An Efficient Deep Neural Network to Classify Large 3D Images With Small Objects.

3D imaging enables accurate diagnosis by providing spatial information about organ anatomy. However, using 3D images to train AI models is computationally challenging because they consist of 10x or 100x more pixels than their 2D counterparts. To be trained with high-resolution 3D images, convolutional neural networks resort to downsampling them or projecting them to 2D. We propose an effective alternative, a neural network that enables efficient classification of full-resolution 3D medical images. Compared to off-the-shelf convolutional neural networks, our network, 3D Globally-Aware Multiple Instance Classifier (3D-GMIC), uses 77.98%-90.05% less GPU memory and 91.23%-96.02% less computation. While it is trained only with image-level labels, without segmentation labels, it explains its predictions by providing pixel-level saliency maps. On a dataset collected at NYU Langone Health, including 85,526 patients with full-field 2D mammography (FFDM), synthetic 2D mammography, and 3D mammography, 3D-GMIC achieves an AUC of 0.831 (95% CI: 0.769-0.887) in classifying breasts with malignant findings using 3D mammography. This is comparable to the performance of GMIC on FFDM (0.816, 95% CI: 0.737-0.878) and synthetic 2D (0.826, 95% CI: 0.754-0.884), which demonstrates that 3D-GMIC successfully classified large 3D images despite focusing computation on a smaller percentage of its input compared to GMIC. Therefore, 3D-GMIC identifies and utilizes extremely small regions of interest from 3D images consisting of hundreds of millions of pixels, dramatically reducing associated computational challenges. 3D-GMIC generalizes well to BCS-DBT, an external dataset from Duke University Hospital, achieving an AUC of 0.848 (95% CI: 0.798-0.896).

Thyroid Cancer: Ultrasound Imaging and Fine-Needle Aspiration Biopsy.

Ultrasound is critical in detection, diagnosis, and management of thyroid nodules. Ultrasound detection of regional nodal metastatic disease is based on abnormal nodal morphology rather than size and is critical to initial surgical and long-term management of thyroid cancer. Fine-needle aspiration biopsy is the gold standard for malignancy diagnosis in thyroid cancer. Thyroglobulin assay of nodal aspirates improves accuracy in diagnosis of metastases. Reporting lexicons assign risk levels to thyroid nodules with the goal of improving and standardizing patient management. Surveillance ultrasound in papillary microcarcinomas is being evaluated and compared with surgical management.

Prevalence of Age-Related Macular Degeneration in Chinese American Adults: The Chinese American Eye Study.

IMPORTANCE: Population-based prevalence estimates of age-related macular degeneration (AMD) need to be determined to assess its burden among Chinese Americans, the fastest growing racial group in the United States. OBJECTIVE: To determine the age- and sex- specific prevalence of AMD among Chinese Americans. DESIGN: The Chinese American Eye Study (CHES) was conducted in a general urban community of 10 census tracts in Monterey Park, California. A total of 4582 Chinese American adults aged 50 years or older participated in this population-based, cross-sectional study from February 16, 2010, through October 9, 2013, and underwent an interview as well as comprehensive clinical and eye examinations, including detailed retinal photography of both eyes. Fundus photographs were graded for drusen and retinal pigment epithelium abnormalities and were evaluated for AMD. MAIN OUTCOMES AND MEASURES: The prevalence of early and advanced AMD, drusen, geographic atrophy, and neovascular AMD were determined by using a modified Wisconsin Age-Related Maculopathy Grading Scale (a 6-level scale: 10, no AMD; 60, advanced AMD). RESULTS: Of the 4582 participants completing both the home survey and clinical examination, 4172 individuals (91.1%) had at least 1 gradable photograph. A total of 1526 (36.6%) participants were men, and the mean (SD) age was 61.2 (8.8) years. When examined by 10-year age groups, the prevalence of early AMD ranged from 5.8% (n = 119) in participants aged 50 to 59 years to 17.6% (n = 37) in those 80 years or older, retinal pigment epithelium abnormalities from 4.1% (n = 85) to 7.2% (n = 16), large drusen (≥125 µm) from 9.8% to 32.4%, soft drusen from 27.6% (n = 567) to 58.6% (n = 123), and soft indistinct drusen from 3.7% (n = 76) to 15.2% (n = 32). The prevalence of advanced AMD ranged from 0.2% (n = 3) in participants aged 50 to 59 years to 1.0% (n = 2) in those 80 years or older. Of the 14 cases of advanced AMD, 85.7% (95% CI, 57.2%-98.2%; n = 12) were neovascular AMD and 14.3% (95% CI, 2.0%-42.8%; n = 2) were geographic atrophy. Acute macular degeneration was more common in men (10.9% [9.3%-12.5%]; n = 166) than women (5.8% [4.9%-6.7%]; n = 154) in this cohort. CONCLUSIONS AND RELEVANCE: Data from CHES suggest that Chinese Americans have a lower prevalence of early and advanced AMD compared with non-Hispanic white individuals. The prevalence of early AMD, advanced AMD, and large drusen was higher among Chinese Americans in CHES than among the Chinese population living in urban/rural China but lower than that in urban-dwelling Taiwanese.