Biomedical Imaging Basics

Imaging Basics

• Biomedical imaging techniques extend beyond the visible light spectrum, enabling the observation of internal structures and functions unseen by the human eye.
• The electromagnetic spectrum encompasses waves arranged by wavelength or frequency.
• Ionizing radiation has enough energy to remove electrons from atoms, while non-ionizing radiation does not.
• Biomedical imaging is vital for diagnosing diseases, assessing treatment responses, and reducing unnecessary procedures.

Medical Imaging

• Medical images are represented as a matrix of numbers (pixels), often in grayscale, where black indicates the lowest and white the highest light intensity.
• Each pixel value provides information about the tissue or structure at a specific location.

Image Quality

• Important factors in good image quality include no acquisition issues, sharp resolution, absence of artifacts (e.g., rings on fingers in x-rays), good signal and low noise, as well as good contrast.
• CNR (contrast-to-noise ratio) and SNR (signal-to-noise ratio) are critical measures of image quality. Lower values for both indicate poor image quality.

Data Acquisition

• Analog detectors convert physical signals (e.g., light, sound, radiation) into continuous electrical signals, processed and converted to readable outputs. Modern systems favor digital detectors due to accuracy, integration, and the ability to create real-time images.
• Quantization errors arise when converting analog signals to digital signals because of imperfect matches of sampled discrete levels to continuous readings. Higher bit depths equate to smaller quantization errors with more possible discrete values for representing signal intensity.

Time Limitations

• High-resolution images require more time to capture data, which often limits their use in emergency situations or for patients who cannot remain still. Patient movement during data acquisition can introduce motion artifacts into the resulting images.
• The Nyquist theorem states that to accurately represent a signal, the sampling rate must be at least twice the highest frequency component of the signal.

Spatial Resolution

• Spatial resolution, the ability to distinguish between objects close together, depends on several factors including detector type, sampling rate, and bit depth.
• Specific measures used to evaluate spatial resolution of imaging modalities are line spread function (LSF), point spread function (PSF), and modulation transfer function (MTF), which represent blur, spreading, and how well the system captures details in an image.

Noise

• Noise in medical imaging is electronic, quantum, or environmental signals that are not part of the intended image.
• The signal-to-noise ratio (SNR) is the ratio of signal strength to noise strength, representing the overall image clarity. Low SNR indicates a poorly formed image.
• The contrast-to-noise ratio (CNR) assesses the ability of the technique to distinguish between tissues or structures with different signal intensities (contrast) by accounting for the presence of noise.

ROC Curves

• ROC (Receiver Operating Characteristic) curves visually represent the performance of a diagnostic test, imaging system, or classifier by examining the tradeoff between sensitivity (true positive rate) and specificity (false positive rate).

Confusion Matrices

• Confusion matrices provide raw data (true positives, false positives, true negatives, false negatives) for evaluating model performance in image classification and can be used to compute accuracy, sensitivity, specificity, etc to determine model performance.

Image Rendering

• Image rendering creates visual representations of medical data, based on the unique advantages and limitations of different imaging modalities.
• Common rendering techniques include gray-scale, superimposing information from different modalities, and highlighting surface information for better visualization and analysis.

Image Characteristics

• Histograms are graphical representations of pixel intensity distribution in an image, allowing for assessment of the intensity range for various tissues or areas of interest.

Fourier Transform

• Fourier transformation breaks down images into their frequency components. It's used for advanced image processing such as filtering and analyzing image patterns.
• The inverse Fourier Transform changes frequency domain images back into their spatially represented equivalent.

DICOM

• DICOM (Digital Imaging and Communications in Medicine) is a standard format for storing, transmitting, and sharing medical images and associated data. It's a critical element for enabling data exchange between various medical imaging devices and software systems. Its hierarchical structure of tags allows organization for effective data transfer.

Image Processing

• Image processing encompasses techniques for enhancing, manipulating (e.g., sharpening, smoothing), restoring (e.g., removing noise, blurring), and performing analysis on medical images. These techniques can often improve image quality and allow better visibility of specific features or structures.
• Histogram equalization is a technique for enhancing contrast by changing the pixel intensity distribution across the whole range of available values in the image.

Filtering

• Filters in image processing modify an image by changing the contribution of pixels within either the spatial domain or through manipulation of frequency components.
• Low-pass filters smooth out rapid intensity changes by averaging pixel values, while high-pass filters highlight rapid intensity changes between pixels, helping reveal edges and borders.

Segmentation & Registration

• Segmentation divides an image into meaningful parts (eg: tissues).
• Registration aligns images obtained at different times or from different imaging modalities.