Neural classifier construction using regularization, pruning and test error estimation

In this paper we propose a method for construction of feed-forward neural classifiers based on regularization and adaptive architectures. Using a penalized maximum likelihood scheme, we derive a modified form of the entropic error measure and an algebraic estimate of the test error. In conjunction with optimal brain damage pruning, a test error estimate is used to select the network architecture. The scheme is evaluated on four classification problems.     

Coil embolization of major bleeding after femoral fracture: A case report

Successful internal arterial embolization with steel coils in life-threatening bleeding after femoral fracture is described in an older patient.     

Epidural Epinephrine and Clonidine: Segmental Analgesia and Effects on Different Pain Modalities

Background: It is not known whether epidural epinephrine has an analgesic effect per se. The segmental distribution of clonidine epidural analgesia and its effects on temporal summation and different types of noxious stimuli are unknown. The aim of this study was to clarify these issues.

Methods: Fifteen healthy volunteers received epidurally (L2-L3 or L3-L4) 20 ml of either epinephrine, 100 micro gram, in saline; clonidine, 8 micro gram/kg, in saline; or saline, 0.9%, alone, on three different days in a randomized, double-blind, cross-over fashion. Pain rating after electrical stimulation, pinprick, and cold perception were recorded on the dermatomes S1, L4, L1, T9, T6, T1, and forehead. Pressure pain tolerance threshold was recorded at S1, T6, and ear. Pain thresholds to single and repeated (temporal summation) electrical stimulation of the sural nerve were determined.

Results: Epinephrine significantly reduced sensitivity to pinprick at L1-L4-S1. Clonidine significantly decreased pain rating after electrical stimulation at L1-L4 and sensitivity to pinprick and cold at L1-L4-S1, increased pressure pain tolerance threshold at S1, and increased thresholds after single and repeated stimulation of the sural nerve.     

Development of a new dynamic method for quantitative evaluation of in vitro hemocompatibility of biomedical materials.

In this study a new dynamic method is introduced allowing the estimation of blood cell adhesion on flat test surfaces by measuring the cell loss in the bulk phase of surface contacting test blood under defined rheological conditions. This was achieved by constructing a novel test chamber permitting the contact of small amounts of blood with a large geometrical test surface. The construction consists of a spiral-shaped flow channel of 0.3 cm width, 0.02 cm height and 78 cm length covered with the biomaterials to be tested from both sides. Laminarity of blood flow in the conduit was confirmed theoretically by the calculation of an equivalent to the Reynolds number for curved systems the so-called Dean number. Furthermore, flow laminarity was proved experimentally finding that the flow rate of blood with different hematocrit values was proportional to the hydrostatic pressure applied. The applicability of the novel 'spiral method' for the estimation of hemocompatibility was demonstrated by evaluation of platelet adhesion onto different polymers in comparison to siliconized and fibrinogen coated glass as reference surfaces. Additionally, it was possible under distinct conditions to determine the adhesion of leucocytes and the detachment of platelet aggregates. Therefore, it was concluded that the spiral method can be used for the assessment of the hemocompatibility of flat biomedical polymers. As main advantages of the new method can be considered the high time efficiency and accuracy without labelling or optical detection of adherent cells.     

Bromine-76 and carbon-11 labelled NNC 13-8199, metabolically stable benzodiazepine receptor agonists as radioligands for positron emission tomography (PET)

NNC 13-8241 has recently been labelled with iodine-123 and developed as a metabolically stable benzodiazepine receptor ligand for single-photon emission computed tomography (SPECT) in monkeys and man. NNC 13-8199 is a bromo-analogue of NNC 13-8241. This partial agonist binds selectively and with subnanomolar affinity to the benzodiazepine receptors. We prepared ⁷⁶Br labelled NNC 13-8199 from the trimethyltin precursor by the chloramine-T method. Carbon-11 labelled NNC 13-8199 was synthesised by N-alkylation of the nitrogen of the amide group with [¹¹C]methyl iodide. Positron emission tomography (PET) examination with the two radioligands in monkeys demonstrated a high uptake of radioactivity in the occipital, temporal and frontal cortex. In the study with [⁷⁶Br]NNC 13-8199, the monkey brain uptake continued to increase until the time of displacement with flumazenil at 215 min after injection. For both radioligands the radioactivity in the cortical brain regions was markedly reduced after displacement with flumazenil. More than 98% of the radioactivity in monkey plasma represented unchanged radioligand 40 min after injection. The low degree of metabolism indicates that NNC 13-8199 is metabolically much more stable than hitherto developed PET radioligands for imaging of benzodiazepine receptors in the primate brain. [⁷⁶Br]NNC 13-8199 has potential as a radioligand in human PET studies using models where a slow metabolism is an advantage. Received 19 April and in revised form 10 June 1997     

Severe Neurological Complications after Central Neuraxial Blockades in Sweden 1990-1999

Background: Central neuraxial blockades find widespread applications. Severe complications are believed to be extremely rare, but the incidence is probably underestimated.

Methods: A retrospective study of severe neurologic complications after central neuraxial blockades in Sweden 1990-1999 was performed. Information was obtained from a postal survey and administrative files in the health care system. During the study period approximately 1,260,000 spinal blockades and 450,000 epidural blockades were administered, including 200,000 epidural blockades for pain relief in labor.

Results: The 127 complications found included spinal hematoma (33), cauda equina syndrome (32), meningitis (29), epidural abscess (13), and miscellaneous (20). Permanent neurologic damage was observed in 85 patients. Incidence of complications after spinal blockade was within 1:20-30,000 in all patient groups. Incidence after obstetric epidural blockade was 1:25,000; in the remaining patients it was 1:3600 (P < 0.0001). Spinal hematoma after obstetric epidural blockade carried the incidence 1:200,000, significantly lower than the incidence 1:3,600 females subject to knee arthroplasty (P < 0.0001).     

Hematogenous infection after knee arthroplasty

Twenty-five hematogenously infected knee arthroplasties in 20 patients (17 with rheumatoid arthritis and 3 with arthrosis) were followed for 3 years. Staphylococcus aureus was the major infecting organism. Three patients with four arthroplasties died of sepsis. Two patients had removal of the arthroplasty, one of which resulted in an above-the-knee amputation. Four out of five arthrodeses fused. Two knees healed after early debridement and two healed without surgery. Ten knees had successful revision arthroplasty.

Rheumatoid arthritis and constrained prostheses increase the risk of hematogenous infection. Any infection and especially cutaneous lesions in a patient with a knee arthroplasty should be treated vigorously.