• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.11a
• /
• pp.608-610
• /
• 2023

의료 분야에서 AI 모델의 활용이 증가하고 있지만, 모델의 예측 불확실성을 정확하게 평가하고 표현하는 것이 중요하다. 본 연구는 이러한 문제를 해결하기 위해 AI-driven 방식을 제안하며, 특히 의료 영상 변환 모델에 대한 불확실성 표현과 데이터 한계 극복 방법론을 제안한다. 제안된 AI-driven 안저영상 변환 모델은 기존 GAN과는 다르게 구조가 이루어져 있으며, 신뢰도가 낮은 영역을 구분하고 시각화하여 표현할 수 있다. 실험 결과, 제안된 방법은 기존 모델과 비교하여 영상 변환 성능이 크게 향상되었으며, 불확실성에 대한 정확도 평가에서도 AI-driven 방식이 높은 성능을 보인다. 결론적으로, 본 연구는 AI-driven 방식을 통해 의료 AI에서의 불확실성 표현의 가능성을 확인하였으며, 이 방식이 데이터의 한계와 불확실성을 극복할 수 있을 것으로 기대된다.

• Journal of The Korean Ophthalmological Society
• /
• v.58 no.11
• /
• pp.1313-1316
• /
• 2017

Purpose: To report fundus photography using a smartphone in an infant with abusive head trauma. Case summary: An 8-month-old male infant presented to the emergency room with decreased consciousness and epileptic seizures that the parents attributed to a fall from a chair. He had no external wounds or fractures to the skull or elsewhere. However, computerized tomography of the brain revealed an acute subdural hematoma in the right cranial convexity and diffuse cerebral edema, leading to a midline shift to the left and effacement of the right lateral ventricle and basal cistern. The attending neurosurgeon promptly administered a decompressive craniectomy. Immediately after the emergency surgery, a fundus examination revealed numerous multi-layered retinal hemorrhages in the posterior pole extending to the periphery in each eye. He also had white retinal ridges with cherry hemorrhages in both eyes. We acquired retinal photographs using the native camera of a smartphone in video mode. The photographer held the smartphone with one hand, facing the patient's eye at 15-20 cm, and held a 20 diopter condensing lens at 5 cm from the eye in the other hand. Our documentation using a smartphone led to a diagnosis of abusive head trauma and to obtain the criminal's confession, because the findings were specific for repetitive acceleration-deceleration forces to an infant's eye with a strong vitreoretinal attachment. Conclusions: This ophthalmic finding had a key role in the diagnosis of abusive head trauma. This case presented the diagnostic use of a smartphone for fundus photography in this important medicolegal case.

• Journal of Biomedical Engineering Research
• /
• v.16 no.2
• /
• pp.121-128
• /
• 1995

Information about retinal blood vessels can be used in grading disease severity or as part of the process of automated diagnosis of diseases with ocular menifestations. In this paper, we address the problem of detecting retinal blood vessels and optic disk (papilla) in eye-fundus images. We introduce an algorithm for feature extraction based on Fuzzy Clustering algorithm (fuzzy c-means). A method of finding the optic disk (papilla) is proposed in the eye-fundus images. Additionally, the inrormations such as position and area of the optic disk are extracted. The results are compared to those obtained from other methods. The automatic detection of retinal blood vessels and optic disk in the eye-rundus images could help physicians in diagnosing ocular diseases.

• Proceedings of the KOSOMBE Conference
• /
• v.1993 no.11
• /
• pp.27-30
• /
• 1993

Information about retinal blood vessels can be used in grading disease severity or as part of the process of automated diagnosis of diseases with ocular menifestations. In this paper, we address the problem of detecting retinal blood vessels and optic disk (papilla) in Eye-fundus images. We introduce an algorithm for feature extraction based on Fuzzy festering(FCM). The results ore compared to those obtained with other methods. The automatic detection of retinal blood vessels and optic disk in the Eye-fundus images could help physicians in diagnosing ocular diseases.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2022.11a
• /
• pp.556-558
• /
• 2022

Diabetic retinopathy (DR) is one of the leading diseases causing vision loss. Early detection of this disease has a crucial role in protecting patients' eyes. Recent works have achieved impressive result when performing DR detection on fundus images using deep learning. In the deep learning-based approach, data augmentation has significant impact on the result. Recently, many data augmentation policies have been proposed and achieved state-of-the-art performance on different tasks. In this work, we compare effects of three data augmentation policies on DR grading in fundus images.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.11a
• /
• pp.557-560
• /
• 2023

In this paper, we introduce a pre-training method leveraging the capabilities of the Vision Transformer (ViT) for disease diagnosis in conventional Fundus images. Recognizing the need for effective representation learning in medical images, our method combines the Vision Transformer with a Masked Autoencoder to generate meaningful and pertinent image augmentations. During pre-training, the Masked Autoencoder produces an altered version of the original image, which serves as a positive pair. The Vision Transformer then employs contrastive learning techniques with this image pair to refine its weight parameters. Our experiments demonstrate that this dual-model approach harnesses the strengths of both the ViT and the Masked Autoencoder, resulting in robust and clinically relevant feature embeddings. Preliminary results suggest significant improvements in diagnostic accuracy, underscoring the potential of our methodology in enhancing automated disease diagnosis in fundus imaging.

• Journal of The Korean Ophthalmological Society
• /
• v.59 no.11
• /
• pp.1097-1102
• /
• 2018

Purpose: To report a case of toxic optic neuropathy caused by chlorfenapyr ingestion accompanied by central nervous system involvement. Case summary: A 44-year-old female visited our clinic complaining of reduced visual acuity in both eyes for 7 days. She had ingested a mouthful of chlorfenapyr for a suicide attempt 2 weeks prior to the visit. Gastric lavage was performed immediately after ingestion at the other hospital. Her best-corrected visual acuity was finger count 30 cm in the right eye and hand motion in the left eye. Both pupils were dilated by 5.0 mm and the response to light was sluggish in both eyes. A relative afferent pupillary defect was detected in her left eye. Funduscopy revealed optic disc swelling in both eyes. Magnetic resonance imaging of the brain showed a symmetric hyper-intense signal in the white matter tract including the internal capsule, corpus callosum, middle cerebellar peduncle, and brainstem. The patient was diagnosed with toxic optic neuropathy induced by chlorfenapyr ingestion, and underwent high-dose intravenous corticosteroid pulse therapy. Three days later, the best-corrected visual acuity was no light perception in both eyes. Three months later, optic atrophy was observed in both eyes. Optical coherence tomography revealed a reduction in the thicknesses of the retinal nerve fiber layer and ganglion cell and inner plexiform layer in the macular area. Conclusions: Ingestion of even a small amount of chlorfenapyr can cause severe optic nerve damage through the latent period, despite prompt lavage and high-dose steroid treatment.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.12
• /
• pp.29-40
• /
• 2022

In this paper, we studied a system that detects and analyzes the pathological features of diabetic retinopathy using Mask R-CNN and a Random Forest classifier. Those are one of the deep learning techniques and automatically diagnoses diabetic retinopathy. Diabetic retinopathy can be diagnosed through fundus images taken with special equipment. Brightness, color tone, and contrast may vary depending on the device. Research and development of an automatic diagnosis system using artificial intelligence to help ophthalmologists make medical judgments possible. This system detects pathological features such as microvascular perfusion and retinal hemorrhage using the Mask R-CNN technique. It also diagnoses normal and abnormal conditions of the eye by using a Random Forest classifier after pre-processing. In order to improve the detection performance of the Mask R-CNN algorithm, image augmentation was performed and learning procedure was conducted. Dice similarity coefficients and mean accuracy were used as evaluation indicators to measure detection accuracy. The Faster R-CNN method was used as a control group, and the detection performance of the Mask R-CNN method through this study showed an average of 90% accuracy through Dice coefficients. In the case of mean accuracy it showed 91% accuracy. When diabetic retinopathy was diagnosed by learning a Random Forest classifier based on the detected pathological symptoms, the accuracy was 99%.

• Journal of The Korean Ophthalmological Society
• /
• v.59 no.11
• /
• pp.1087-1090
• /
• 2018

Purpose: To report a case of trochlear nerve palsy caused by quadrigeminal cistern lipoma located in the dorsal midbrain. Case summary: A 65-year-old male visited our clinic for intermittent vertical diplopia over 2-year period. Symptoms of diplopia had worsened over the past two weeks. He had no previous medical history except having had diabetes for 1 month. The best-corrected visual acuity was 20/25 in the right eye and 20/20 in the left eye. Pupillary examination was not remarkable. Extraocular examination showed 4 prism diopters (PD) left hypertropia at distant gaze and 4 PD exotropia at near gaze, with adduction elevation of the left eye. The Bielschowsky head tilt test revealed 6 PD left hypertropia on the left gaze and orthotropia on the right tilt. Fundus examination showed excyclotorsion of the right eye and incyclotorsion of the left eye. Brain magnetic resonance imaging revealed quadrigeminal cistern lipoma. Prism glasses were prescribed to alleviate diplopia, and we followed up the lesions without further treatment. Conclusions: Trochlear nerve palsy can be caused by quadrigeminal cistern lipoma; however, it is uncommon for this condition to be caused by a compressive lesion. Prompt neuroimaging can be helpful to rule out the causes of this condition in patients with atypical symptoms.

• Journal of the Korean Child Neurology Society
• /
• v.26 no.4
• /
• pp.284-287
• /
• 2018

Tuberous sclerosis complex (TSC) is an autosomal dominant disorder caused by mutation of one of two genes, TSC1 (encoding hamartin, 9q34) and TSC2 (encoding tuberin, 16p13). It invades the central nervous system and various parts of the body, causing various symptoms. Crohn's disease (CD) is a chronic immune-mediated disease that has not been clearly elucidated. It is thought to be caused by an excessive immune response of the body to bacteria that normally exist in the digestive tract with genetic factors. No cases have been reported in which both of the above-mentioned diseases occurred simultaneously. We report a case of CD in a patient with TSC. A 12-year-old boy was brought to our hospital because of abdominal pain. Skin lesions were observed in the TSC. Fundus examination revealed a hamartoma in the right retina. Brain magnetic resonance imaging revealed a subendothelial giant cell astrocytoma (SEGA). On the basis of these findings, he was diagnosed as having TSC. Blood test results showed increased levels of inflammatory markers. On abdominal ultrasonography, his colon walls were observed to be thickened with increased vascularity of the proximal ascending colon, ileocecal valve, and terminal ileum. Colonoscopy revealed discontinuous ulcerations and inflammations of the ileum, IC valve, and cecum, similar to those found in CD. Everolimus was administered orally for the SEGA but was discontinued frequently owing to the exacerbation of CD. The possibility of CD should be kept in mind in patients with TSC considering to undergo treatment for SEGA.