Universal adversarial attacks on deep neural networks for medical image classification.

BACKGROUND: Deep neural networks (DNNs) are widely investigated in medical image classification to achieve automated support for clinical diagnosis. It is necessary to evaluate the robustness of medical DNN tasks against adversarial attacks, as high-stake decision-making will be made based on the diagnosis. Several previous studies have considered simple adversarial attacks. However, the vulnerability of DNNs to more realistic and higher risk attacks, such as universal adversarial perturbation (UAP), which is a single perturbation that can induce DNN failure in most classification tasks has not been evaluated yet. METHODS: We focus on three representative DNN-based medical image classification tasks (i.e., skin cancer, referable diabetic retinopathy, and pneumonia classifications) and investigate their vulnerability to the seven model architectures of UAPs. RESULTS: We demonstrate that DNNs are vulnerable to both nontargeted UAPs, which cause a task failure resulting in an input being assigned an incorrect class, and to targeted UAPs, which cause the DNN to classify an input into a specific class. The almost imperceptible UAPs achieved > 80% success rates for nontargeted and targeted attacks. The vulnerability to UAPs depended very little on the model architecture. Moreover, we discovered that adversarial retraining, which is known to be an effective method for adversarial defenses, increased DNNs' robustness against UAPs in only very few cases. CONCLUSION: Unlike previous assumptions, the results indicate that DNN-based clinical diagnosis is easier to deceive because of adversarial attacks. Adversaries can cause failed diagnoses at lower costs (e.g., without consideration of data distribution); moreover, they can affect the diagnosis. The effects of adversarial defenses may not be limited. Our findings emphasize that more careful consideration is required in developing DNNs for medical imaging and their practical applications.

Spectral domain optical coherence tomography classification of diabetic macular edema: a new proposal to clinical practice.

PURPOSE: To classify the types of diabetic macular edema (DME) and evaluate its morphological features on spectral domain optical coherence tomography (SD-OCT) and determine correlations between visual acuity and OCT findings. METHODS: We assessed 406 eyes of 309 patients with a diagnosis of DME retrospectively. Three types based on SD-OCT were identified: diffuse macular edema, cystoid macular edema, and cystoid degeneration. Morphological features such as serous macular detachment (SMD), vitreomacular interface abnormalities (VMAI), hard exudates, photoreceptor status, and correlations between visual acuity and those morphological features were also evaluated by SD-OCT. RESULTS: The most common type of DME was cystoid edema (68.5%). No statistically significant difference was found between groups in sex (P = 0.40), type of diabetes (P = 0.50), or diabetic retinopathy (P = 0.78). However, the duration of symptoms and BCVA was significantly lower in the group with cystoid degeneration compared with the group with cystoid edema (P < 0.001) and the group with diffuse macular edema (P < 0.001). In the group with cystoid degeneration compared with the groups with cystoid and diffuse edema, the central fovea and central subfield were significantly thicker (both (P < 0.001), the subfoveal choroid was significantly thinner (P = 0.049), rate of serous macular detachment was significantly lower (P < 0.001), and the rate of outer retinal damage was significantly higher (P < 0.001). CONCLUSIONS: Cystoid macular degeneration, which is consistent with poor functional and morphological outcomes, should be differentiated from cystoid macular edema. Serous macular detachment, which is mostly seen in eyes with early stages of DME, should be evaluated as an accompanying morphological finding rather than a type of DME.

DISEASE CLASSIFICATION OF MACULAR OPTICAL COHERENCE TOMOGRAPHY SCANS USING DEEP LEARNING SOFTWARE: Validation on Independent, Multicenter Data.

PURPOSE: To evaluate Pegasus optical coherence tomography (OCT), a clinical decision support software for the identification of features of retinal disease from macula OCT scans, across heterogenous populations involving varying patient demographics, device manufacturers, acquisition sites, and operators. METHODS: Five thousand five hundred and eighty-eight normal and anomalous macular OCT volumes (162,721 B-scans), acquired at independent centers in five countries, were processed using the software. Results were evaluated against ground truth provided by the data set owners. RESULTS: Pegasus-OCT performed with areas under the curve of the receiver operating characteristic of at least 98% for all data sets in the detection of general macular anomalies. For scans of sufficient quality, the areas under the curve of the receiver operating characteristic for general age-related macular degeneration and diabetic macular edema detection were found to be at least 99% and 98%, respectively. CONCLUSION: The ability of a clinical decision support system to cater for different populations is key to its adoption. Pegasus-OCT was shown to be able to detect age-related macular degeneration, diabetic macular edema, and general anomalies in OCT volumes acquired across multiple independent sites with high performance. Its use thus offers substantial promise, with the potential to alleviate the burden of growing demand in eye care services caused by retinal disease.

The design and validation of an optical coherence tomography-based classification system for focal vitreomacular traction.

PURPOSE: To develop and validate a classification system for focal vitreomacular traction (VMT) with and without macular hole based on spectral domain optical coherence tomography (SD-OCT), intended to aid in decision-making and prognostication. METHODS: A panel of retinal specialists convened to develop this system. A literature review followed by discussion on a wide range of cases formed the basis for the proposed classification. Key features on OCT were identified and analysed for their utility in clinical practice. A final classification was devised based on two sequential, independent validation exercises to improve interobserver variability. RESULTS: This classification tool pertains to idiopathic focal VMT assessed by a horizontal line scan using SD-OCT. The system uses width (W), interface features (I), foveal shape (S), retinal pigment epithelial changes (P), elevation of vitreous attachment (E), and inner and outer retinal changes (R) to give the acronym WISPERR. Each category is scored hierarchically. Results from the second independent validation exercise indicated a high level of agreement between graders: intraclass correlation ranged from 0.84 to 0.99 for continuous variables and Fleiss' kappa values ranged from 0.76 to 0.95 for categorical variables. CONCLUSIONS: We present an OCT-based classification system for focal VMT that allows anatomical detail to be scrutinised and scored qualitatively and quantitatively using a simple, pragmatic algorithm, which may be of value in clinical practice as well as in future research studies.

Classification of image artefacts in optical coherence tomography angiography of the choroid in macular diseases.

BACKGROUND: To evaluate and classify image artefacts in optical coherence tomography (OCT) angiography (OCTA) of the choroid in a group of patients with macular diseases. DESIGN: Retrospective observational study. PARTICIPANTS: Five patients with age-related macular degeneration, three with central serous retinopathy, one with polypoidal choroidal vasculopathy and one with multiple evanescent white dot syndrome. METHODS: OCTA and OCT reflectivity (OCTR) maps were reviewed along with their fluorescein angiography and indocyanine green angiography. Sixty OCTA images (20 outer retina, 20 Sattler and 20 Haller layers) were graded for image artefacts by two examiners independently. MAIN OUTCOME MEASURES: OCTA artefacts and their correlation with OCTR maps, angiography and OCT B-scans. RESULTS: Artefacts (frequency) were classified into (i) motion (70-100%), (ii) fringe washout (100%), (iii) decorrelation projection (0-20%), (iv) masking and unmasking (50-65%) and (v) stromal decorrelation signal (100%). Motion artefact in OCTA is characterized by horizontal dark lines or bands not apparent on OCTR map. Fringe washout creates signal void within choroidal vessels because of fast blood flow. Decorrelation projection from retinal vasculature and choroidal new vessels above the Bruch's membrane are seen within the choroidal OCTA image. Masking and unmasking artefacts occur in regions of pigment epithelial detachment and atrophy. Decorrelation signals can also be seen in the choroidal stroma. CONCLUSIONS: Our classification system of artefact in choroidal OCTA establishes a common terminology for clinical interpretation. This is important in enhancing our understanding of the principles of OCTA acquisition, and it also serves as a bench mark for reading centres.

RETINAL ASTROCYTIC HAMARTOMA: Optical Coherence Tomography Classification and Correlation With Tuberous Sclerosis Complex.

PURPOSE: To propose a classification of retinal astrocytic hamartoma based on spectral domain optical coherence tomography and correlate each class with systemic manifestations of tuberous sclerosis complex. METHODS: Retrospective chart review conducted at four international referral medical retina centers. There were 43 consecutive patients with an established diagnosis of tuberous sclerosis complex based on presence of at least 2 major or 1 major and 2 minor features of the diagnostic criteria. Clinical and spectral domain optical coherence tomography features regarding retinal astrocytic hamartoma were documented. RESULTS: The mean patient age at presentation was 16.2 years. The retinal astrocytic hamartoma was classified as Type I (n = 41), Type II (n = 25), Type III (n = 20), or Type IV (n = 12). Patients with Type II showed greater number of cutaneous fibrous plaques (odds ratio = 64.8; 92% confidence interval: 64.2-65; P < 0.001); those with Type III displayed higher incidence of subependymal giant-cell astrocytomas (odds ratio = 43.2; 95% confidence interval: 43.0-43.3; P < 0.001); and those with Type IV showed higher incidence of pulmonary lymphangiomyomatosis (odds ratio = 126; 95% confidence interval: 122-128; P < 0.001). CONCLUSION: Retinal astrocytic hamartoma can be classified into four morphologic groups, based on spectral domain optical coherence tomography. There are important systemic tuberous sclerosis complex correlations with each class.

Evolución, recientes avances y futuro desarrollo de la tomografía de coherencia óptica / Evolution, recent progress and future development of optical coherence tomography

No disponible

Learning from healthy and stable eyes: A new approach for detection of glaucomatous progression.

UNLABELLED: Glaucoma is a chronic neurodegenerative disease characterized by loss of retinal ganglion cells, resulting in distinctive changes in the optic nerve head (ONH) and retinal nerve fiber layer. Important advances in technology for non-invasive imaging of the eye have been made providing quantitative tools to measure structural changes in ONH topography, a crucial step in diagnosing and monitoring glaucoma. Three dimensional (3D) spectral domain optical coherence tomography (SD-OCT), an optical imaging technique, is now the standard of care for diagnosing and monitoring progression of numerous eye diseases. METHOD: This paper aims to detect changes in multi-temporal 3D SD-OCT ONH images using a hierarchical fully Bayesian framework and then to differentiate between changes reflecting random variations or true changes due to glaucoma progression. To this end, we propose the use of kernel-based support vector data description (SVDD) classifier. SVDD is a well-known one-class classifier that allows us to map the data into a high-dimensional feature space where a hypersphere encloses most patterns belonging to the target class. RESULTS: The proposed glaucoma progression detection scheme using the whole 3D SD-OCT images detected glaucoma progression in a significant number of cases showing progression by conventional methods (78%), with high specificity in normal and non-progressing eyes (93% and 94% respectively). CONCLUSION: The use of the dependency measurement in the SVDD framework increased the robustness of the proposed change-detection scheme with comparison to the classical support vector machine and SVDD methods. The validation using clinical data of the proposed approach has shown that the use of only healthy and non-progressing eyes to train the algorithm led to a high diagnostic accuracy for detecting glaucoma progression compared to other methods.

Novel optical coherence tomography classification of torpedo maculopathy.

BACKGROUND: Torpedo maculopathy is a rare condition with a twofold clinical significance. Firstly, it is a differential of atypical congenital hypertrophy of the retinal pigment epithelium. Secondly, visual field loss has been reported. We demonstrate the spectrum of structural abnormality of torpedo maculopathy as seen on optical coherence tomography, and correlate this with age of presentation, fundus autofluorescence, retinal sensitivity loss and visual field abnormality. DESIGN: A retrospective, observational case series. PARTICIPANTS: Five Australian patients seen between 2008 and 2013. METHODS: Fundus photography, optical coherence tomography, fundus autofluorescence and visual field analysis. One patient underwent fluorescein angiography. MAIN OUTCOME MEASURES: Lesion appearance on each imaging modality, and visual field analysis. RESULTS: We consistently observed a flat, hypopigmented lesion located in the temporal macula, with a distinctive tip pointing toward the fovea. Optical coherence tomography demonstrated variable retinochoroidal features ranging from mild outer retinal disturbance (type 1) to outer retinal cavitation (type 2). Lesion appearance on short-wave autofluorescence showed varying degrees of hypo-autofluorescence. Near-infrared autofluorescence was performed in two patients and showed a well-defined region of hypo-autofluorescence. Microperimetry showed normal sensitivity over the lesion in one patient and a dense paracentral scotoma over the temporal portion of the lesion in another. On Humphrey field analysis, only one of two patients tested had a paracentral scotoma. CONCLUSION: Two types of torpedo maculopathy lesions are described here with unique optical coherence tomography, demographic, fundus autofluorescence and retinal sensitivity features. These may represent different stages of the same disease that evolve over several decades.

SAVE: a grading protocol for clinically significant diabetic macular oedema based on optical coherence tomography and fluorescein angiography.

AIM: To analyse a new grading protocol for clinically significant diabetic macular oedema (CSME) based on spectral domain optical coherence tomography (SD-OCT) and fluorescein angiography (FA). METHODS: 56 eyes of 40 patients with CSME were examined by Cirrus OCT (Carl Zeiss Meditec), Spectralis HRA and OCT (Heidelberg Engineering) on the same day. Three graders analysed images based on a newly developed grading protocol integrating all relevant information from OCT and FA. The protocol defined four categories: (1) subretinal fluid (category 'S'); (2) the planimetrically measured oedematous area (category 'A'); (3) vitreo-retinal interface abnormalities (category 'V'); and (4) CSME aetiology (category 'E') defining the leakage source. RESULTS: The new grading protocol allowed for a detailed characterisation of each individual type of CSME. It defines four aetiological types of CSME and analyses four further categories important in diagnosis and during follow-up in clinical and study settings. Atrophic, a new type of CSME, was described and characteristic combinations of triggers of CSME were revealed. Inter-grader agreement, analysed using Fleiss' κ values for Cirrus OCT and Spectralis OCT, respectively, was good for 'S' (0.9; 0.82), 'A' (1.0; 1.0) and 'E' (range 0.63-0.8; 0.57-0.77), and lower for 'V' (0.25; 0.42). CONCLUSIONS: The 'SAVE' grading protocol of CSME integrates information from two imaging techniques, OCT and FA. Its clinical approach allows examiners to define and further categorise clinical characteristics to find tailored therapeutic strategies.

Proposed lexicon for anatomic landmarks in normal posterior segment spectral-domain optical coherence tomography: the IN•OCT consensus.

PURPOSE: To develop a consensus nomenclature for the classification of retinal and choroidal layers and bands visible on spectral-domain optical coherence tomography (SD-OCT) images of a normal eye. DESIGN: An international panel with expertise in retinal imaging (International Nomenclature for Optical Coherence Tomography [IN • OCT] Panel) was assembled to define a consensus for OCT imaging terminology. PARTICIPANTS: A panel of retina specialists. METHODS: A set of 3 B-scan images from a normal eye was circulated to the panel before the meeting for independent assignment of nomenclature to anatomic landmarks in the vitreous, retina, and choroid. The outputs were scrutinized, tabulated, and used as the starting point for discussions at a roundtable panel meeting. The history of anatomic landmark designations over time was reviewed for the various cellular layers of the ocular structures that are visible by SD-OCT. A process of open discussion and negotiation was undertaken until a unanimous consensus name was adopted for each feature. MAIN OUTCOME MEASURES: Definitions of normal eye features showed by SD-OCT. RESULTS: Definitions for various layers changed frequently in the literature and were often inconsistent with retinal anatomy and histology. The panel introduced the term "zone" for OCT features that seem to localize to a particular anatomic region that lacks definitely proven evidence for a specific reflective structure. Such zones include the myoid, ellipsoid, and the interdigitation zones. CONCLUSIONS: A nomenclature system for normal anatomic landmarks seen on SD-OCT outputs has been proposed and adopted by the IN • OCT Panel. The panel recommends this standardized nomenclature for use in future publications. The proposed harmonizing of terminology serves as a basis for future OCT research studies.

Comprehensive overview of definitions for optical coherence tomography-based plaque and stent analyses.

Optical coherence tomography (OCT) is the current state-of-the-art intracoronary imaging modality that allows visualization of detailed morphological characteristics of both atherosclerotic plaque and stent. So far, three expert review documents have been released for standardization of OCT image analysis. In the real world, a variety of definitions are being used by different groups and by different core laboratories to analyze OCT findings because of different clinical/procedural contexts in which OCT research has been carried out. This comprehensive overview is aimed to summarize different applicable definitions used by different research groups in plaque and stent analysis using OCT. In addition, it presents readers with a panoramic view to select the best definition of OCT measurement for one's own study purpose. We divided this review article into two parts: Part I - Plaque analysis, and Part II - Stent analysis. The plaque analysis section summarizes the definitions of plaque composition, rupture, erosion, protruding calcific nodules, macrophages, microvessels, and cholesterol crystal. The stent analysis section includes the classification of stent struts, features of neointimal hyperplasia, and other stent-related findings such as tissue protrusion, thrombus, intrastent, and stent edge dissections. In each case of controversy, an explanation for the specific context is provided.

Spectral-domain optical coherence tomography characteristics of intermediate age-related macular degeneration.

PURPOSE: Describe qualitative spectral-domain optical coherence tomography (SD-OCT) characteristics of eyes classified as intermediate age-related macular degeneration (nonadvanced AMD) from Age-Related Eye Disease Study 2 (AREDS2) color fundus photography (CFP) grading. DESIGN: Prospective cross-sectional study. PARTICIPANTS: We included 345 AREDS2 participants from 4 study centers and 122 control participants who lack CFP features of intermediate AMD. METHODS: Both eyes were imaged with SD-OCT and CFP. The SD-OCT macular volume scans were graded for the presence of 5 retinal, 5 subretinal, and 4 drusen characteristics. In all, 314 AREDS2 participants with ≥1 category-3 AMD eye and all controls each had 1 eye entered into SD-OCT analysis, with 63 eyes regraded to test reproducibility. MAIN OUTCOME MEASURES: We assessed SD-OCT characteristics at baseline. RESULTS: In 98% of AMD eyes, SD-OCT grading of all characteristics was successful, detecting drusen in 99.7%, retinal pigment epithelium (RPE) atrophy/absence in 22.9%, subfoveal geographic atrophy in 2.5%, and fluid in or under the retina in 25.5%. Twenty-eight percent of AMD eyes had characteristics of possible advanced AMD on SD-OCT. Two percent of control eyes had drusen on SD-OCT. Vision loss was not correlated with foveal drusen alone, but with foveal drusen that were associated with other foveal pathology and with overlying focal hyperreflectivity. Focal hyperreflectivity over drusen, drusen cores, and hyper- or hyporeflectivity of drusen were also associated with RPE atrophy. CONCLUSIONS: Macular pathologies in AMD can be qualitatively and reproducibly evaluated with SD-OCT, identifying pathologic features that are associated with vision loss, RPE atrophy, and even possibly the presence of advanced AMD not apparent on CFP. Qualitative and detailed SD-OCT analysis can contribute to the anatomic characterization of AMD in clinical studies of vision loss and disease progression. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found after the references.

Classification algorithms based on anterior segment optical coherence tomography measurements for detection of angle closure.

OBJECTIVE: A recent study found that a combination of 6 anterior segment optical coherence tomography (ASOCT) parameters (anterior chamber area, volume, and width [ACA, ACV, ACW], lens vault [LV], iris thickness at 750 µm from the scleral spur, and iris cross-sectional area) explain >80% of the variability in angle width. The aim of this study was to evaluate classification algorithms based on ASOCT measurements for the detection of gonioscopic angle closure. DESIGN: Cross-sectional study. PARTICIPANTS: We included 2047 subjects aged ≥50 years. METHODS: Participants underwent gonioscopy and ASOCT (Carl Zeiss Meditec, Dublin, CA). Customized software (Zhongshan Angle Assessment Program, Guangzhou, China) was used to measure ASOCT parameters in horizontal ASOCT scans. Six classification algorithms were considered (stepwise logistic regression with Akaike information criterion, Random Forest, multivariate adaptive regression splines, support vector machine, naïve Bayes' classification, and recursive partitioning). The ASOCT-derived parameters were incorporated to generate point and interval estimates of the area under the receiver operating characteristic (AUC) curves for these algorithms using 10-fold cross-validation as well as 50:50 training and validation. MAIN OUTCOME MEASURES: We assessed ASOCT measurements and angle closure. RESULTS: Data on 1368 subjects, including 295 (21.6%) subjects with gonioscopic angle closure were available for analysis. The mean (±standard deviation) age was 62.4±7.5 years and 54.8% were females. Angle closure subjects were older and had smaller ACW, ACA, and ACV; greater LV; and thicker irides (P<0.001 for all). For both, the 10-fold cross-validation and the 50:50 training and validation methods, stepwise logistic regression was the best algorithm for detecting eyes with gonioscopic angle closure with testing set AUC of 0.954 (95% confidence interval [CI], 0.942-0.966) and 0.962 (95% CI, 0.948-0.975) respectively, whereas recursive partitioning had relatively the poorest performance with testing set AUC 0.860 (95% CI, 0.790-0.930) and 0.905 (95% CI, 0.876-0.933), respectively. This algorithm performed similarly well (AUC, 0.957) in a second independent sample of 200 angle closure subjects and 302 normal controls. CONCLUSIONS: A classification algorithm based on stepwise logistic regression that used a combination of 6 parameters obtained from a single horizontal ASOCT scan identified subjects with gonioscopic angle closure>95% of the time. FINANCIAL DISCLOSURE(S): The authors have no proprietary or commercial interest in any of the materials discussed in this article.

Classification algorithms enhance the discrimination of glaucoma from normal eyes using high-definition optical coherence tomography.

PURPOSE: To evaluate the diagnostic performance of classification algorithms based on Linear Discriminant Analysis (LDA) and Classification And Regression Tree (CART) methods, compared with optic nerve head (ONH) and retinal nerve fiber layer (RNFL) parameters measured by high-definition optical coherence tomography (Cirrus HD-OCT) for discriminating glaucoma subjects. METHODS: Consecutive glaucoma subjects (Training data = 184; Validation data = 102) were recruited from an eye center and normal subjects (n = 508) from an ongoing Singaporean Chinese population-based study. ONH and RNFL parameters were measured using a 200 × 200 scan protocol. LDA and CART were computed and areas under the receiver operating characteristic curve (AUC) compared. RESULTS: Average RNFL thickness (AUC 0.92, 95% confidence interval [CI] 0.91, 0.93), inferior RNFL thickness (AUC 0.92, 95% CI 0.91, 0.93), vertical cup-disc ratio (AUC 0.91, 95% CI 0.90, 0.92) and rim area/disc area ratio (AUC 0.90, 95% CI 0.86, 0.93) discriminated glaucoma better than other parameters (P ≤ 0.033). LDA (AUC 0.96, 95% CI 0.95, 0.96) and CART (0.98, 95% CI 0.98, 0.99) outperformed all parameters for diagnostic accuracy (P ≤ 0.005). Misclassification rates in LDA (8%) and CART (5.6%) were found to be low. The AUC of LDA for the validation data was 0.98 (0.95, 0.99) and CART was 0.99 (0.99, 0.994). CART discriminated mild glaucoma from normal better than LDA (AUC 0.94 vs. 0.99, P < 0.0001). CONCLUSIONS: Classification algorithms based on LDA and CART can be used in HD-OCT analysis for glaucoma discrimination. The CART method was found to be superior to individual ONH and RNFL parameters for early glaucoma discrimination.

Factors associated with false positives in retinal nerve fiber layer color codes from spectral-domain optical coherence tomography.

PURPOSE: To determine the factors that contribute to false-positive retinal nerve fiber layer (RNFL) color code results from spectral-domain optical coherence tomography (OCT). DESIGN: A prospective, cross-sectional study. PARTICIPANTS: This study included 149 eyes from 77 healthy participants. METHODS: Participants, who were consecutively enrolled from June 2009 to December 2009, underwent Cirrus OCT. Recorded demographic and clinical factors included age, gender, eye side, intraocular pressure, central corneal thickness, spherical equivalent, axial length, anterior chamber depth, disc area, and the extent of retinal vasculature. MAIN OUTCOME MEASURES: An abnormal finding in RNFL color codes was defined as ≥1 yellow or red sectors by quadrant and clock-hour maps and a wedge-shaped color pattern represented by yellow or red in the deviation map. The incidence of false-positive color codes was determined. The influence of clinical and demographic factors on the incidence of false-positive RNFL color codes was assessed using generalized linear mixed model analysis. RESULTS: The false-positive rate for ≥1 of the quadrant, clock-hour, and deviation maps was 26.2%. Longer axial length and smaller disc area were significantly associated with an increased incidence of false-positives when other factors were controlled (odds ratios, 2.422 and 0.165; P = 0.008 and 0.035, respectively). CONCLUSIONS: The factors that significantly affected the false-positive RNFL color code results using spectral-domain OCT were axial length and disc area, which may significantly affect the specificity of spectral-domain OCT. Therefore, axial length and disc area should be considered during RNFL thickness profile analysis.