Classification of rib fracture types from postmortem computed tomography images using deep learning.

Human or time resources can sometimes fall short in medical image diagnostics, and analyzing images in full detail can be a challenging task. With recent advances in artificial intelligence, an increasing number of systems have been developed to assist clinicians in their work. In this study, the objective was to train a model that can distinguish between various fracture types on different levels of hierarchical taxonomy and detect them on 2D-image representations of volumetric postmortem computed tomography (PMCT) data. We used a deep learning model based on the ResNet50 architecture that was pretrained on ImageNet data, and we used transfer learning to fine-tune it to our specific task. We trained our model to distinguish between "displaced," "nondisplaced," "ad latus," "ad longitudinem cum contractione," and "ad longitudinem cum distractione" fractures. Radiographs with no fractures were correctly predicted in 95-99% of cases. Nondisplaced fractures were correctly predicted in 80-86% of cases. Displaced fractures of the "ad latus" type were correctly predicted in 17-18% of cases. The other two displaced types of fractures, "ad longitudinem cum contractione" and "ad longitudinem cum distractione," were correctly predicted in 70-75% and 64-75% of cases, respectively. The model achieved the best performance when the level of hierarchical taxonomy was high, while it had more difficulties when the level of hierarchical taxonomy was lower. Overall, deep learning techniques constitute a reliable solution for forensic pathologists and medical practitioners seeking to reduce workload.

Anterior cervical discectomy and fusion: is surgical education safe?

BACKGROUND: Operative skills are key to neurosurgical resident training. They should be acquired in a structured manner and preferably starting early in residency. The aim of this study was to test the hypothesis that the outcome and complication rate of anterior cervical discectomy and fusion with or without instrumentation (ACDF(I)) is not inferior for supervised residents as compared to board-certified faculty neurosurgeons (BCFN). METHODS: This was a retrospective single-center study of all consecutive patients undergoing ACDF(I)-surgery between January 2011 and August 2014. All procedures were dichotomized into two groups according to the surgeon's level of experience: teaching cases (postgraduate year (PGY)-2 to PGY-6 neurosurgical residents) and non-teaching cases operated by BCFN. The primary study endpoint was patients' clinical outcome 4 weeks after surgery, categorized into a binary responder and non-responder variable. Secondary endpoints were complications, need for re-do surgery, and clinical outcome until the last follow-up. RESULTS: After exclusion of six cases because of incomplete data, a total of 287 ACDF(I) operations were enrolled into the study, of which 82 (29.2 %) were teaching cases and 199 (70.8 %) were non-teaching cases. Teaching cases required a longer operation time (131 min (95 % confidence interval (CI) 122-141 min) vs. 102 min (95-108 min; p < 0.0001) and were associated with a slightly higher estimated blood loss (84 ml (95 % CI 56-111 ml) vs. 57 ml (95 % CI 47-66 ml); p = 0.0017), while there was no difference in the rate of intraoperative complications (2.4 vs. 1.5 %; p = 0.631). Four weeks after surgery, 92.7 and 93 % of the patients had a positive response to surgery (p = 1.000), respectively. There was no difference in the postoperative complication rate (4.9 vs. 3.0 %; p = 0.307). Around 30 % of the study patients were followed up in outpatient clinics for more than once up until a mean period of 6.4 months (95 % CI 5.3-7.6 months). At the last follow-up, the clinical outcome was similar with a 90 % responder rate for both groups (p = 0.834). In total, five patients from the teaching group and eight patients from the non-teaching group required re-do surgery (p = 0.602). CONCLUSIONS: Short- and mid-term outcomes and complication rates following microscopic ACDF(I) were comparable for patients operated on by supervised neurosurgical residents or by senior surgeons. Our data thus indicate that a structured neurosurgical education of operative skills does not lead to worse outcomes or increase the complication rates after ACDF(I). Confirmation of the results by a prospective study is desired.

Shunts: Is Surgical Education Safe?

BACKGROUND: More data regarding complications in neurosurgery residents' cases are needed to assess patients' safety during hands-on surgical education. METHODS: A retrospective 2-center study was performed comparing consecutive patients undergoing shunt implantation by a supervised neurosurgery resident (teaching cases) versus a board-certified faculty neurosurgeon (nonteaching cases). The primary end point was surgical revision after shunting. Univariate and multivariate Cox proportional hazard models (Breslow method for ties) with time censored at 2 years were used to examine time-to-event data. Operation time, length of hospitalization, intracranial hemorrhage, and misplacement of the shunt catheter were other outcome measures to be compared between the groups. RESULTS: A total of 320 shunts (180 [56.3%] teaching and 140 [43.7%] nonteaching cases) with a mean follow-up of 563 ± 771 days (standard deviation) were analyzed. Revision rates for the entire cohort were 9.3% at 90 days, 13.3% at 6 months, 18.4% at 1 year, and 26.5% at 2 years. In univariate analysis, teaching cases were 96% as likely as nonteaching cases to be surgically revised (hazard ratio, 0.96; 95% confidence interval, 0.54-1.70; P = 0.877). In multivariate analysis adjusted for indication and shunt type, teaching cases were 94% as likely as nonteaching cases to undergo surgical revision (hazard ratio, 0.94; 95% confidence interval, 0.53-1.69; P = 0.847). There were no group differences in operation time, length of hospitalization, intracranial hemorrhage, and rates of shunt misplacement. CONCLUSIONS: The results of the current study in addition to the literature on neurosurgery resident training support the safety of supervised early surgical education for shunt surgery.

The timed up and go test for lumbar degenerative disc disease.

We report on the use and performance of an objective measure of functional impairment, the timed up and go (TUG) test, in clinical practice for patients with lumbar degenerative disc disease (DDD). We illustrate nine representative patients with lumbar DDD, who were selected from an ongoing prospective study, to report our clinical experience with the TUG test. In addition, a preliminary sample of 30 non-selected consecutive patients is presented. The following parameters were assessed preoperatively, and 3 days and 6 weeks postoperatively: back and leg pain using the visual analogue scale (VAS); functional impairment using the Oswestry disability index (ODI) and Roland-Morris disability index (RMDI); health-related quality of life using the EuroQol 5D (EQ5D) and Short-Form 12 (SF-12). The TUG test results improved by 2.6 and 5.4s after 3 days and 6 weeks compared to the baseline assessment. The mean VAS for back and leg pain decreased by 2.3 and 5.3, respectively, after 3 days, and by 2.7 and 4.6 after 6 weeks. The mean RMDI and ODI decreased by 3.4 and 23.3, respectively, after 3 days, and by 7.0 and 28.0 after 6 weeks. The mean EQ5D increased by 0.38 after 3 days and 0.358 after 6 weeks. The mean SF-12 mental component scale decreased by 0.2 after 3 days and increased by 5.6 after 6 weeks, whereas the mean SF-12 physical component scale increased by 6.4 after 3 days and by 9.8 after 6 weeks. The TUG test proved to be a useful, easy to use tool that could add a new, objective dimension to the armamentarium of clinical tests for the diagnosis and management of DDD. From our preliminary experience, we conclude that the TUG test accurately reflects a patient's objective functional impairment before and after surgery.