En löytänyt tätä, mutta löysin tältä vuodelta toisen mielenkiintoisen artikkelin, jossa käytetty Optomedin Aurora IQ:ta ja tekoälyä tunnistamaan aivoinfarkti- ja TIA-potilaat terveistä. Ei ole varmaan vielä linkattu tänne. Näkisin isona plussana, jos perähikiällä tk:ssa olisi kamera auttamassa diagnostiikassa ja pitäisi konsultoida neurologia keskussairaalasta.
Kohtalaisen pieni aineisto (N=220), mutta jos tulkkaan oikein, niin kuitenkin jo ilmeisen asiallisia tuloksia näin pienellä otannalla. Tässä hyödynnetty ensimmäistä kertaa silmänpohjakuvia kahdesta kuvakulmasta (makulasta ja näköhermon päästä) ja kummastakin silmästä, mikä voi parantaa sensitiivisyyttä ja spesifisyyttä. MVS-Net käyttää siis neljää kuvaa per potilas tehdäkseen johtopäätöksensä:
https://arxiv.org/pdf/2502.00079
Muutaman kohdan siteerasin:
“Therefore, in this study, we propose a multiview stroke network (MVS-Net) to detect stroke and transient ischemic attack (TIA) using retinal fundus images. Contrary to existing studies, our study proposes for the first time a solution to discriminate stroke and TIA with deep multi-view learning by proposing an end-to-end deep network, consisting of multiview inputs of fundus images captured from both right and left eyes. Accordingly, the proposed MVS-Net defines representative features from fundus images of both eyes and determines the relation within their macula-centered and optic nerve headcentered views. Experiments performed on a dataset collected from stroke and TIA patients, in addition to healthy controls, show that the proposed framework achieves an AUC score of 0.84 for stroke and TIA detection.”
( AUC (area under the curve) value (0–1):
- 1.0 = Perfect (100% correct every time)
- 0.9 = Excellent
- 0.8 = Good
- 0.5 = Useless (random guessing))
“Research study nurses acquired retinal fundus images using the Optomed Aurora IQ fundus camera at the Stroke Units of hospitals.”
"Accordingly, retinal fundus images from both eyes of the participants were captured. From each eye, dual views were obtained, which refer to macula-centric and optic nerve head-centric images that visualize the central visual field and papilla in the middle, respectively. "
“Hence, results reveal that detection from single-view retinal fundus images proposed by many studies in the literature is not a suitable approach for stroke detection. Therefore, this study for the first time shows that using multi-view retinal fundus images for stroke detection outperforms the single-view approach.”
“Stroke diagnosis seeks new diagnostics that can be used in case of emergency to identify stroke symptoms and transfer patients to a hospital quickly. In this study, we propose a method to standardize stroke diagnosis with retinal fundus images that can be used even in remote areas, where urgent access to a hospital and neurologist is restricted. The proposed method can be used in a fundus camera, which is non-invasive and cheap, accelerating the first-aid detection of stroke with minimal subjectivity.”
vilpukkaulpukka