2(summary_df, roc_data, importances_data, output_folder, config): """ Generates the publication-ready 2-panel Prediction Figure (ROC + Feature Sig). Labeled as Figure 2. """ try: apply_pub_style() CONFIG_FT = 'LR_RFE15_FT_OriginalFeats' CONFIG_HM = 'LR_RFE15_HM_OriginalFeats' COLOR_FT = '#332288' # Indigo COLOR_HM = '#117733' # Green fig = plt.figure(figsize=(18, 8)) gs = gridspec.GridSpec( 1, 2, figure=fig, left=0.06, right=0.98, bottom=0.08, top=0.90, wspace=0.35 ) ax_A = fig.add_subplot(gs[0, 0]) ax_B = fig.add_subplot(gs[0, 1]) # --- Panel A: ROC Curves --- ax_A.plot([0, 1], [0, 1], color='gray', linestyle='--', lw=2, label='Chance (AUC = 0.50)') for plot_cfg in [ {'name': CONFIG_FT, 'label': 'Finger Tapping', 'color': COLOR_FT}, {'name': CONFIG_HM, 'label': 'Hand Movement', 'color': COLOR_HM} ]: cfg_name = plot_cfg['name'] roc_runs = roc_data.get(cfg_name, []) metrics_row = summary_df[summary_df['Config_Name'] == cfg_name] if roc_runs and not metrics_row.empty: metrics_row = metrics_row.iloc[0] base_fpr = np.linspace(0, 1, 101) tprs_interp = [np.interp(base_fpr, fpr, tpr) for fpr, tpr in roc_runs if len(fpr) > 1] if tprs_interp: mean_tprs = np.mean(tprs_interp, axis=0) std_tprs = np.std(tprs_interp, axis=0) tprs_upper = np.minimum(mean_tprs + std_tprs, 1) tprs_lower = np.maximum(mean_tprs - std_tprs, 0) auc_val = metrics_row['Mean_ROC_AUC'] auc_std = metrics_row['Std_ROC_AUC'] auc_label = f"{plot_cfg['label']}\nAUC = {auc_val:.2f} ± {auc_std:.2f}" ax_A.plot(base_fpr, mean_tprs, label=auc_label, color=plot_cfg['color'], lw=3) ax_A.fill_between(base_fpr, tprs_lower, tprs_upper, color=plot_cfg['color'], alpha=0.15) ax_A.set_title('A) Prediction Performance', fontsize=16, weight='bold', loc='left') ax_A.set_xlabel('False Positive Rate', weight='bold') ax_A.set_ylabel('True Positive Rate', weight='bold') ax_A.legend(loc='lower right', fontsize=13) ax_A.set_aspect('equal', adjustable='box') # --- Panel B: Feature Importance (Finger Tapping signature) --- ft_coeffs_df = importances_data.get('group', {}).get(CONFIG_FT, {}).get('ft') if ft_coeffs_df is not None and not ft_coeffs_df.empty: plot_df = ( ft_coeffs_df .reindex(ft_coeffs_df['Mean_Importance'].abs().sort_values(ascending=False).index) .head(config.PLOT_TOP_N_FEATURES) .sort_values('Mean_Importance', ascending=False) ) base_feature_names = plot_df.index.to_series().str.replace('ft_', '', regex=False) plot_df = plot_df.copy() plot_df['Readable_Feature'] = base_feature_names.map(VARIABLE_NAMES_MAP).fillna(base_feature_names) sns.barplot( x='Mean_Importance', y='Readable_Feature', data=plot_df, ax=ax_B, edgecolor='black' ) colors = ['#D55E00' if c < 0 else '#0072B2' for c in plot_df['Mean_Importance']] for patch, c in zip(ax_B.patches, colors): patch.set_facecolor(c) ax_B.axvline(0, color='black', linewidth=1.5, linestyle='--') ax_B.set_title('B) Kinematic Signature (Finger Tapping)', fontsize=16, weight='bold', loc='left') ax_B.set_xlabel('Mean Coefficient (Log-Odds)', weight='bold') ax_B.set_ylabel('') ax_B.tick_params(axis='y', pad=6) # small readability boost else: ax_B.text(0.5, 0.5, "No coefficient data.", ha='center', va='center', transform=ax_B.transAxes) ax_B.set_title('B) Kinematic Signature (Finger Tapping)', fontsize=16, weight='bold', loc='left') ax_B.set_xlabel('Mean Coefficient (Log-Odds)', weight='bold') ax_B.set_ylabel('') fig.suptitle('Figure 2: Prediction of Dopaminergic Deficit', fontsize=20, weight='bold', y=0.98) figure_filename = os.path.join(output_folder, "Figure 2.pdf") plt.savefig(figure_filename, dpi=300, bbox_inches='tight') tiff_filename = figure_filename.rsplit('.', 1)[0] + '.tiff' try: plt.savefig(tiff_filename, dpi=600, bbox_inches='tight', pil_kwargs={"compression": "tiff_lzw"}) except TypeError: plt.savefig(tiff_filename, dpi=600, bbox_inches='tight') plt.close(fig) logger.info(f"--- SUCCESS! Figure 2 saved to: {figure_filename} ---") logger.info(f"--- SUCCESS! Figure 2 TIFF saved to: {tiff_filename} ---") except Exception as e: logger.error(f"Error generating Figure 2: {e}") traceback.print_exc() def plot_performance_vs_threshold(summary_df, x_col, y_col, filename, config): """ Generates sensitivity analysis plot. """ apply_pub_style() std_col = y_col.replace('Mean_', 'Std_') if not {x_col, y_col, std_col, 'Config_Name'}.issubset(summary_df.columns): return fig, ax = plt.subplots(figsize=(12, 7)) configs = sorted(summary_df['Config_Name'].unique()) color_map = { 'LR_RFE15_FT_OriginalFeats': '#332288', 'LR_RFE15_HM_OriginalFeats': '#117733', } for cfg_name in configs: df_cfg = summary_df[summary_df['Config_Name'] == cfg_name].copy() df_cfg = df_cfg.sort_values(by=x_col).dropna(subset=[x_col, y_col, std_col]) if df_cfg.empty: continue x = df_cfg[x_col] y_mean = df_cfg[y_col] y_std = df_cfg[std_col] color = color_map.get(cfg_name, 'gray') label = cfg_name.replace('_', ' ').replace('OriginalFeats', '').replace('LR RFE15', '').strip() ax.plot(x, y_mean, marker='o', linestyle='-', label=label, color=color, markersize=8, lw=3) ax.fill_between(x, y_mean - y_std, y_mean + y_std, color=color, alpha=0.15) ax.axhline(y=0.5, color='black', linestyle='--', label='Chance', lw=2) ax.set_title('Sensitivity Analysis: Performance vs. Threshold', fontsize=18, weight='bold', pad=15) ax.set_xlabel('DaTscan Z-Score Threshold (More Severe -->)', fontsize=14, weight='bold') ax.set_ylabel('Mean ROC AUC', fontsize=14, weight='bold') ax.legend(title='Model', loc='best', fontsize=12) ax.grid(True, which='both', linestyle=':', linewidth=1) ax.invert_xaxis() ax.set_ylim(0.40, 0.85) sns.despine() fig.tight_layout() try: pdf_filename = filename.replace('.png', '.pdf') tiff_filename = filename.replace('.png', '.tiff') plt.savefig(pdf_filename, dpi=300, bbox_inches='tight') try: plt.savefig(tiff_filename, dpi=600, bbox_inches='tight', pil_kwargs={"compression": "tiff_lzw"}) except TypeError: plt.savefig(tiff_filename, dpi=600, bbox_inches='tight') logger.info(f"Performance vs. Threshold plot saved to: {pdf_filename}") logger.info(f"Performance vs. Threshold TIFF saved to: {tiff_filename}") except Exception as e: logger.error(f"Error saving threshold plot: {e}") finally: plt.close(fig)