ObjectiveTo observe multimodal imaging characteristics in eyes with focal choroidal excavation (FCE) and preliminarily analyze the risk factors in FCE with complications correlated with RPE.MethodsA retrospective case series. Thirty-one patients (31 eyes) with monocular FCE, first identified by spectral-domain (SD)-OCT in the Eye Center of The Second People’s Hospital of Foshan from December 2014 to December 2018, were involved in this study. There were 14 males and 17 females, with the mean age of 45.84±13.57 years. All patients underwent BCVA, optometry, and SD-OCT examinations. FFA and ICGA were simultaneously performed in 3 FCE patients with RPE complications. The subfoveal choroidal thickness (SFCT) and excavation width were measured with enhanced depth imaging OCT (EDI-OCT). The eyes with FCE were divided into two groups (FCE alone group 17 eyes vs. FCE complication group 14 eyes), based on whether complicated by RPE dysfunction. Among 14 eyes of FCE complication group, 7 (22.6%) with choroidal neovascularization, 4 (12.9%) with central serous chorioretinopathy, 1 (3.2%) with polypoidal choroidal vasculopathy, and 2 (6.5%) with RPE detachment. No significant difference was found in the mean age (t=0.87), gender composition (χ2=0.06), ocular laterality (χ2=2.58), and spherical equivalent (t=?0.81) between two groups, respectively (P＞0.05), except that the BCVA was significantly different (t=?2.11, P＜0.05). The SFCT and excavation width of eyes in both groups and the ICGA imaging characteristics of eyes in FCE complication group were analyzed. Risk factors of FCE with RPE complications were analyzed by logistic regression analysis.ResultsThirty-three excavations were identified in 31 eyes with FCE. The mean SFCT was 167.00±85.18 μm in FCE alone group vs. 228.36±67.95 μm in FCE complication group, while the excavation width was 645.00±231.93 μm vs. 901.00±420.55 μm and they were both significantly different (P＜0.05). Logistic regression analysis showed the SFCT (OR=1.016, P=0.026) and excavation width (OR=1.004, P=0.034) were risk factors for RPE complications of FCE. EDI-OCT showed the RPE at the excavation was impaired or vulnerable in all eyes of the FCE alone group, especially at the boundary area of excavation. The RPE damages were located at the boundary area of excavation in 10 eyes (71.4%) of FCE complication group. Constant choroidal hypofluorescence and filling defect were observed under the excavation in 3 eyes with ICGA imaging.ConclusionsSFCT and excavation width may be risk factors for RPE complications of FCE. Impairment of RPE at boundary area of excavation and focal choroidal ischemia or aberrant circulation under the excavation may correlate with the development of FCE complications.
With the rapid development of ophthalmic imaging methods, there are many ways of examination in the diagnosis and treatment of fundus diseases, such as FFA, ICGA, FAF, OCT and emerging blood vessels by OCT angiography in recent years. Multi-model image can understand the changes of anatomical structure and function of different levels and parts of the fundus from different aspects. A variety of imaging examinations are combined and complemented each other, which makes us have a further understanding of the location and pathological changes of many fundus diseases. But at the same time, the emergence of multi-modal images also brings a series of problems. How to standardize the use of multi-modal imaging platform to better serve the clinic is a problem that ophthalmologists need to understand.
The task of automatic generation of medical image reports faces various challenges, such as diverse types of diseases and a lack of professionalism and fluency in report descriptions. To address these issues, this paper proposes a multimodal medical imaging report based on memory drive method (mMIRmd). Firstly, a hierarchical vision transformer using shifted windows (Swin-Transformer) is utilized to extract multi-perspective visual features of patient medical images, and semantic features of textual medical history information are extracted using bidirectional encoder representations from transformers (BERT). Subsequently, the visual and semantic features are integrated to enhance the model's ability to recognize different disease types. Furthermore, a medical text pre-trained word vector dictionary is employed to encode labels of visual features, thereby enhancing the professionalism of the generated reports. Finally, a memory driven module is introduced in the decoder, addressing long-distance dependencies in medical image data. This study is validated on the chest X-ray dataset collected at Indiana University (IU X-Ray) and the medical information mart for intensive care chest x-ray (MIMIC-CXR) released by the Massachusetts Institute of Technology and Massachusetts General Hospital. Experimental results indicate that the proposed method can better focus on the affected areas, improve the accuracy and fluency of report generation, and assist radiologists in quickly completing medical image report writing.
ObjectiveTo observe the multimodal imaging characteristics of multiple evanescent white dot syndrom (MEWDS).MethodsThis was a retrospective series case study. Eighteen patients (18 eyes) diagnosed with MEWDS in Eye Center of The Second People’s Hospital of Foshan from September 2015 to April 2017 were enrolled in this study. There were 12 females and 6 males, with the mean age of 35.9 years. The disease course ranged from 3 to 90 days, with the mean of 14 days. All the patients underwent BCVA, slit-lamp microscope with +90D preset lens, fundus photography, spectral domain OCT (SD-OCT) and FAF examinations. FFA was simultaneously performed in 6 eyes, FFA and ICGA were simultaneously performed in 12 eyes. Ten patients received the treatment of glucocorticoids and vasodilator substance, and other 8 patients without any treatment. The follow-up duration was 4.5 months. The multimodal imaging characteristics were reviewed and analyzed.ResultsFundus color photography showed a variable number of small dots and large spots lesions (14 eyes), and/or fovea granularity (7 eyes) and disk swelling (5 eyes). A variable number of little dots and larger spots lesions showed respectively in FFA, FAF and ICGA were needle-like dots distributed in a wreathlike pattern and a large plaque occasionally confluent of early highly fluorescent, highly autofluorescence and hypofluoresence. Combined hypofluorescent spots with overlying dots were observed in 10 eyes of the late stages of the ICGA. Black lesions in the gray background show in ICGA were the most obvious and the most extensive, gray-white lesions in the gray-black show in FAF were the second, light gray-black lesions in the gray show in FFA were the least. Gray-white lesions in an orange background show in fundus photography were not obvious and transient. SD-OCT showed disruption of the ellipsoid zone and/or accumulations of hyperreflective material from the ellipsoid layer toward the outer plexiform layer and vitreous cells. During the period of following-up, some patients were prescribed low-dose glucocorticoid and some not, almost all the patients except one patient experienced recovery in BCVA and the lesions in fundus imaging.ConclusionsThe lesions in MEWDS eyes in modern multimodal imaging modalities among fundus photography (fovea granularity), FFA (needle-like dots distributed in a wreathlike pattern and a large plaque occasionally confluent of early highly fluorescent), ICGA (flake hypofluorescent) and SD-OCT (disruption of the ellipsoid zone) showed good consistency. Almost eyes were recovery.
Stellate multiform amelanotic choroidopathy (SMACH) is a rare choroidal disease that can cause persistent subretinal fluid (SRF). It is more common in young individuals, with a similar male-to-female ratio, it is most often unilateral, and its clinical manifestations are diverse. The pathogenesis of SMACH is not well understood, but it may be a form of congenital choroidal dysplasia. The progressive impact of the lesion on the choroidal capillaries and the retinal pigment epithelium may be the cause of SRF. Its characteristic multimodal imaging changes include optical coherence tomography showing hyperreflective fibrous-like changes located in the inner choroidal stroma. Typical finger-like projections arranged in a stellate configuration are best seen on near-infrared imaging, indocyanine green angiography, and en face optical coherence tomography. The lesion is stable, with no progressive changes, and is unresponsive to treatment. Clinicians have limited knowledge about SMACH, which can lead to patients undergoing unnecessary or inappropriate treatments. Therefore, it is necessary to understand and recognize SMACH early in clinical practice.
Objective To verify the feasibility and effectiveness of the modified fast-track surgery (FTS) in the perioperative period of open liver resection. Methods A prospective randomized controlled trial was carried out in 188 consecutive patients undergoing open liver resection between March and December 2014 in the Department of Liver Surgery of West China Hospital. The modified fast recovery procedure and standard rehabilitation procedure were compared in terms of length of hospital stay after operation, hospitalization cost, complications and readmission rate. Results A total of 188 consecutive patients were enrolled in the trial. The analysis included 87 patients in the modified fast recovery group and 89 in the standard rehabilitation group. Compared with the standard rehabilitation group, the modified fast recovery group had a shorter length of hospital stay [(5.70±1.47)vs. (7.26±1.96) days] and a lower cost [(42.7±6.7)vs. (47.3±12.5) thousand yuan], and the differences were statistically significant (P<0.05). There were 20 complication cases in the modified group and 39 in the standard group with significant difference (P=0.003). There was no significant difference in the rate of readmission between the two groups (P=1.000). Compared with the standard group, patients in the modified group had less pain 8 hours, the 1st and 2nd days after surgery, better postoperative activities of daily living, more initiative cough times and off-bed activity times, longer duration of movement, and earlier bowel recovery and exhausting, and all the above differences were significantly different (P<0.05). Stepwise regression analysis showed that postoperative complications and bowel recovery and exhausting time were independent related factors for postoperative hospital stay (P<0.001). Conclusions Multimodal analgesia-based fast recovery procedure is feasible and effective in the perioperative period of partial hepatectomy. It can shorten the time of hospitalization and reduce the cost of hospitalization.
Cardiovascular diseases is the leading cause of threat to human life and health worldwide. Early risk assessment, timely diagnosis, and prognosis evaluation are critical to the treatment of cardiovascular diseases. Currently, the evaluation of diagnosis and prognosis of cardiovascular diseases mainly relies on imaging examinations such as coronary CT and coronary angiography, which are expensive, time-consuming, partly invasive, and require high professional competence of the operator, making it difficult to promote in the community or in areas where medical resources are scarce. The fundus microcirculation is a part of the human microcirculation and has similar embryological origins and physiopathological features to cardiovascular circulation. Several studies have revealed fundus imaging biomarkers associated with cardiovascular diseases, and developed and validated intelligent diagnosis and treatment models for cardiovascular diseases based on fundus imaging data. Fundus imaging is expected to be an important adjunct to cardiovascular disease diagnosis and treatment given its noninvasive and convenient nature. The purpose of this review is to summarize the current research status, challenges, and future prospects of the application of artificial intelligence based on multimodal fundus imaging data in cardiovascular disease diagnosis and treatment.
Objective To observe the multimodal imaging features and explore the treatment of parafoveal exudative vascular anomaly complex (PEVAC). Methods A retrospective study. Six patients (6 eyes) with PEVAC diagnosed in Tianjin Eye Hospital were included in this study from July 2018 to December 2021. All patients were female with monocular disease. The age was (61.1±9.3) years. All patients showed a sudden painless decline in monocular vision with metamorphopsia. All patients underwent best corrected visual acuity (BCVA), color fundus photography, fundus fluorescein angiography (FFA), optical coherence tomography (OCT) and OCT angiography (OCTA). Indocyanine green angiography (ICGA) was performed in 4 eyes. In 6 eyes, 3 eyes were treated with intravitreal injection of anti-vascular endothelial growth factor drug; 5 eyes were treated with micropulse laser photocoagulation and/or local thermal laser photocoagulation; 1 eye was treated with photodynamic therapy. Five patients were followed up for (9.2±7.4) months, and 1 patient was lost. At follow-up, the same equipment and methods were used as at the initial diagnosis. The clinical manifestations, multimodal image features and treatment response were observed. Results Baseline BCVA of affected eyes were ranged from 0.1 to 0.5. PEVAC was isolated in 6 eyes, and the fundus showed isolated hemangioma-like leision, accompanied by small bleeding and hard exudation. There were 2 isolated hemangiomatous lesions adjacent to each other in 2 eyes. In the early stage of FFA, punctate high fluorescence lesions near the macular fovea were seen, and the leakage was enhanced in the late stage. There was no leakage in the early stage of ICGA, or slight leakage with late scouring. OCT showed an oval lesion with high reflection wall and uneven low reflection. The central macular thickness (CMT) was (431±76) μm. OCTA showed blood flow signals in PEVAC, 2 eyes in the superficial capillary plexus (SCP), and it was also observed in the deep capillary plexus (DCP), but the intensity of blood flow signal was slightly weaker than that in the SCP. The blood flow signal was visible only in DCP in 2 eyes. SCP and DCP showed similar intensity of blood flow signals in 2 eyes. After treatment, the bleeding was absorbed basically in 4 eyes, the hard exudation partially subsided, the CMT decreased, the intercortical cystic cavity of the fovea nerve decreased, the hemangiomatous lesions narrowed, and BCVA increased. In 1 eye, the macular sac was reduced and partially absorbed by hard exudation, which was later relapsed due to blood pressure fluctuation.ConclusionsThe majority of PEVAC patients had monocular onset. The fundus is characterized by solitary or structure with strong reflex walls, with or without retinal cysts, hard exudates, and subretinal fluid, and visible blood flow signals inside.
ObjectiveTo observe the multimodal imaging characteristics of choroidal metastasis.MethodsA retrospective clinical observation study. From January 2016 to November 2018, 28 patients with choroidal metastasis diagnosed in Department of Ophthalmology in the Second People’s Hospital of Yunnan Province were included in the study. There were 12 males and 16 females, with the mean age of 50.8±6.9 years. There were 18 unilateral patients and 10 bilateral patients. The lesion of choroidal metastasis was regressed after systemic antitumor therapy in 3 patients (4 eyes). All patients underwent ultra-wide-angle fundus photography, infrared fundus imaging, fundus autofluorescence, FFA, frequency-domain OCT, and B-ultrasound examinations.ResultsIn the ultra-wide-angle fundus photography, metastatic tumors were located in the posterior or middle part of the retina, of which 26 were isolated lesions and 12 were multifocal. A yellow-white bulge lesion with (11 eyes) or without pigmentation (27 eyes). There were 12 eyes with exudative retinal detachment. Infrared photography of the fundus showed that the tumor area showed varying degrees of mottled brightness change, and the infrared photograph of the exudative retinal detachment area was relatively low. Fundus autofluorescence showed that 14 eyes had plaque-like strong autofluorescence in the tumor, 13 eyes had a mottled autofluorescence formed by strong and weak fluorescence in the tumor; 3 eyes of old lesions showed " leopard-like” autofluorescence. Among the 38 eyes in the fluorescein angiography, 32 eyes of the early lesions showed low fluorescence, and the venous phase showed a needle-like high fluorescence point, and the post-leakage fluorescence gradually increased. Two eyes with old lesions showed a " leopard-like” change. In 38 eyes, OCT showed wavy ridges of the choroid and pigment epithelium, and a large number of fine-grained or cluster-like high-reflector accumulations were observed between the retinal neuroepithelial layer and the pigment epithelial layer. B-ultrasound showed substantial lesions in the posterior pole and uniform internal echo. There were 23 eyes with flat shape, 12 eyes with flat hemisphere, and 3 eyes with irregular shape.ConclusionsColor photography of the fundus showed the size, location, pigmentation and peripheral retinopathy of the metastatic lesions. Infrared photography showed different reflex signals in the tumor, exudation, and atrophy. The autofluorescence of the fundus showed the damage of pigment epithelium in the lesion. In the fluorescein angiography, the fresh tumor showed fluorescence leakage, while the atrophic tumor showed transmitted fluorescenc. OCT reflected the height of the lesion and the change of pigment epithelium.
Recently, deep learning has achieved impressive results in medical image tasks. However, this method usually requires large-scale annotated data, and medical images are expensive to annotate, so it is a challenge to learn efficiently from the limited annotated data. Currently, the two commonly used methods are transfer learning and self-supervised learning. However, these two methods have been little studied in multimodal medical images, so this study proposes a contrastive learning method for multimodal medical images. The method takes images of different modalities of the same patient as positive samples, which effectively increases the number of positive samples in the training process and helps the model to fully learn the similarities and differences of lesions on images of different modalities, thus improving the model's understanding of medical images and diagnostic accuracy. The commonly used data augmentation methods are not suitable for multimodal images, so this paper proposes a domain adaptive denormalization method to transform the source domain images with the help of statistical information of the target domain. In this study, the method is validated with two different multimodal medical image classification tasks: in the microvascular infiltration recognition task, the method achieves an accuracy of (74.79 ± 0.74)% and an F1 score of (78.37 ± 1.94)%, which are improved as compared with other conventional learning methods; for the brain tumor pathology grading task, the method also achieves significant improvements. The results show that the method achieves good results on multimodal medical images and can provide a reference solution for pre-training multimodal medical images.