Caffey's landmark article of 1946 noted an association between healing long-bone fractures and chronic subdural hematomas in infancy, and it was the first to draw attention to physical abuse as a unifying etiology.  In 1962, Caffey and Kempe et al proposed manhandling and violent shaking as mechanisms of injury and emphasized the acute and long-term sequelae of abuse as serious public health problems.  Since these early reports, investigators have more clearly defined the pathophysiology of abusive injuries (see the images below). Community-service and law-enforcement authorities have taken a role in protecting potential victims and in prosecuting perpetrators.
In the United States, in 2015, there were 683,000 victims of child abuse, and approximately 1670 children died of abuse and neglect, a rate of 2.25 per 100,000 children. Almost 75% of those deaths occurred in children younger than 3 years. Most reports of abuse were submitted by educational personnel (18.4%) and legal and law enforcement personnel (18.2%). Approximatley 9% of reports were submitted by medical personnel. 
For infants and children younger than 2 years, a skeletal survey should be performed as the initial screening examination when child abuse is being considered. [4, 5, 6, 7] The survey consists of the acquisition of a series of images collimated to each body region.  The series includes frontal and lateral views of the skull, frontal and lateral views of the spine, frontal views of the chest (ribs) and pelvis, and frontal views of the extremities, including the hands and feet. [9, 10]
The skeletal survey is widely available and inexpensive in comparison with alternative imaging modalities. Other important advantages of the skeletal survey include a high sensitivity for most acute and healing fractures and a relatively low radiation burden. 
A babygram, in which the entire skeleton is depicted on a single image, is not an appropriate substitute for a properly performed survey. Geometric distortion and varying exposures are unacceptable limitations of this image. Use of a high-detail, high-contrast, screen-film system with good spatial resolution is mandatory. All abnormal areas should be viewed on at least 2 projections.
Computed tomography (CT) scanning of the head is the imaging modality of choice for evaluating a child with acute neurologic findings or retinal hemorrhage on physical examination. It is more sensitive to acute intracerebral and extra-axial hemorrhages than is magnetic resonance imaging (MRI). Brain MRI may be helpful as an adjunct for the evaluation of axonal shear injuries and for a precise dating of intracranial hemorrhage. [12, 13]
Subarachnoid hemorrhages (SAHs) are best demonstrated on CT scans. The use of MRI to detect acute SAH remains controversial. However, MRI is superior to CT scanning for differentiating a hypoattenuating subdural hematoma from cerebrospinal fluid (CSF) and for detecting small and chronic extra-axial fluid collections. (See the images below; both images reveal a subdural hematoma, the first with a CT scan and the second with an MRI scan.)
(See the images below.)
Most child abuse–related injuries are readily detectable during imaging. Radiologic examination is the mainstay for diagnosing physical abuse in children. [14, 15, 16] Careful correlation of the observed radiologic findings with the proposed mechanism of injury and with the child's clinical status is imperative in the evaluation of any child in whom abuse is suspected. If such correlation is not performed, important clues of an inflicted injury may be overlooked, and the child may be returned to an abusive environment—with potentially disastrous consequences. [17, 18, 19, 20]
The radiologist has important medical and legal roles in cases of child abuse. The radiologist may be the first to raise a question of abuse if characteristic or unexplained findings are encountered during imaging. Immediate, direct communication with the referring physician is imperative in such cases.
A radiologist's testimony regarding findings of possible abuse may be required, and the radiologist may be asked to give an opinion as to the likely age of the child's fractures and about the possibility of alternative diagnoses. Familiarity with radiographic patterns and mechanisms of abuse generally allows the radiologist to give an interpretation with a high degree of certainty.
Because imaging studies document occult injuries, they may justify the implementation of protective measures when the patient's clinical presentation suggests abuse.
Kleinman et al reported the utility of high-detail, postmortem radiography in identifying skeletal injuries that had otherwise been overlooked. 
Skeletal injury is the most common form of injury (excluding external soft-tissue injuries) in child abuse. Fractures are documented in 11-55% of physically abused children. 
ACR Appropriateness Criteria
The appropriate imaging of pediatric patients being evaluated for suspected physical abuse depends on the age of the child, the presence of neurologic signs and symptoms, and evidence of visceral thoracic or abdominopelvic injuries.
A skeletal survey is indicated in the initial imaging evaluation of children 24 mo of age or younger. In
older children, it is usually appropriate to target imaging to the area(s) of suspected injury.
Skeletal survey and CT of the head without contrast are indicated in the emergent/initial imaging evaluation of a child with neurologic signs and symptoms, complex skull fracture, apnea, multiple fractures, spine trauma, or facial injury. These examinations are not indicated for general screening.
MRI of the head may provide additional diagnostic information to head CT in about 25% of children.
MRI of the cervical spine should be considered at the time of head MRI, because unsuspected injury (usually ligamentous) may be present in over 33% of children with intracranial injury.
Skeletal survey and CT of the chest/abdomen/pelvis with IV contrast are indicated if there are signs or symptoms of intrathoracic or intra-abdominal visceral injury (eg, abdominal pain/distention/bruising, abnormal liver, or pancreatic enzymes).
In children 24 mo of age or younger with equivocal skeletal survey or with a high clinical suspicion for abuse and a negative initial skeletal survey, a repeat limited/focused skeletal survey performed at 2 wk may add diagnostic information.
The resulting fracture may cross the diaphysis in an oblique or transverse plane, or it may create the highly specific and classic metaphyseal lesion (CML). The classic metaphyseal lesion is also referred to as a corner fracture or a bucket-handle fracture. A CML occurs when a torsional force is applied to the immature primary spongiosa adjacent to a cartilaginous growth plate.
Rib fractures occur when a compressive force is applied simultaneously to the sternum and to the costovertebral junction during violent shaking as the perpetrator compresses the child's chest using both hands. 
The posterior ribs are most commonly fractured, because the greatest force is imparted to the articulation of the head and to the neck of the rib with the transverse process of the vertebral body.
However, fractures are not limited to the posterior aspects of the ribs. Anterolateral fractures are also common. Rib fractures are typically noted at several contiguous levels; they are frequently bilateral.
(See the image below.)
Head injury accounts for 80% of deaths associated with abuse in children younger than age 2 years. Mechanisms of injury include forceful shaking, either by itself or accompanied by abrupt impact. [12, 13, 25, 26] (See the images below.)
In an acute setting, CT scanning is more readily available and more cost-effective than MRI. MRI is used as a problem-solving modality when CT findings are unexplained or confusing.
CT scanning of the abdomen is indicated if abdominal injury is suspected. CT scanning demonstrates visceral injuries and retroperitoneal hematomas.
Gastrografin upper-GI study under fluoroscopic guidance is occasionally indicated for evaluating submucosal hemorrhages of the duodenum or for identifying perforation of the duodenum.
Skeletal scintigraphy may be used when clinical suspicion remains high despite normal findings on a skeletal survey.  Advantages of scintigraphy include increased sensitivity for acute posterior rib fractures, because the spine tends to obscure these injuries on radiographic examination. Fractures of the spine may be better depicted with scintigraphy than with radiography.
Limitations of techniques
Fractures that are parallel or nearly parallel to the section orientation may be missed during CT scanning. Therefore, radiography of the skull is preferred over CT scanning for examining these injuries.
The disadvantages of scintigraphy include diminished sensitivity in detecting skull fractures and CMLs, as these are contiguous with the normal isotope-avid growth plates. In addition, because all abnormal sites must be confirmed radiographically, an osseous survey is the preferred initial examination.
Scintigraphy is also limited because of its expense, lack of availability, amount of gonadal radiation exposure relative to plain radiography, and lack of reader expertise.
The radiographic appearance of a CML of the long bone is highly specific for physical abuse. In infants younger than 1 year, these fractures are usually found in the distal femur, the proximal tibia, the distal tibia, or the proximal humerus as a result of a series of microfractures across the metaphysis. The fracture is parallel to the growth plate and perpendicular to the long axis of the bone. Differential horizontal motion across the metaphysis is a feature of abusive injury, but it is not characteristic of falls or blunt trauma.
The fracture line of the CML courses through the primary spongiosa of the metaphysis, and the metaphyseal fragment tends to be thicker peripherally than centrally. Depending on how it is viewed, the fracture may appear as a corner injury or as a bucket-handle fracture.
Other injuries that are highly specific to child abuse are posterior rib fractures (see the image below) and fractures of the scapula, spinous process, and sternum, because these bones are ordinarily difficult to break. Fractures in different stages of healing are also highly specific findings of child abuse.
Degree of confidence
Acute fractures may be missed on plain radiographs, especially if optimal technique is not followed. A follow-up skeletal survey may be obtained 2 weeks after the initial study when abuse is strongly suspected but initial radiographic findings are negative.
Skeletal dysplasias, especially osteogenesis imperfecta, may mimic injuries resulting from physical abuse. Osteogenesis imperfecta types I and IV may be confused with abusive injury.
Differentiating features include the invariable presence of osteopenia in osteogenesis imperfecta and telltale abnormalities of modeling and bowing of the long bones. Other helpful clues are wormian (intrasutural) bones of the skull and the site of the fracture. Abusive injuries and the fractures seen in osteogenesis imperfecta typically occur in different locations.
Other causes of multiple fractures include Menkes syndrome and congenital indifference to pain. Additional radiographic or clinical signs of these diseases are invariably present, and they are rarely confused with findings of child abuse.
Periosteal reaction may be seen with congenital syphilis, leukemia, or Caffey disease.
A fracture-dislocation of the hip occurring before the femoral head ossifies may be mistaken for developmental dysplasia of the hip. Hip ultrasonography is indicated to differentiate the conditions under the appropriate clinical circumstances.
CT scanning of the head is useful in patients with head injury, especially those with skull fractures (see the image below). With advancements in technology, low-dose techniques are now available for pediatric head CT scanning, and automatic regional adjustments in radiation doses are now possible during scanning. [12, 13, 25, 26]
CT scans may reveal subarachnoid, subdural, or epidural hemorrhages. Subdural hemorrhage (see the first and second images below), especially interhemispheric bleeds, and SAH (see the third image below) are common injuries of abuse.
In infants and small children, subdural hemorrhage indicates rotational brain movement or shear injury; it is commonly associated with retinal hemorrhages and retinoschisis. Epidural hemorrhage is more common with accidental injury than with inflicted injury; it may result from falls from a relatively short distance.
Focal or diffuse cerebral edema (see the images below) may also be seen. This edema may be a manifestation of primary injury or a consequence of hypoxia resulting from strangulation, suffocation, posttraumatic apnea, or other causes. Edema may manifest as sulcal effacement, loss of distinction between gray matter and white matter, and/or cerebral hypoattenuation (ie, the reversal sign).
The abdominal organs most commonly injured in cases of child abuse are the duodenum and the proximal jejunum (see the image below). Hematomas, lacerations, or mesenteric injuries may be evident on CT scans.
Solid-organ contusions, lacerations, hemoperitoneum, and pneumoperitoneum are easily diagnosed on abdominal and pelvic CT scans. None of these organ injuries are specific for child abuse, but making the diagnosis is important for the clinical treatment of the patient.
CT scanning has had a major effect on the early diagnosis of visceral injuries of the liver, spleen, pancreas, and kidneys (see the images below). Grading of injuries by means of CT scanning facilitates patient care and surgical planning.
Degree of confidence
Subdural hematomas may be hypoattenuating on the day of trauma because of various factors. For example, in children with anemia, the blood may be hypoattenuating. Disruption of the arachnoid membrane during trauma may allow CSF to mix with the subdural hematoma, diluting the hemorrhage and reducing its attenuation on CT scans.
On CT scanning, nondepressed fractures of the calvaria may be missed, especially if the fracture lies in an axial plane parallel to the plane of imaging.
Nonhemorrhagic brain contusions, early global ischemia, and shear injury (diffuse axonal injury) may be missed on CT scanning because of limitations in resolution that cause subtle differences in tissue attenuation to be overlooked.
Magnetic Resonance Imaging
The brain parenchymal injuries most commonly observed in cases of child abuse are shear injury, edema, and contusion. Shear injury most often occurs at the gray matter–white matter junction, at the corpus callosum, and at the midbrain. Findings appear as hyperintense foci on T2-weighted or inversion recovery images.
Brain contusions are admixtures of edema, hemorrhage, and necrosis. They usually are cortical and are adjacent to the bony surfaces; they result from direct contact forces. Contusions are distinctly rare in infants. Shear injury and brain edema account for most abnormalities.
Degree of confidence
MRI is useful for detecting small, extra-axial fluid collections; early global ischemia; and shear injury, which may be missed on CT scans. MRI also aids in dating hemorrhage, depending on the blood products present in the collections. Diffusion-weighted images are useful in detecting early ischemia.
MRI is superior to CT scanning for differentiating hypoattenuating subdural hemorrhage from prominent extracerebral spaces of infancy (see the image below).
Scattered reports mention that MR spectroscopy may provide complementary information in diagnosing posttraumatic neuronal loss. 
The use of MRI to detect acute SAH remains controversial. Although one may observe SAH on fluid-attenuated inversion recovery (FLAIR) or susceptibility on MRI scans, CT is definitely preferable to MRI, because it easily depicts sulcal hyperattenuation, which confirms early SAH.
Head ultrasonography has no role in the evaluation of acute abusive injury.
In unstable patients being examined in the emergency department, portable abdominal ultrasonography may be used for initial screening for visceral injuries and free fluid.
Ultrasonography is less sensitive than CT scanning in the detection of acute intracranial hemorrhage and cerebral edema, especially after the fontanelles partially close.
Limitations of ultrasonography include its relatively poor depiction of the posterior fossa and of the far convexities of the brain, as well as its inability to demonstrate the nature of abnormal fluid collections. In the presence of excessive gas or ileus, abdominal ultrasonography may be difficult to perform.
Scintigraphy complements the radiographic skeletal survey.
Abnormalities detected on scintigraphy should be confirmed radiographically. Scintigraphy improves sensitivity in the detection of acute rib fractures, especially those at the costovertebral junction.
Skeletal scintigraphy has high sensitivity and low specificity in cases of child abuse. On bone scanning, a hot lesion may be indicative of trauma, infection, or healing infarction. Therefore, correlation with radiographs is always necessary when abnormalities are identified on scintigrams.
Skull fractures are difficult to identify on scintigrams, and plain radiographs of the skull are always required.
On scintigraphy, a CML may be overlooked, because it lies adjacent to a metabolically active and isotope-avid epiphyseal growth plate.