AI News, BOOK REVIEW: Machine Learning and Medical Imaging

Machine Learning and Medical Imaging

It first summarizes cutting-edge machine learning algorithms in medical imaging, including not only classical probabilistic modeling and learning methods, but also recent breakthroughs in deep learning, sparse representation/coding, and big data hashing.

In the second part leading research groups around the world present a wide spectrum of machine learning methods with application to different medical imaging modalities, clinical domains, and organs.

Machine Learning and Medical Imaging

It first summarizes cutting-edge machine learning algorithms in medical imaging, including not only classical probabilistic modeling and learning methods, but also recent breakthroughs in deep learning, sparse representation/coding, and big data hashing.

In the second part leading research groups around the world present a wide spectrum of machine learning methods with application to different medical imaging modalities, clinical domains, and organs.

Medical imaging

Medical imaging is the technique and process of creating visual representations of the interior of a body for clinical analysis and medical intervention, as well as visual representation of the function of some organs or tissues (physiology).

As a discipline and in its widest sense, it is part of biological imaging and incorporates radiology which uses the imaging technologies of X-ray radiography, magnetic resonance imaging, medical ultrasonography or ultrasound, endoscopy, elastography, tactile imaging, thermography, medical photography and nuclear medicine functional imaging techniques as positron emission tomography (PET) and Single-photon emission computed tomography (SPECT).

Up until 2010, 5 billion medical imaging studies had been conducted worldwide.[1] Radiation exposure from medical imaging in 2006 made up about 50% of total ionizing radiation exposure in the United States.[2] Medical imaging is often perceived to designate the set of techniques that noninvasively produce images of the internal aspect of the body.

As a field of scientific investigation, medical imaging constitutes a sub-discipline of biomedical engineering, medical physics or medicine depending on the context: Research and development in the area of instrumentation, image acquisition (e.g., radiography), modeling and quantification are usually the preserve of biomedical engineering, medical physics, and computer science;

magnetic resonance imaging instrument (MRI scanner), or 'nuclear magnetic resonance (NMR) imaging' scanner as it was originally known, uses powerful magnets to polarize and excite hydrogen nuclei (i.e., single protons) of water molecules in human tissue, producing a detectable signal which is spatially encoded, resulting in images of the body.[4] The MRI machine emits a radio frequency (RF) pulse at the resonant frequency of the hydrogen atoms on water molecules.

For example, imaging of prostate tumors is better accomplished using T2-MRI and DWI-MRI than T2-weighted imaging alone.[5] The number of applications of mpMRI for detecting disease in various organs continues to expand, including liver studies, breast tumors, pancreatic tumors, and assessing the effects of vascular disruption agents on cancer tumors.[6][7][8] Nuclear medicine encompasses both diagnostic imaging and treatment of disease, and may also be referred to as molecular medicine or molecular imaging &

By this method, functional information from SPECT or positron emission tomography can be related to anatomical information provided by magnetic resonance imaging (MRI).[12] Similarly, fiducial points established during MRI can be correlated with brain images generated by magnetoencephalography to localize the source of brain activity.

A path of reflected sound waves in a multilayered structure can be defined by an input acoustic impedance (ultrasound sound wave) and the Reflection and transmission coefficients of the relative structures.[11] It is very safe to use and does not appear to cause any adverse effects.

Main branches of ultrasound elastography include Quasistatic Elastography/Strain Imaging, Shear Wave Elasticity Imaging (SWEI), Acoustic Radiation Force Impulse imaging (ARFI), Supersonic Shear Imaging (SSI), and Transient Elastography.[14] In the last decade a steady increase of activities in the field of elastography is observed demonstrating successful application of the technology in various areas of medical diagnostics and treatment monitoring.

NIRS (near infrared spectroscopy) is used for the purpose of functional neuroimaging and has been widely accepted as a brain imaging technique.[17] Using superparamagnetic iron oxide nanoparticles, magnetic particle imaging (MPI) is a developing diagnostic imaging technique used for tracking superparamagnetic iron oxide nanoparticles.

Magnetic resonance imaging (MRI) without MRI contrast agents as well as obstetric ultrasonography are not associated with any risk for the mother or the fetus, and are the imaging techniques of choice for pregnant women.[18] Projectional radiography, X-ray computed tomography and nuclear medicine imaging result some degree of ionizing radiation exposure, but have with a few exceptions much lower absorbed doses than what are associated with fetal harm.[18] At higher dosages, effects can include miscarriage, birth defects and intellectual disability.[18] The amount of data obtained in a single MR or CT scan is very extensive.

Some of the data that radiologists discard could save patients time and money, while reducing their exposure to radiation and risk of complications from invasive procedures.[19] Another approach for making the procedures more efficient is based on utilizing additional constraints, e.g., in some medical imaging modalities one can improve the efficiency of the data acquisition by taking into account the fact the reconstructed density is positive.[20] Volume rendering techniques have been developed to enable CT, MRI and ultrasound scanning software to produce 3D images for the physician.[21] Traditionally CT and MRI scans produced 2D static output on film.

Many medical imaging software applications (3DSlicer, ImageJ, MIPAV, ImageVis3D, etc.) are used for non-diagnostic imaging, specifically because they don't have an FDA approval[22] and not allowed to use in clinical research for patient diagnosis.[23] Note that many clinical research studies are not designed for patient diagnosis anyway.[24] Used primarily in ultrasound imaging, capturing the image produced by a medical imaging device is required for archiving and telemedicine applications.

The DICOM Standard incorporates protocols for imaging techniques such as radiography, computed tomography (CT), magnetic resonance imaging (MRI), ultrasonography, and radiation therapy.[26] DICOM includes standards for image exchange (e.g., via portable media such as DVDs), image compression, 3-D visualization, image presentation, and results reporting.[27] Medical imaging techniques produce very large amounts of data, especially from CT, MRI and PET modalities.

Imaging biomarkers (a characteristic that is objectively measured by an imaging technique, which is used as an indicator of pharmacological response to a therapy) and surrogate endpoints have shown to facilitate the use of small group sizes, obtaining quick results with good statistical power.[29] Imaging is able to reveal subtle change that is indicative of the progression of therapy that may be missed out by more subjective, traditional approaches.

Imaging techniques such as positron emission tomography (PET) and magnetic resonance imaging (MRI) are routinely used in oncology and neuroscience areas,.[30][31][32][33] For example, measurement of tumour shrinkage is a commonly used surrogate endpoint in solid tumour response evaluation.

In Alzheimer's disease, MRI scans of the entire brain can accurately assess the rate of hippocampal atrophy, while PET scans can measure the brain's metabolic activity by measuring regional glucose metabolism,[29] and beta-amyloid plaques using tracers such as Pittsburgh compound B (PiB).

For example, in the United States the Health Insurance Portability and Accountability Act (HIPAA) sets restrictions for health care providers on utilizing protected health information, which is any individually identifiable information relating to the past, present, or future physical or mental health of any individual.[35] While there has not been any definitive legal decision in the matter, at least one study has indicated that medical imaging may contain biometric information that can uniquely identify a person, and so may qualify as PHI.[36] The UK General Medical Council's ethical guidelines indicate that the Council does not require consent prior to secondary uses of X-ray images.[37] As per Compendium: Chapter 300 by the US Copyright Office, 'the Office will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author.'

In Germany, X-ray images as well as MRT, ultrasound, PET and scintigraphy images are protected by (copyright-like) related rights or neighbouring rights.[39] This protection does not require creativity (as would be necessary for regular copyright protection) and lasts only for 50 years after image creation, if not published within 50 years, or for 50 years after the first legitimate publication.[40] The letter of the law grants this right to the 'Lichtbildner'[9], i.e.

Medical images created in the United Kingdom will normally be protected by copyright due to 'the high level of skill, labour and judgement required to produce a good quality x-ray, particularly to show contrast between bones and various soft tissues'.[41] The Society of Radiographers believe this copyright is owned by employer (unless the radiographer is self-employed—though even then their contract might require them to transfer ownership to the hospital).

Multimodal Correlative Preclinical Whole Body Imaging and Segmentation

The algorithm proposed is composed of several key steps: First, after performing the multi modal imaging with the designed bed and markers we register all imaging modalities to one pre-selected channel (in our experiments the MR 9.4T T2w).

In the third step a set of bounding boxes are determined automatically for each structure region by defining a coordinate systems on the mouse body and generating an average heat map per structure based on a the training set coordinates.

Mice were anesthetized and placed on a custom-made cross-modality bed loaded with fluorescent markers, and imaged sequentially with high field MRI (9.4T Bruker), low field MRI (1T Aspect), optical imaging (IVIS, Caliper) for bioluminescence (BLI), and micro-CT (Tomoscope).

Bed size and design was matched to the available insert in the IVIS spectrum stag, holes were created at the bottom of the bed at the exact pattern found in the IVIS stage in order to enable the use of trans-illumination.

Finally, mice were secured to a portal bed using small rubber bands on limbs and teeth holder and imaged sequentially by different modalities while positioned on bed with markers detectable by CT, IVIS and MRI to ensure accurate alignment between modalities.

For BLI imaging mice were given an intra-peritoneal injection of 1.5 mg of D-luciferin (Caliper Life Sciences), sequential imaging iterations of 1 min exposure were performed until signal reached maximal plateau (about 15 min post D-luciferin IP injection).

A single mouse was imaged at a time, Signal was acquired in BLI mode for 60 sec’ (no excitation and open emission filter) field of view of 12.6 cm’, Field stop and binning were selected according to signal in order to enable maximal signal without saturation.

Due to the maximum length limit, to cover the whole mouse body, imaging was performed in two parts with overlapping area and then all slices merged to one dataset representing the entire ROI.

The advantage of the selected mouse holder is the ability to transport a small size anaesthetized mouse on bed from one scanner to the other, allowing us to make simple rigid body assumptions for inter-modality images.

The manual ground truth segmentation included nine structures: heart, lungs, liver, stomach, left kidney, right kidney, tumor, vena cava and bladder (Fig.

Manual segmentation for four additional classes was generated in the CT data by manually assigning bright supervoxels (above a predefined threshold) to one of the following categories (upper limbs, ribs, spine and lower limbs).

In this section, we first briefly explain the SWA approach for generating supervoxels together with our extensions40 (we refer the reader to the appendix for a detailed formulation) and then we explain the machine learning formulation.

The algorithm recursively coarsens the graph, level after level, by softly aggregating several similar nodes of a finer level into a single node of the next coarser level.

The intermediate levels allow supervoxels to gather enough statistics before they merge with other structures and are determined based on the volume characteristics of the anatomical structures (scales 4–6 of ~14).

Two randomly chosen data sets were used to determine the weight coefficient [0.14, 0.14, 0.13, 0.13, 0.23, 0.25] and these sets were not used later in either the training or testing experiments.

First, the training sets are aligned with the test set and then the prior map is created by voxel-wise averaging of the tissue structures over the manually labeled training data sets.

The heatmap BB constraint yields a fast coarse categorization followed by the kNN model obtaining a smaller focused NN set of training examples and completed by the SVM executing a fine discrimination of the test supervoxels.

Assuming a training set of q supervoxels, k = 1, …q extracted from intermediate levels of the pyramid, each represented by a d dimensional feature vector, the features are first normalized to zero mean and unit variance.

SVMs attempt to find a separating hyperplane which maximizes the margins between the classes while minimizing the error on the training set with a cost depending on the number of misclassifications (C = 28 is the penalty parameter for misclassification on the training data).

Summary of Machine Learning Steps Optical imaging modalities such as Bioluminescence imaging (BLI) are widely used in-vivo to monitor biochemistry with high sensitivity specifically to follow tumors, albeit BLI does not provide anatomical information and therefore it is commonly fused with high resolution micro-CT images.

Yelawolf - American You

Pre-order the album Love Story now On iTunes: Google Play: Amazon MP3: .

"Pathophysiology and Diagnosis of Heart Failure" by Christina Vanderpluym, MD, for OPENPediatrics

Please visit: OPENPediatrics™ is an interactive digital learning platform for healthcare clinicians sponsored by Boston Children's ..

Demystifying Medicine 2017: Hepatocellular Cancer and Liver Transplantation

Demystifying Medicine 2017: Hepatocellular Cancer and Liver Transplantation Air date: Tuesday, March 7, 2017, 4:00:00 PM Category: Demystifying Medicine ...

TRACO 2016: Introduction and Tumor imaging

TRACO 2016: Introduction and Tumor imaging Air date: Tuesday, September 6, 2016, 4:00:00 PM Category: TRACO Runtime: 01:54:34 Description: Introduction ...

Cutting to cure cancer and 'the limits set by nature'

Professor Tom Treasure (UCL Clinical Operational Research Unit) Surgeons look back and see the expectations of their predecessors exceeded, and obstacles ...

Health Matters- Breathe Easy: Treating Lung Diseases

The lungs are among the hardest-working organs in the body. During a normal day, you breathe nearly 25000 times. If all types of lung disease are lumped ...

The Promises and Challenges of Digital Imaging as Applied to Pathology: Dr. Ulysses G.J. Balis

Dr. Ulysses G. J. Balis from the University of Michigan Medical School presents: Ephemeral Images, Indelible Data: The Promises and Challenges of Digital ...

CSULB COE Lecture - Robotics and Medicine

Over the past century, engineering has made numerous fundamental contribution to the field of medicine. From the most prosaic forms of engineering like sewer ...

The Gerson Miracle (2004) [Eng. Captions]

The Gerson Miracle introduces us to the physician who developed The Gerson Therapy more than 75 years ago, a therapy that has proven to cure cancer and ...

Organ donation

Organ donation is the donation of biological tissue or an organ of the human body, from a living or dead person to a living recipient in need of a transplantation.