Yukun Zhou

Abstract Medical artificial intelligence (AI) offers great potential for recognizing signs of health conditions in retinal images and expediting the diagnosis of eye diseases and systemic disorders 1 . However, the development of AI models requires substantial annotation and models are usually task-specific with limited generalizability to different clinical applications 2 . Here, we present RETFound, a foundation model for retinal images that learns generalizable representations from unlabelled retinal images and provides a basis for label-efficient model adaptation in several applications. Specifically, RETFound is trained on 1.6 million unlabelled retinal images by means of self-supervised learning and then adapted to disease detection tasks with explicit labels. We show that adapted RETFound consistently outperforms several comparison models in the diagnosis and prognosis of sight-threatening eye diseases, as well as incident prediction of complex systemic disorders such as heart failure and myocardial infarction with fewer labelled data. RETFound provides a generalizable solution to improve model performance and alleviate the annotation workload of experts to enable broad clinical AI applications from retinal imaging.

Purpose: To externally validate a deep learning pipeline (AutoMorph) for automated analysis of retinal vascular morphology on fundus photographs. AutoMorph has been made publicly available, facilitating widespread research in ophthalmic and systemic diseases. Methods: AutoMorph consists of four functional modules: image preprocessing, image quality grading, anatomical segmentation (including binary vessel, artery/vein, and optic disc/cup segmentation), and vascular morphology feature measurement. Image quality grading and anatomical segmentation use the most recent deep learning techniques. We employ a model ensemble strategy to achieve robust results and analyze the prediction confidence to rectify false gradable cases in image quality grading. We externally validate the performance of each module on several independent publicly available datasets. Results: The EfficientNet-b4 architecture used in the image grading module achieves performance comparable to that of the state of the art for EyePACS-Q, with an F1-score of 0.86. The confidence analysis reduces the number of images incorrectly assessed as gradable by 76%. Binary vessel segmentation achieves an F1-score of 0.73 on AV-WIDE and 0.78 on DR HAGIS. Artery/vein scores are 0.66 on IOSTAR-AV, and disc segmentation achieves 0.94 in IDRID. Vascular morphology features measured from the AutoMorph segmentation map and expert annotation show good to excellent agreement. Conclusions: AutoMorph modules perform well even when external validation data show domain differences from training data (e.g., with different imaging devices). This fully automated pipeline can thus allow detailed, efficient, and comprehensive analysis of retinal vascular morphology on color fundus photographs. Translational Relevance: By making AutoMorph publicly available and open source, we hope to facilitate ophthalmic and systemic disease research, particularly in the emerging field of oculomics.

Colorectal cancer (CRC) is the third most common cancer worldwide and has poor prognosis. To identify the oncofetal proteins involved in CRC carcinogenesis, differentially expressed proteins among fetal colorectal tissues, CRC, and the paired tumor-adjacent normal colorectal tissues were investigated by a two-dimensional gel electrophoresis and MALDI-TOF/TOF-based proteomics approach. 42 protein spots were differentially expressed among these tissues, and 22 proteins were identified by MS analysis. Desmin and zinc finger protein 829 were found to be elevated in CRC tissue and fetal colorectal tissue compared with normal colorectal tissue. The elevated expression of desmin in CRC tissue and different developmental stages of fetus colon was confirmed by RT-PCR and Western blot analysis. Immunohistochemical analysis showed that the elevated expression of desmin was correlated with the severity and differentiation of CRC and decreased survival rate of CRC patients. Finally by developing a highly sensitive immunoassay, desmin could be detected in human serum and was significantly elevated in CRC patients compared with healthy volunteers. We propose that desmin be considered a potential oncofetal serum tumor marker for CRC that may have significance in the detection of patients with CRC.

<h3>Background and objectives:</h3> Cadaveric studies have shown disease-related neurodegeneration and other morphological abnormalities in the retina of individuals with Parkinson disease (PD), however it remains unclear whether this can be reliably detected with in vivo imaging. We investigated inner retinal anatomy, measured using optical coherence tomography (OCT), in prevalent PD and subsequently assessed the association of these markers with the development of PD using a prospective research cohort. <h3>Methods:</h3> This cross-sectional analysis used data from two studies. For the detection of retinal markers in prevalent PD, we used data from AlzEye, a retrospective cohort of 154,830 patients aged 40 years and over attending secondary care ophthalmic hospitals in London, UK between 2008 and 2018. For the evaluation of retinal markers in incident PD, we used data from UK Biobank, a prospective population-based cohort where 67,311 volunteers aged 40-69 years were recruited between 2006 and 2010 and underwent retinal imaging. Macular retinal nerve fibre layer (mRNFL), ganglion cell-inner plexiform layer (GCIPL), and inner nuclear layer (INL) thicknesses were extracted from fovea--centred OCT. Linear mixed effects models were fitted to examine the association between prevalent PD and retinal thicknesses. Hazard ratios for the association between time to PD diagnosis and retinal thicknesses were estimated using frailty models. <h3>Results:</h3> Within the AlzEye cohort, there were 700 individuals with prevalent PD and 105,770 controls (mean age 65.5 ± 13.5 years, 51.7% female). Individuals with prevalent PD had thinner GCIPL (-2.12 μm, 95% confidence interval: -3.17, -1.07, <i>p</i> = 8.2 × 10^-5) and INL (-0.99 μm, 95% confidence interval: -1.52, -0.47, <i>p</i> = 2.1 × 10^-4). The UK Biobank included 50,405 participants (mean age 56.1 ± 8.2 years, 54.7% female), of whom 53 developed PD at a mean of 2653 ± 851 days. Thinner GCIPL (hazard ratio: 0.62 per standard deviation increase, 95% confidence interval: 0.46, 0.84, <i>p</i>=0.002) and thinner INL (hazard ratio: 0.70, 95% confidence interval: 0.51, 0.96, <i>p</i>=0.026) were also associated with incident PD. <h3>Discussion:</h3> Individuals with PD have reduced thickness of the INL and GCIPL of the retina. Involvement of these layers several years before clinical presentation highlight a potential role for retinal imaging for at-risk stratification of PD.

Abstract The purpose of this study was to establish and validate a new deep learning system that generates quantitative computed tomography (CT) reports for the diagnosis of pulmonary tuberculosis (PTB) in clinic. 501 CT imaging datasets were collected from 223 patients with active PTB, while another 501 datasets, which served as negative samples, were collected from a healthy population. All the PTB datasets were labeled and classified manually by professional radiologists. Then, four state-of-the-art 3D convolution neural network (CNN) models were trained and evaluated in the inspection of PTB CT images. The best model was selected to annotate the spatial location of lesions and classify them into miliary, infiltrative, caseous, tuberculoma, and cavitary types. The Noisy-Or Bayesian function was used to generate an overall infection probability of this case. The results showed that the recall and precision rates of detection, from the perspective of a single lesion region of PTB, were 85.9% and 89.2%, respectively. The overall recall and precision rates of detection, from the perspective of one PTB case, were 98.7% and 93.7%, respectively. Moreover, the precision rate of type classification of the PTB lesion was 90.9%. Finally, a quantitative diagnostic report of PTB was generated including infection possibility, locations of the lesion, as well as the types. This new method might serve as an effective reference for decision making by clinical doctors.