We report the case of a 17-year-old man presenting with bilateral visual loss, panuveitis, exudative retinal detachments, and diffuse choroidal infiltration. The patient also complained of headaches and ear pain. Based on the fundus image analysis, fluoangiography, indocyanine green angiography, and optical coherence tomography, Vogt-Koyagani-Harada disease was diagnosed. The patient received high-dose methylprednisolone along with azathioprine. He responded well to treatment with progressive visual improvement and subretinal fluid disappearance. One month later, his right eye exhibited a sectorial retinal vasculitis with infero-temporal vein branch occlusion and massive sub-retinal hemorrhages. Coagulation screening as well as HIV and syphilis serology were all negative; whereas QuantiFERON was positive. An 18F-FDG PET/CT revealed hypermetabolic mediastinal lymph nodes, and a biopsy confirmed active tuberculosis (TB) infection. Anti-TB quadritherapy was initiated, resulting in cytochrome P450 enzyme induction and lowered steroid treatment efficacy, with consecutive recurrence of choroidal inflammation.

Diagnosis is an incredibly dynamic and complex process, which is prone to errors. We all make many diagnostic errors and this, in every healthcare setting, but particularly so in the emergency department. However, these errors are preventable. It is thus essential to better understand how and why these errors occur; then, some simple steps must be taken to avoid their repetition. Many of these errors pertain to the clinical reasoning process and are due to cognitive errors, although other system-related factors are likely present in most cases.

Artificial intelligence (AI) refers to the simulation of human intelligence by machines with integrated computer programs. AI analyzes data and contributes to decision making. The computer performs human-like tasks based on algorithms. Moore's law says that computer performance increases exponentially and doubles every 18 months. On this basis, AI would surpass human intelligence by 2050. Machine learning mimics the human brain. It is fed by big data, which is a mass of heterogeneous data that are processed at a rate far beyond human capacity and in which exploitable data can be gathered. In addition to this virtual branch of AI, there is also the physical branch of robots. It has been shown that physicians surpass current diagnostic programs with respect to medical diagnosis. However, in the case of mammography, while no single AI algorithm outperformed radiologists, overall diagnostic accuracy was improved when using a set of AI algorithms combined with radiologist assessment in a single-reader screening. Only the physicians who have empathy, unlike the machines that lack this skill, may diagnose the disease. Patients’ confidence must come at the top of researchers’ priority list. Nat Med 26, 301 (2020).

This article presents the different pharmacological classes and therapeutic strategies employed at the different stages of Parkinson's disease. To date, only symptomatic treatments exist. Levodopa remains the most effective treatment with the best benefit-risk ratio. It is the initial treatment of choice for most patients. In order to delay the onset of motor complications, such as dyskinesia or end-of-dose akinesia, dopamine agonists or monoamine oxidase B inhibitors may be proposed as first-line treatment to young patients with mild disability. Motor fluctuations can be improved by adjusting the levodopa dosing frequency or by adding a dopamine agonist, monoamine oxidase inhibitor, or catechol-O-methyltransferase inhibitor. Disabling dyskinesias can be improved by reducing the levodopa dose or using amantadine. Patients with severe motor complications may benefit from a treatment with Duodopa® or subcutaneous apomorphine.