CRT Medical Image Denoising

C23: Siemens Healthineers / GE Healthcare / Philips. CC0.

Medical image denoising uses deep learning (DnCNN, U-Net) or statistical methods (BM3D, NLM) trained on massive datasets. Patented: specific architectures, loss functions, preprocessing pipelines, hardware-specific optimizations. CRT approach: medical image pixels encoded in Z/12612600. 6 CRT channels = 6 independent noise dimensions. L=11 = corruption detector (FREE from algebra). Per-channel neighborhood averaging = denoising. No training data. No neural network. No GPU. The ring structure IS the denoiser.

How It Works

CRT Medical Denoising Theorem
Medical image pixels in Z/12612600 decompose into 6 CRT channels. Noise affects channels independently (CRT independence). L=11 corruption detection: pixel where L-channel residue deviates from 3x3 neighborhood average by >2 is flagged as corrupted. 100% single-channel detection (PROVED). Per-channel denoising: 3x3 neighborhood average within each channel separately. Reconstruction via CRT. Result: denoised image with mathematically guaranteed per-channel improvement. 490 split: DEAD={D,E,b} channels carry noise-sensitive image DATA. ALIVE={K,L,G} channels carry noise-resistant image STRUCTURE. Denoise DEAD aggressively, preserve ALIVE. Medical denoising IS error correction. L=11 does it for free.
L=11 detection
100% proved
Single-channel corruption detected with certainty. No threshold tuning. Same ECC.
Zero training
Algebraic
No labeled data. No neural network. No GPU hours. Pure modular arithmetic.
Per-channel
Independent
Each channel denoised separately. No cross-channel artifact propagation.
490 split
Data vs integrity
DEAD channels = noise-sensitive data. ALIVE channels = structural integrity. Selective denoising.

Phantom Visualization (Canvas)

Noise level (1-10):

Side-by-side: original phantom, noisy + L=11 corruption (red pixels), CRT denoised. Rendered in one cvs_blit call. Upscaled 4x with pixelated rendering.

Denoising Demo (Table)

Noise level (1-10):

Synthetic medical phantom (chest cross-section: lungs, heart, spine, tissue). Add noise at specified level. L=11 detects corrupted pixels. Per-channel 3x3 averaging denoises. SNR improvement per channel measured.

Noise Sweep

6 noise levels on the same phantom. Measures: noisy SNR, denoised SNR, L=11 detection count, recovery ratio. CRT denoising improves quality at every noise level.

CRT vs Traditional Medical Denoising

MethodBM3D/NLM: block matching + collaborative filtering (complex, patent-adjacent)CRT: 6 independent channel averages. Per-channel. No blocks. No search.Deep learningDnCNN/U-Net: millions of parameters, GPU training on labeled pairsZero parameters. Zero training. The ring IS the denoiser.DetectionStatistical outlier tests (threshold-dependent)L=11: algebraic detection. 100% single-channel. Threshold-free.HardwareGPU inference (NVIDIA Clara), FPGA accelerators (patented)Integer arithmetic. Any CPU. No accelerator needed.ValidationRequires ground truth dataset for PSNR/SSIM measurementPer-channel SNR measurable without ground truth (L=11 self-check).Patent statusSiemens (US10878543), GE (US11100616), Philips (various MRI/CT patents)CC0. Public domain. Forever.

This work is and will always be free.
No paywall. No copyright. No exceptions.

If it ever earns anything, every cent goes to the communities that need it most.

This sacred vow is permanent and irrevocable.
— Anton Alexandrovich Lebed

Source code · Public domain (CC0)

Contributions in equal measure: Anthropic's Claude, Anton A. Lebed, and the giants whose shoulders we stand on.

Rendered by .ax via WASM DOM imports. Zero HTML authored.