Introduction to Radiomics and Artificial Intelligence: A Primer for Radiologists.

Health informatics and artificial intelligence (AI) are expected to transform the healthcare enterprise and the future practice of radiology. There is an increasing body of literature on radiomics and deep learning/AI applications in medical imaging. There are also a steadily increasing number of FDA cleared AI applications in radiology. It is therefore essential for radiologists to have a basic understanding of these approaches, whether in academia or private practice. In this article, we will provide an overview of the field and familiarize the readers with the fundamental concepts behind these approaches.

Applications of Artificial Intelligence in the Radiology Roundtrip: Process Streamlining, Workflow Optimization, and Beyond.

There are many impactful applications of artificial intelligence (AI) in the electronic radiology roundtrip and the patient's journey through the healthcare system that go beyond diagnostic applications. These tools have the potential to improve quality and safety, optimize workflow, increase efficiency, and increase patient satisfaction. In this article, we review the role of AI for process improvement and workflow enhancement which includes applications beginning from the time of order entry, scan acquisition, applications supporting the image interpretation task, and applications supporting tasks after image interpretation such as result communication. These non-diagnostic workflow and process optimization tasks are an important part of the arsenal of potential AI tools that can streamline day to day clinical practice and patient care.

Pattern of brexucabtagene autoleucel-related neurotoxicity on magnetic resonance imaging of the brain in a patient with relapsed/refractory B-cell acute lymphoblastic leukemia and prior leptomeningeal disease.

Immune effector cell-associated neurotoxicity syndrome (ICANS) secondary to chimeric antigen receptor T-cell therapy is common in adult patients with relapsed/refractory (R/R) B-cell acute lymphoblastic leukemia (ALL), but imaging findings during neurologic toxicity and their meaning have yet to be systematically described in this patient population. Brexucabtagene autoleucel (brexu-cel) is a CD19-directed autologous T-cell immunotherapy for the treatment of adult patients with R/R B-cell ALL that can enter the central nervous system. We present a case of an adult patient with R/R B-cell ALL and prior leptomeningeal disease who developed neurologic toxicity and new findings on magnetic resonance imaging of the brain while receiving brexu-cel. We interpret the patient's neuroimaging studies within clinical context to differentiate ICANS from active treatment of residual leukemia.

GCS 15: when mild TBI isn't so mild.

OBJECTIVE: The present study characterizes patients with the mildest of mild traumatic brain injury (TBI), as defined by a Glasgow coma score (GCS) of 15. METHODS: This is an IRB approved observational cohort study of adult patients who presented to the emergency department of a Level-1 trauma center, with the primary diagnosis of TBI and a GCS score of 15 on arrival. Data collected included demographic variables such as age, gender, race, mechanisms of injury, signs and symptoms including associated vomiting, seizures, loss of consciousness (LOC), alteration of consciousness (AOC), and post-traumatic amnesia (PTA).Pre- hospital GCS, Emergency Department (ED) GCS, and results of brain CT scans were also collected as well as patient centered outcomes including hospital or intensive care unit (ICU) admission, neurosurgical intervention, and in hospital death. Data were stored in REDCap (Research Electronic Data Capture), a secure, web- based application. Descriptive and inferential analysis was done using JMP 14.0 for the Mac. RESULTS: Univariate predictors of hospital admission included LOC, AOC, and PTA, all p < 0.0001. Patients admitted to ICU were significantly more likely to be on an antiplatelet or anticoagulant (P < 0.0001), have experienced PTA (p = 0.0025), LOC (p < 0.0001), or have an abnormal brain CT (p < 0.0001). Patients who died in the hospital were significantly more likely to be on an antiplatelet or anticoagulant (P = 0.0005. All who died in the hospital had intracranial hemorrhage on ED head CT, despite having presented to the ED with GCS of 15. Patients were also significantly more likely to have had vomiting (p < 0.0001). Patients who underwent neurosurgical intervention were significantly more likely to be male (P = 0.0203), to be on an antiplatelet or anticoagulant (P = < 0.0001) likely to have suffered their TBI from a fall (P = 0.0349), and experienced vomiting afterwards (P = 0.0193). CONCLUSIONS: This study underscores: 1) the importance of neuroimaging in all patients with TBI, including those with a GCS 15. Fully 10% of our cohort was not imaged. Extrapolating, these would represent 2.5% bleeds, and 1.47% fractures. 2) The limitations of GCS in classifying TBI, as patients with even the mildest of mild TBI have a high frequency of gross CT abnormalities.

The value of neurocognitive testing for acute outcomes after mild traumatic brain injury.

BACKGROUND: Traditionally, neurocognitive testing is performed weeks to months after head injury and is mostly performed on patients who continue to have symptoms or difficulties. In this study, we sought to determine whether these tests, when administered acutely, could assist in predicting short-term outcomes after acute traumatic brain injury (TBI). METHODS: This is an IRB-approved prospective study of adult patients who came to the emergency department of our Level-1 trauma center with TBI. Patients were enrolled prospectively after providing written informed consent and underwent three separate neurocognitive tests: the Galveston Orientation Amnesia Test (GOAT) the Rivermead Post-Concussion Survey Questionnaire (RPCSQ) and the Mini Mental Status Examination (MMSE). RESULTS: A lower GOAT score was significantly associated with hospitalization (P = 0.0212) and the development of post-concussion syndrome (P = 0.0081) at late follow-up. A higher RPCSQ score was significantly associated with hospital admission (P = 0.0098), re-admission within 30 days of discharge (P = 0.0431) and evidence of post-concussion syndrome (PCS) at early follow-up (P = 0.0004). A higher MMSE score was significantly associated with not being admitted to the hospital (P = 0.0002) and not returning to the emergency department (ED) within 72 hours of discharge (P = 0.0078). Lower MMSE was also significantly associated with bleeding or a fracture on the brain CT (P = 0.0431). CONCLUSIONS: While neurocognitive testing is not commonly performed in the ED in the setting of acute head injury, it is both feasible and appears to have value in predicting hospital admission and PCS. These data are especially important in terms of helping patients understand what to expect, thus, aiding in their recovery.

Understanding Why Patients Return to the Emergency Department after Mild Traumatic Brain Injury within 72 Hours.

INTRODUCTION: Although there are approximately 1.1 million case presentations of mild traumatic brain injury (mTBI) in the emergency department (ED) each year, little data is available to clinicians to identify patients who are at risk for poor outcomes, including 72-hour ED return after discharge. An understanding of patients at risk for ED return visits during the hyperacute phase following head injury would allow ED providers to develop clinical interventions that reduce its occurrence and improve outcomes. METHODS: This institutional review board-approved consecutive cohort study collected injury and outcome variables on adults with the purpose of identifying positive predictors for 72-hour ED return visits in mTBI patients. RESULTS: Of 2,787 mTBI patients, 145 (5%) returned unexpectedly to the ED within 72 hours of hospital discharge. Positive predictors for ED return visits included being male (p=0.0298), being black (p=0.0456), having a lower prehospital Glasgow Coma Score (p=0.0335), suffering the injury due to a motor vehicle collision (p=0.0065), or having a bleed on head computed tomography (CT) (p=0.0334). ED return visits were not significantly associated with age, fracture on head CT, or symptomology following head trauma. Patients with return visits most commonly reported post-concussion syndrome (43.1%), pain (18.7%), and recall for further clinical evaluation (14.6%) as the reason for return. Of the 124 patients who returned to the ED within 72 hours, one out of five were admitted to the hospital for further care, with five requiring intensive care unit stays and four undergoing neurosurgery. CONCLUSION: Approximately 5% of adult patients who present to the ED for mTBI will return within 72 hours of discharge for further care. Clinicians should identify at-risk individuals during their initial visits and attempt to provide anticipatory guidance when possible.

TBI ADAPTER: traumatic brain injury assessment diagnosis advocacy prevention and treatment from the emergency room--a prospective observational study.

There is no standard treatment algorithm for patients who present to the emergency department (ED) with acute traumatic brain injury (TBI). This is in part because of the heterogeneity of the injury pattern and the patient profile, and the lack of evidence-based guidelines, especially for mild TBI in adults. As TBI is seen more and more frequently in the ED, a standardized assessment would be beneficial in terms of efficiency. The authors present their ED approach to mild TBI evaluation in the ED, along with results to date. These data represent a prospective observational cohort study, where each patient provided individual, written informed consent.

Who gets post-concussion syndrome? An emergency department-based prospective analysis.

BACKGROUND: The objective of this study was to determine who gets post-concussion syndrome (PCS) after mild traumatic brain injury or head injury. METHODS: Patients presented within an hour of mild traumatic brain injury (mTBI). Written informed consent was obtained from all patients, who then provided detailed answers to surveys at the time of injury as well as at 1 week and 1 month follow-up. Statistical analyses were performed using JMP 11.0 for the Macintosh. RESULTS: The most commonly reported symptoms of PCS at first follow-up were headache (27%), trouble falling asleep (18%), fatigue (17%), difficulty remembering (16%), and dizziness (16%). Furthermore, only 61% of the cohort was driving at 1 week follow-up, compared to 100% prior to the injury. Linear regression analysis revealed the consumption of alcohol prior to head injury, the mechanism of head injury being a result of motor vehicle collision (MVC) or fall, and the presence of a post-injury headache to be significantly associated with developing PCS at 1 week follow-up, while the occurrence of a seizure post-injury or having an alteration in consciousness post-injury was significantly associated with developing PCS at 1 month follow-up. On multivariate regression analysis, the presence of a headache post-injury was the most robust predictor, retaining statistical significance even after controlling for age, gender, and presence of loss of consciousness (LOC), alteration of consciousness (AOC), post-traumatic amnesia (PTA), seizure, or vomiting. CONCLUSIONS: The results of this prospective study suggest that headache right after the head injury, an alteration of consciousness after the head injury, and alcohol consumption prior to the head injury are significant predictors of developing PCS, which occurs with equal frequency in men and women. Early identification of those who are at risk of developing PCS would diminish the burden of the injury and could potentially reduce the number of missed work and school days.

Impact of helmet use in traumatic brain injuries associated with recreational vehicles.

Objective. To study the impact of helmet use on outcomes after recreational vehicle accidents. Methods. This is an observational cohort of adult and pediatric patients who sustained a TBI while riding a recreational vehicle. Recreational vehicles included bicycles, motorcycles, and all-terrain vehicles (ATVs), as well as a category for other vehicles such as skateboards and scooters. Results. Lack of helmet use was significantly associated with having a more severe traumatic brain injury and being admitted to the hospital. Similarly, 25% of those who did wearing a helmet were admitted to the ICU versus 36% of those who did not (P = 0.0489). The hospital length of stay was significantly greater for patients who did not use helmets. Conclusion. Lack of helmet use is significantly correlated with abnormal neuroimaging and admission to the hospital and ICU; these data support a call for action to implement more widespread injury prevention and helmet safety education and advocacy.

TBI surveillance using the common data elements for traumatic brain injury: a population study.

BACKGROUND: To characterize the patterns of presentation of adults with head injury to the Emergency Department. METHODS: This is a cohort study that sought to collect injury and outcome variables with the goal of characterizing the very early natural history of traumatic brain injury in adults. This IRB-approved project was conducted in collaboration with our Institution's Center for Translational Science Institute. Data were entered in REDCap, a secure database. Statistical analyses were performed using JMP 10.0 pro for Windows. RESULTS: The cohort consisted of 2,394 adults, with 40% being women and 79% Caucasian. The most common mechanism was fall (47%) followed by motor vehicle collision (MVC) (36%). Patients sustaining an MVC were significantly younger than those whose head injury was secondary to a fall (P < 0.0001). Ninety-one percent had CT imaging; hemorrhage was significantly more likely with worse severity as measured by the Glasgow Coma Score (chi-square, P < 0.0001). Forty-four percent were admitted to the hospital, with half requiring ICU admission. In-hospital death was observed in 5.4%, while neurosurgical intervention was required in 8%. For all outcomes, worse TBI severity per GCS was significantly associated with worse outcomes (logistic regression, P < 0.0001, adjusted for age). CONCLUSION: These cohort data highlight the burden of TBI in the Emergency Department and provide important demographic trends for further research.

Assessment of the tracer delay effect in whole-brain computed tomography perfusion: results in patients without known neuroanatomic abnormalities.

OBJECTIVE: Whole-brain computed tomography perfusion (CTP) data sets generated by tracer delay-insensitive singular value decomposition plus (SVD+) and standard singular value decomposition (sSVD) deconvolution algorithms were evaluated to quantify relatedness and discrepancies in CTP results. METHODS: Twenty females with symmetrical hemispheric CTP maps indicative of brain tissue without apparent abnormalities were studied. Tissue-specific CTP values were analyzed. RESULTS: Standard SVD values were higher than SVD+ for cerebral blood flow. Other CTP values had minimal differences across brain regions. All simple linear regression models were statistically significant (P < 0.05) except for cerebral blood flow in white matter (P = 0.06). Cerebral blood volume had a good model fit, and mean transit time, a poor fit. CONCLUSIONS: Corresponding fitted CTP values for sSVD and SVD+ based on regression equations for brain-tissue types are presented. Additional research is required to compare SVD+ and sSVD in disease states when significant hemodynamic brain alterations are present.

Assessment of tracer delay effect in whole-brain computed tomography perfusion: results for selected regions of interest in middle cerebral artery acute ischemic strokes.

OBJECTIVE: The objective of this study was to compare the variability of computed tomography perfusion (CTP) results in identical data sets of middle cerebral artery (MCA) acute ischemic stroke (AIS) generated by standard singular value decomposition (sSVD) deconvolution and tracer delay-insensitive singular value decomposition (SVD+) algorithm analyses. METHODS: Whole-brain 320-detector-row CTP data sets from 9 unilateral MCA AIS cases and 9 controls were retrospectively analyzed. Computed tomography perfusion values for the combined core/penumbra, contralateral hemispheres and arterial territories were measured and compared with literature values. Simple linear regression models are provided to predict corresponding SVD+ value and sSVD CTP values. RESULTS: In the core/penumbra, sSVD generated lower cerebral blood flow (CBF) values, higher mean transit time (MTT) values, and a broader range of CBF and MTT values as compared with SVD+. Mean transit time value differences between the core/penumbra and contralateral hemispheres were statistically significant using sSVD, whereas those of SVD+ were not. Goodness of fit between algorithms for the core/penumbra was lower for CBF (0.483) and MTT (0.494), as compared with time to peak (0.891) and cerebral blood volume (0.997). CONCLUSIONS: In this study using identical source data for patients with MCA AIS, use of either sSVD or SVD+ analyses created statistically significant differences in the CTP value results. Tracer delay-sensitive and -insensitive algorithms impact CTP results in AIS and controls, highlighting the need to pursue additional studies that assess the variability, accuracy, and clinical implications of CTP results generated when using heterogeneous deconvolution algorithms.

Radiotherapy and concomitant intraarterial cisplatin (RADPLAT) for advanced squamous cell carcinomas of the head and neck.

PURPOSE: Review of the University of Florida experience in treating advanced squamous cell carcinomas of the head and neck with concurrent intraarterial cisplatin and radiotherapy (RADPLAT). PATIENTS AND METHODS: Thirty-five patients with Stage III or IV head and neck squamous cell carcinomas were treated between November 2000 and October 2001 with 3 to 4 cycles of weekly intraarterial cisplatin (150 mg/m2) and a rapid infusion of the neutralizing agent sodium thiosulfate (9 g/m2), followed by a continuous infusion of sodium thiosulfate (12 g/m2), with concomitant radiotherapy (RT). The primary site was treated with 70 Gy in 35 fractions with 6 MV photons, and the low neck received 50 to 70 Gy depending on nodal involvement. Median follow-up for all patients was 4.06 years; median follow-up for living patients was 4.34 years. RESULTS: The 4-year outcomes were: local-regional control, 78%; distant metastasis-free survival, 83%; cause-specific survival, 65%; and survival, 57%. The majority of patients with recurrent disease failed at the primary site (73%); the remaining 27% of patients with recurrent disease failed only in distant sites. Nine patients required a neck dissection after chemoradiation because of suspected residual disease; 2 of 9 patients had residual tumor in the neck nodes. Severe osteoradionecrosis occurred in 26% of patients, and 9 of 35 patients (26%) required a permanent gastrostomy. CONCLUSIONS: RADPLAT results in excellent local-regional control and survival in patients with advanced squamous cell carcinomas of the head and neck. Outcomes from the University of Florida are comparable to those reported by other institutions. However, in our limited experience, the likelihood of osteoradionecrosis and permanent gastrostomy may be higher than what might be anticipated after RT alone or combined with intravenous chemotherapy.

Orbital hemangioma requiring enucleation in a premature infant.

A preterm male infant developed a rapidly growing, treatment-resistant orbital hemangioma. Despite aggressive management, the patient required enucleation of his right eye and image-guided ethanol sclerosis to limit tumor progression intracranially.

Delayed exacerbation of third nerve palsy due to aneurysmal regrowth after endovascular coil embolization.

A 72-year-old woman with a painful left third cranial nerve palsy due to a basilar artery aneurysm situated between the superior cerebellar and posterior cerebral arteries was treated with Guglielmi detachable coils (GDCs). Despite a good initial angiographic result with a small residual neck and improvement in the ocular motility and pain, the patient experienced worsening of the third cranial nerve palsy 15 months later. Cerebral angiography confirmed coil compaction with aneurysmal regrowth. A second endovascular coil embolization resulted in complete obliteration of the aneurysm. The patient experienced complete resolution of the pain and partial resolution of the third cranial nerve palsy. In some patients, a small residual aneurysm neck after endovascular embolization therapy with GDCs can result in delayed aneurysmal regrowth due to coil compaction. Clinical manifestations may herald this dangerous regrowth.

A red eye and then a really red eye.

A 90-year-old woman presented with a 4-week history of a presumed infectious conjunctivitis resistant to topical antibiotic medications. Examination revealed tortuous, dilated conjunctival vessels in the right eye, retinal hemorrhages, and an orbital bruit suggestive of a carotid-cavernous sinus fistula (CCF). While awaiting a magnetic resonance imaging study, she returned to the clinic the next day with a painful, swollen right eye and an intraocular pressure of 69 mm Hg. A cerebral arteriogram confirmed a direct CCF. Because of the tortuosity of the systemic vascular anatomy, a right carotid artery cut-down with balloon occlusion was performed with successful closure of the fistula and prompt resolution of the orbital congestion. This case illustrates the spectrum of subtle to conspicuous ocular manifestations that can be seen in patients with CCF and its potential to present as an emergency. CCF should be included in the differential diagnosis of an "atypical" red eye. Recognition of arteriolized conjunctival vessels and auscultation of an orbital bruit raises the possibility of a CCF, requiring prompt diagnostic studies.

Effect of radiographic techniques (kVp and mAs) on image quality and patient doses in digital subtraction angiography.

We investigated how varying the x-ray tube voltage and image receptor input exposure affected image quality and patient radiation doses in interventional neuroradiologic imaging. Digital subtraction angiography (DSA) images were obtained of a phantom with 1 mm diameter vessels containing iodine at concentrations between 4.5 and 50 mg/cc. The detection threshold concentration of iodine was determined by inspecting DSA images obtained at a range of x-ray tube voltages and input exposure levels. Surface doses were obtained from measured x-ray tube output data, and corresponding values of energy imparted were determined using the exposure-area product incident on the phantom. In one series of experiments, the air kerma at the image intensifier (X) was varied between 0.44 microGy per frame and 8.8 microGy per frame at a constant x-ray tube voltage of 70 kVp. In a second series of experiments, the tube voltage was varied between 50 and 100 kVp, and the mAs adjusted to maintain a constant exposure level at the input of the image intensifier. At a constant x-ray tube voltage, the surface dose and energy imparted were directly proportional to the input exposure per frame used to acquire the DSA images. On our DSA system operated below 2.2 microGy per frame, the threshold iodine concentration was found to be proportional to X(-0.57), which is in reasonable agreement with the theoretical prediction for a quantum noise limited imaging system. Above 2.2 microGy per frame, however, the threshold iodine concentration was proportional to X(-0.26), indicating that increasing the input exposure above this value will only achieve modest improvements in image quality. At a constant image intensifier input exposure level, increasing the x-ray tube voltage from 50 kVp to 100 kVp reduced the surface dose by a factor of 6.1, and the energy imparted by a factor of 3.5. The detection threshold iodine concentration was found to be proportional to kVp(n), where n was 2.1 at 1.1 microGy per frame, and 1.6 at 3.9 microGy per frame. For clinical situations that can be modeled by a uniform phantom, reducing the x-ray tube voltage rather than increasing the exposure level would best achieve improvements on our DSA imaging system performance.