Artificial intelligence (AI), with its subdivisions (machine and deep learning), is a new branch of computer science that has shown impressive results across a variety of domains. The applications of AI to medicine and biology are being widely investigated. Medical specialties that rely heavily on images, including radiology, dermatology, oncology and ophthalmology, were the first to explore AI approaches in analysis and diagnosis. Applications of AI in ophthalmology have concentrated on diseases with high prevalence, such as diabetic retinopathy, retinopathy of prematurity, age-related macular degeneration (AMD), and glaucoma. Here we provide an overview of AI applications for diagnosis, classification, and clinical management of AMD and other macular dystrophies.
INTRODUCTION: We evaluated the associations of clinical and demographic characteristics with visual acuity (VA) with over 5 years in a subspecialty noninfectious uveitis population. METHODS: Retrospective data from 5,530 noninfectious uveitis patients were abstracted by expert reviewers, and contemporaneous associations of VA with demographic and clinical factors were modeled. RESULTS: Patients were a median of 41 years old, 65% female, and 73% white. Eyes diagnosed ≥5 years prior to cohort entry had worse VA (-1.2 lines) than those diagnosed <6 months prior, and eyes with cataract surgery performed prior to entry had worse VA (-5.9 lines) than those performed during follow-up. Vitreous haze (-4.2 lines for 3+ vs quiet), hypotony (-2.5 lines for ≤5 mm Hg vs 6-23 mm Hg), and CNV (-1.8 lines) all were strongly associated with reduced VA. CONCLUSION: Factors associated with reduced VA included well-known structural complications, and lack of subspecialty care during cataract surgery.
PURPOSE: The International Neonatal Consortium recently published a proposed retinopathy of prematurity (ROP) activity scale intended for use in clinical trials after validation. The aim of this study was to validate the ROP activity scale (ROP-ActS) in a ROP screened cohort with protocol based collected data by evaluating the ability of the ROP-Act scores to predict ROP treatment. In addition, we aimed to evaluate the scale's sensitivity characteristic of disease severity by studying association with gestational age (GA) in comparison with conventionally used ROP stage and zone. METHODS: A cohort of 535 preterm infants with 3324 ROP examinations with an end-point of ROP treatment or end of screening in Gothenburg, Sweden, was included. Median GA was 28.1 weeks, 47.5% were girls, and 74 (13.8%) infants were treated for ROP. The validation was performed by estimating probabilities for ROP treatment, and by applying logistic and linear regression. RESULTS: The original ROP-ActS was overall well-ordered with respect to ability to predict ROP treatment but could be improved by re-ordering score 3 (zone II stage 1) and 5 (zone III stage 3) based on our clinical cohort data. The modified ROP-ActS was superior to ROP stage and zone in the prediction analysis of ROP treatment. Modified ROP-ActS was more strongly related to GA than currently used ROP stage, but not zone. CONCLUSION: In the studied cohort, the modified ROP-ActS could better predict ROP treatment compared to ROP stage and zone. Retinopathy of Prematurity Activity Scale (ROP-ActS) had a superior sensitivity characteristic studied through association to GA than conventionally used ROP stage.
BACKGROUND/AIMS: As swept-source optical coherence tomography (SS-OCT) simultaneously obtains 128 meridional scans, it is important to identify which scans are playing the main role in classifying gonioscopic angle closure to simplify the analysis. We aimed to evaluate the diagnostic performance of every meridional scan in its ability to detect gonioscopic angle closure. METHODS: Observational study with 2027 phakic subjects consecutively recruited from a community polyclinic. Gonioscopy and SS-OCT were performed. Gonioscopic angle closure was defined as non-visibility of the posterior trabecular meshwork in ≥180° of the angle, while SS-OCT was defined as iridotrabecular contact anterior to the scleral spur. The area under the receiver operating characteristic curve (AUC) was calculated to assess the diagnostic performance of each single scan, the sequential anticlockwise cumulative effect of those single scans and different combinations of them. RESULTS: The AUCs of each scan ranged from 0.73 to 0.82. The single scan at 80°-260° had the highest AUC (0.82, 95% CI 0.79 to 0.84) and performed significantly better than most of the temporonasal scans (from 0° to 52° and from 153° to 179°). The superoinferior scans achieved higher AUCs compared with the temporonasal ones. When assessing the cumulative effect of adding individual scans consecutively, the peak AUC (0.80) was obtained when considering the superoinferior scans closer to 80°-85°, but no further positive cumulative effect was seen when adding the rest of the temporonasal scans of the circumference. CONCLUSIONS: In conclusion, the single SS-OCT scan at 80°-260° had the highest diagnostic performance. Our study suggests that the 360° evaluation may not translate to better clinical utility for detection of gonioscopic angle closure.
PURPOSE: Quantification of dark adaptation (DA) response using the conventional rod intercept time (RIT) requires very long testing time and may not be measurable in the presence of impairments due to diseases such as age-related macular degeneration (AMD). The goal of this study was to investigate the advantages of using area under the DA curve (AUDAC) as an alternative to the conventional parameters to quantify DA response. METHODS: Data on 136 eyes (AMD: 98, normal controls: 38) from an ongoing longitudinal study on AMD were used. DA was measured using the AdaptDx 20 min protocol. AUDAC was computed from the raw DA characteristic curve at different time points, including 6.5 min and 20 min (default). The presence of AMD in the given eye was predicted using a logistic regression model within the leave-one-out cross-validation framework, with DA response as the predictor while adjusting for age and gender. The DA response variable was either the AUDAC values computed at 6.5 min (AUDAC6.5) or at 20 min (AUDAC20) cut-off, or the conventional RIT. RESULTS: AUDAC6.5 was strongly correlated with AUDAC20 (β=86, p<0.001, R=0.87). The accuracy of predicting the presence of AMD using AUDAC20 was 76%, compared with 79% when using RIT, the current gold standard. In addition, when limiting AUDAC calculation to 6.5 min cut-off, the predictive accuracy of AUDAC6.5 was 80%. CONCLUSIONS: AUDAC can be a valuable measure to quantify the overall DA response and can potentially facilitate shorter testing duration while maintaining diagnostic accuracy.
PRECIS: A cross-sectional sample of the United States ophthalmology residency graduating class of 2018 revealed that 18.4% of residents logged less than 5 traditional glaucoma surgeries, and 63.4% logged at least one microinvasive glaucoma surgery. PURPOSE: Describe the state of microinvasive glaucoma surgery in United States ophthalmology residency training and propose a glaucoma procedure classification system for residents' surgical case logs. METHODS: Deidentified case logs from residents graduating in 2018 were requested from United States residency program directors. RESULTS: Case logs were received for 152/488 (31%) residents from 36/115 (31%) programs. The mean number of traditional glaucoma surgeries per resident was 9.0±5.9 (range 0-31). The mean number of microinvasive glaucoma surgery per resident was 5.2±8.9 cases (range 0-58). There were 28/152 (18.4%) residents from 16/36 (44.4%) programs who logged <5 traditional glaucoma surgeries, and 3/152 (2.0%) residents from 3/36 (8.3%) programs who logged zero traditional glaucoma surgeries. There were 98/152 (64.5%) residents from 32/36 (88.8%) programs who logged <5 microinvasive glaucoma surgery as primary surgeon, and 48/152 (31.6%) residents from 25 of 36 (69.4%) programs who logged zero microinvasive glaucoma surgery. There were 104/152 (63.4%) residents from 33/36 (91.6%) programs who logged at least 1 microinvasive glaucoma surgery; there were 3/36 (8.3%) residency programs where no resident logged any microinvasive glaucoma surgery. CONCLUSIONS: United States ophthalmology residents' microinvasive glaucoma surgery experience varies widely. Residents can satisfy glaucoma surgery requirements with some microinvasive glaucoma surgery, even in the absence of adequate traditional glaucoma surgeries. We propose a residency case log classification system that better reflects the growing role of microinvasive glaucoma surgery in clinical practice and helps ophthalmic educators more accurately track procedures requiring related skills.
BACKGROUND: The severity and extent of microaneurysms (MAs) have been used to determine diabetic retinopathy (DR) severity and estimate the risk of DR progression over time. The recent introduction of ultrawide field (UWF) imaging has allowed ophthalmologists to readily image nearly the entire retina. Manual counting of MAs, especially on UWF images, is laborious and time-consuming, limiting its potential use in clinical settings. Automated MA counting techniques are potentially more accurate and reproducible compared to manual methods. METHOD: Review of available literature on current techniques of automated MA counting techniques on both ultrawide field (UWF) color images (CI) and fluorescein angiography (FA) images. RESULTS: Automated MA counting techniques on UWF images are still in the early phases of development with UWF-FA counts being further along. Early studies have demonstrated that these techniques are accurate and reproducible. CONCLUSION: Automated techniques may be an appropriate option for detecting and quantifying MAs on UWF images, especially in eyes with earlier DR severity. Larger studies are needed to appropriately validate these techniques and determine if they add substantially to clinical practice compared to standard DR grading.
Since the new coronavirus known as 2019-nCoV (severe acute respiratory syndrome coronavirus 2, SARS-CoV-2) has widely spread in Wuhan, China, with severe pneumonia, scientists and physicians have made remarkable efforts to use various options such as monoclonal antibodies, peptides, vaccines, small-molecule drugs and interferon therapies to control, prevent or treatment infections of 2019-nCoV. However, no vaccine or drug has yet been confirmed to completely treat 2019-nCoV. In this review, we focus on the use of potential available small-molecule drug candidates for treating infections caused by 2019-nCoV.
PURPOSE: To compare onset times of glaucoma progression among different glaucoma tests: disc photography (DP), visual field (VF) testing, two-dimensional (2D) retinal nerve fiber layer (RNFL) thickness, and three-dimensional (3D) spectral-domain optical coherence tomography (SD-OCT) neuroretinal rim measurements. DESIGN: Prospective longitudinal cohort study PARTICIPANTS: One hundred and twenty-four eyes of 124 open angle glaucoma patients METHODS: Over a 5-year period, 124 open angle glaucoma patients had yearly DP, VFs, SD-OCT RNFL thickness scans, and optic nerve volume scans (Spectralis, Heidelberg Engineering, Heidelberg, Germany), all performed on the same day. From high-density optic nerve volume scans, custom-built software calculated the minimum distance band (MDB) thickness, a 3D neuroretinal rim parameter. Patients were classified as glaucoma progressors or non-glaucoma progressors using event-based analysis. Progression by DP and VF occurred when 3 masked glaucoma specialists unanimously concurred. Progression by RNFL and MDB thickness occurred if there was change greater than test-retest variability. Kaplan-Meier curves were constructed to analyze time-to-progression data. Kappa coefficients were used to measure agreement of progressing eyes among modalities. MAIN OUTCOME MEASURES: Time to glaucoma progression among all 4 modalities RESULTS: Global MDB thickness detected glaucoma progression in the highest percentage of eyes (52.4%) compared to DP (16.1%, P<0.001), and global RNFL thickness (15.3%, P<0.001) respectively. Global MDB thickness detected glaucoma progression earlier than either DP (23 versus 44 months; P<0.001) or global RNFL thickness (23 versus 33 months; P<0.001). Among MDB progressing eyes, 46.2% were simultaneously or later confirmed by other conventional modalities. Agreement of glaucoma progressing eyes for all 4 modalities in paired fashion were slight to fair (k=0.095-0.300). CONCLUSION: High-density 3D SD-OCT neuroretinal rim measurements detected glaucoma progression approximately 1-2 years earlier compared to current clinically available structural tests (i.e. disc photos and 2D RNFL thickness measurements).
Primary central nervous system lymphoma-ophthalmic variant (PCNSL-O) is an ocular subset of PCNSL predominantly involving subretinal pigment epithelium (RPE) space, retina, and vitreous. The ophthalmic manifestations can precede, occur simultaneously, or follow other compartments of the CNS. Clinical trials have resulted in a significantly improved outcome in PCNSL patients over the past 2 decades, with a higher proportion of patients receiving frontline high dose methotrexate-based polychemotherapy regimens with curative intent; however, the current management of PCNSL-O remains controversial due to lack of prospective data. The goals of PCNSL-O treatment are both to achieve local (ocular) control and to prevent tumor-specific mortality from further CNS involvement. Despite achieving high rates of ocular control with intravitreal agents like methotrexate and rituximab, the overall survival is poor, as 65-85% of patients eventually succumb to CNS disease. Few studies define the role of systemic chemotherapy with/without local treatment as a first line induction treatment for PCNSL-O considering limiting factors such as ocular penetration of systemically administered drugs and treatment related neurotoxicity. Also, the role of adjuvant treatment for PCNSL-O to prevent CNS progression and to improve overall survival is unknown. In this systematic review of the literature, we analyze treatment outcomes of various regimens (local, systemic, and combination) in terms of local control, CNS progression, and overall survival.
Historically, surgical access to orbital tumors has required a transcutaneous, transconjunctival or transcranial approach. Resection of orbital tumors is notoriously challenging due to the surrounding dense network of critical structures in a confined bony cavity. Advances in endoscopic endonasal surgery, initially used for sinonasal and skull base conditions, have allowed for expansion of its applications beyond the sinorbital interface. In the past decade, the evolution of techniques has enabled a purely endoscopic, minimally invasive approach to medially located orbital pathology with good outcomes. With experience and multidisciplinary collaboration between orbit and rhinologic surgeons, this has expanded to allow for a safe and effective transnasal approach to nearly all regions of the orbit with or without assistance from the orbital side. This review summarizes the relevant anatomy, variations of surgical approaches, and literature regarding outcomes of the endoscopic endonasal approach to orbital tumors.
To analyze functional and anatomical response patterns to dexamethasone (DEX) implant in diabetic macular edema (DME), to describe proportion of responders and non-responders, and to propose a new DME grading system. Retrospective, multicenter, observational cohort study. Naïve and non-naïve DME patients were treated with DEX, with visual acuity (VA) ≥ 0.2 logMAR and central subfield thickness (CST) of ≥ 300 µm. Functional and anatomical responses were graded after 2 and 4 months, and categorized as early and stable improvement, early and progressive improvement, pendular response, delayed improvement, and persistent non-response. 417 eyes were included (175 treatment naïve eyes). Compared to non-naïve eyes, naïve eyes showed a very good functional response (VA gain ≥ 10 letters) more frequently after 2 and 4 months (56% and 57% [naïve] vs. 33% and 28% [non-naïve], p < 0.001). A VA gain < 5 letters (non-response) after 2 and 4 months was seen in 18% and 16% of naïve eyes, and in 49% and 53% of non-naïve eyes (p < 0.001). A lack of anatomical response was rare in both groups, but more frequently in non-naïve eyes (12% vs. 4%, p = 0.003). Functionally and anatomically, naïve eyes showed most frequently an early and stable improvement (functionally: 77/175 44%; anatomically: 123/175 eyes, 70%). Most non-naïve eyes experienced no significant improvement functionally (97/242 eyes, 40%), despite a mostly early and stable improvement anatomical response pattern (102/242 eyes, 42%). Functional but not anatomical response patterns were influenced by baseline VA. Naïve and non-naïve eyes show different functional and anatomical response patterns to DEX implant. Functional non-responders are rare in naïve eyes, whereas anatomical non-response is unusual in both groups.
To describe a database of longitudinally graded telemedicine retinal images to be used as a comparator for future studies assessing grader recall bias and ability to detect typical progression (e.g. International Classification of Retinopathy of Prematurity (ICROP) stages) as well as incremental changes in retinopathy of prematurity (ROP). Cohort comprised of retinal images from 84 eyes of 42 patients who were sequentially screened for ROP over 6 consecutive weeks in a telemedicine program and then followed to vascular maturation or treatment, and then disease stabilization. De-identified retinal images across the 6 weekly exams (2520 total images) were graded by an ROP expert based on whether ROP had improved, worsened, or stayed the same compared to the prior week's images, corresponding to an overall clinical "gestalt" score. Subsequently, we examined which parameters might have influenced the examiner's ability to detect longitudinal change; images were graded by the same ROP expert by image view (central, inferior, nasal, superior, temporal) and by retinal components (vascular tortuosity, vascular dilation, stage, hemorrhage, vessel growth), again determining if each particular retinal component or ROP in each image view had improved, worsened, or stayed the same compared to the prior week's images. Agreement between gestalt scores and view, component, and component by view scores was assessed using percent agreement, absolute agreement, and Cohen's weighted kappa statistic to determine if any of the hypothesized image features correlated with the ability to predict ROP disease trajectory in patients. The central view showed substantial agreement with gestalt scores (κ = 0.63), with moderate agreement in the remaining views. Of retinal components, vascular tortuosity showed the most overall agreement with gestalt (κ = 0.42-0.61), with only slight to fair agreement for all other components. This is a well-defined ROP database graded by one expert in a real-world setting in a masked fashion that correlated with the actual (remote in time) exams and known outcomes. This provides a foundation for subsequent study of telemedicine's ability to longitudinally assess ROP disease trajectory, as well as for potential artificial intelligence approaches to retinal image grading, in order to expand patient access to timely, accurate ROP screening.
BACKGROUND: Embolic events leading to retinal ischemia or cerebral ischemia share common risk factors; however, it has been well documented that the rate of concurrent cerebral infarction is higher in patients with a history of transient ischemic attack (TIA) than in those with monocular vision loss (MVL) due to retinal ischemia. Despite the fact that emboli to the ophthalmic artery (OA) and middle cerebral artery share the internal carotid artery (ICA) as a common origin or transit for emboli, the asymmetry in their final destination has not been fully explained. We hypothesize that the anatomic location of the OA takeoff from the ICA may contribute to the differential flow of small emboli to the retinal circulation vs the cerebral circulation. METHODS: We report a retrospective, comparative, case-control study on 28 patients with retinal ischemia and 26 patients with TIA or cerebral infarction caused by embolic events. All subjects underwent either computed tomography angiography or MRA. The location of the ipsilateral OA origin off the ICA was then graded in a blinded fashion and compared between cohorts. Vascular risk factors were collected for all patients, including age, sex, hypertension, hyperlipidemia, arrhythmia, diabetes, coronary artery disease, and smoking. RESULTS: We find that in patients with retinal ischemia of embolic etiology, the ipsilateral OA takeoff from the ICA is more proximal than in patients with cerebral infarcts or TIA (P = 0.0002). We found no statistically significant differences in demographic, vascular, or systemic risk factors. CONCLUSIONS: We find that the mean anatomical location of the OA takeoff from the ICA is significantly more proximal in patients with MVL due to retinal ischemia compared with patients with TIA or cerebral ischemia. This finding contributes significantly to our understanding of a long observed but poorly understood phenomenon that patients with MVL are less likely to have concurrent cerebral ischemia than are patients with TIA.
Importance: Orbital fractures are common in ocular trauma, and there is a need to develop predictive tools to estimate risk of concurrent ocular injury. Objective: To identify clinical and radiographic features that are associated with increased risk of substantial ocular injury in the setting of orbital fracture. Design, Setting, and Participants: Retrospective consecutive case series of patients who sustained orbital fractures between 2012 and 2018. Examinations were done at 1 of 2 level 1 trauma centers in the emergency or inpatient setting. A total of 430 consecutive patients (500 eyes) between 2012 and 2017 met inclusion criteria for the training sample. After building a predictive model, 88 additional consecutive patients (97 eyes) between 2017 and 2018 who met inclusion criteria were collected as a test sample. Main Outcomes and Measures: The primary outcome measure was substantial ocular injury distinct from orbital fracture. Results: The mean age of our patient population was 53.5 years (range, 16-100 years). The overall rate of substantial ocular injury was 20.4%, and the rate of injury requiring immediate ophthalmic attention was 14.4%. Five variables were found to be associated with increased risk of substantial ocular injury: blunt trauma with a foreign object (odds ratio [OR], 19.4; 95% CI, 6.3-64.1; P < .001), inability to count fingers (OR, 10.1; 95% CI, 2.8-41.1; P = .002), roof fracture (OR, 9.1; 95% CI, 2.8-30.0; P = .002), diplopia on primary gaze (OR, 6.7; 95% CI, 1.7-25.1; P = .003), and conjunctival hemorrhage or chemosis (OR, 4.2; 95% CI, 2.2-8.5; P < .001). The results were translated into a bedside tool that was tested in an independent group of eyes (n = 97) and found to be associated with substantial ocular injury with a 95% sensitivity (95% CI, 77.2-99.9), 40% specificity (95% CI, 28.9-52.0), 31.8% positive predictive value (95% CI, 27.5-36.5), and 96.8% negative predictive value (95% CI, 81.3-99.5). Conclusions and Relevance: A minority of patients with an orbital fracture had a substantial ocular injury. Certain radiographic and clinical findings were associated with substantial ocular injury. Testing of the algorithm in prospective longitudinal settings appears warranted.
Understanding the molecular composition of pathogenic tissues is a critical step in understanding the pathophysiology of disease and designing therapeutics. First described in 2009, single cell RNA sequencing (scRNAseq) is a methodology whereby thousands of cells are simultaneously isolated into individual micro-environments that can be altered experimentally and the genome-wide RNA expression of each cell is captured. It has undergone significant technological improvement over the last decade and gained tremendous popularity. scRNAseq is an improvement over prior pooled RNA analyses which cannot identify the cellular composition and heterogeneity of a tissue of interest. This new approach offers new opportunity for new discovery, as tissue samples can now be sub-categorized into groups of cell types based on genome-wide gene expression in an unbiased fashion. As ophthalmologists, we are uniquely positioned to obtain pathologic samples from the eye for further study. ScRNAseq has already been applied in ophthalmology to characterize retinal tissue, and it may offer the key to understanding various pathological processes in the future.
The onset of scleral necrosis after ocular surgery can have catastrophic ocular and systemic consequences. The two most frequent surgeries causing surgically-induced scleral necrosis (SISN) are pterygium excision and cataract extraction. Several pathogenic mechanisms are involved in SISN. All of them are poorly understood. Ocular trauma increasing lytic action of collagenases with subsequent collagen degradation, vascular disruption leading to local ischemia, and immune complex deposition activating the complement system represents some of the events that lead to scleral necrosis. The complex cascade of events involving different pathogenic mechanisms and the patient's abnormal immune response frequently leads to delayed wound healing that predisposes the development of scleral necrosis. The management of SISN ranges from short-term systemic anti-inflammatory drugs to aggressive immunosuppressive therapy and surgical repair. Therefore, before performing any ocular surgery involving the sclera, a thorough ophthalmic and systemic evaluation must be done to identify high-risk patients that may develop SISN.
PURPOSE: To evaluate the surgical success and need for adjustment due to overcorrection in patients who undergo inferior oblique myectomy (IOM) combined with lateral rectus recession (LRc) for intermittent exotropia in the setting of inferior oblique overaction. METHODS: A retrospective chart review was conducted of patients with intermittent exotropia who underwent LRc using adjustable sutures alone versus LRc combined with IOM between January 2010 and July 2018 at our institution. Binocular alignment was recorded before and within one week of surgery. Evaluation measures noted were surgical success (defined as distance alignment of ⩽10 prism diopters) and need for postoperative adjustment due to overcorrection. RESULTS: Of 48 patients, 24 underwent LRc alone and 24 underwent LRc combined with IOM; all 48 patients had adjustable sutures. Surgical success was significantly higher in the LRc alone group (91.6%) compared with the LRc with IOM group (62.5%) ( = 0.036). The need for postoperative adjustment due to overcorrection was also significantly higher in the LRc with IOM group (20.8%) compared with the LRc alone group (0%) ( = 0.049). CONCLUSIONS: In this study, more patients needed adjustment for overcorrection after undergoing LRc combined with IOM versus LRc alone. Since the tertiary action of the inferior oblique is abduction it is possible that, in patients with inferior oblique overaction, surgically weakening the inferior oblique causes more esodeviation and overcorrection. Thus, surgical correction of exotropia and inferior oblique overaction using LRc combined with IOM may lead to overcorrection and increased need for postoperative adjustment.
: This review provides an overview of the causes and treatment of neurotrophic keratopathy in the pediatric population.: A thorough review of the current literature discussing neurotrophic keratopathy was conducted then summarized.Fourty-nine papers were reviewed. Congenital and acquired causes of neurotrophic keratopathy exist in the pediatric population. Both medical and surgical approaches to treatment have been trialed, albeit to a limited extent, in pediatric patients. Conservative treatment includes topical lubrication and antibiotics to prevent concurrent infectious ulcer formation. Various neurotrophic factors have been trialed in the form of serum drops to restore corneal sensation when conservative measures fail. Surgically, different corneal neurotization techniques have been developed whereby a donor nerve is routed to the anesthetized cornea to restore innervation and sensation. : Advances in the treatment of neurotrophic keratopathy have made corneal reinnervation and restoration of vision more easily attainable in pediatric patients.
Memories are encoded in a manner that depends on our knowledge and expectations ("schemas"). Consistent with this, expertise tends to improve memory: Experts have elaborated schemas in their domains of expertise, allowing them to efficiently represent information in this domain (e.g., chess experts have enhanced memory for realistic chess layouts). On the other hand, in most situations, people tend to remember abnormal or surprising items best-those that are also rare or out-of-the-ordinary occurrences (e.g., surprising-but not random-chess board configurations). This occurs, in part, because such images are distinctive relative to other images. In the current work, we ask how these factors interact in a particularly interesting case-the domain of radiology, where experts actively search for abnormalities. Abnormality in mammograms is typically focal but can be perceived in the global "gist" of the image. We ask whether, relative to novices, expert radiologists show improved memory for mammograms. We also test for any additional advantage for abnormal mammograms that can be thought of as unexpected or rare stimuli in screening. We find that experts have enhanced memory for focally abnormal images relative to normal images. However, radiologists showed no memory benefit for images of the breast that were not focally abnormal, but were only abnormal in their gist. Our results speak to the role of schemas and abnormality in expertise; the necessity for spatially localized abnormalities versus abnormalities in the gist in enhancing memory; and the nature of memory and decision-making in radiologists.