No abstract

The near-infrared photoluminescence intrinsic to semiconducting single-walled carbon nanotubes is ideal for biological imaging owing to the low autofluorescence and deep tissue penetration in the near-infrared region beyond 1 µm. However, biocompatible single-walled carbon nanotubes with high quantum yield have been elusive. Here, we show that sonicating single-walled carbon nanotubes with sodium cholate, followed by surfactant exchange to form phospholipidpolyethylene glycol coated nanotubes, produces in vivo imaging agents that are both bright and biocompatible. The exchange procedure is better than directly sonicating the tubes with the phospholipid-polyethylene glycol, because it results in less damage to the nanotubes and improves the quantum yield. We show whole-animal in vivo imaging using an InGaAs camera in the 1-1.7 µm spectral range by detecting the intrinsic near-infrared photoluminescence of the 'exchange' single-walled carbon nanotubes at a low dose (17 mg l −1 injected dose). The deep tissue penetration and low autofluorescence background allowed high-resolution intravital microscopy imaging of tumour vessels beneath thick skin.Single-walled carbon nanotubes (SWNTs) have shown potential for biological and medical 1 applications because of their intrinsic optical properties and ability to load both targeting ligands and chemotherapy drugs, in vitro 2,3 and in vivo 4-6. Their unique optical properties make SWNTs attractive candidates for biological imaging7 -11 and sensing12. In particular, the near-infrared (NIR) photoluminescence of semiconducting SWNTs13 has made them promising as NIR fluorescent contrast agents in biological systems8 -11. SWNT NIR fluorescent probes with emission mostly in the infrared-A (IR-A, 1-1.4 µm) region (Fig. 1) are ideal as biological probes because of the inherently low autofluorescence in the NIR range (0.8-1.7 µm) 14 and large Stokes shift between the excitation and emission bands, which allows excitation in the biological transparency window15 near 800 nm while further reducing the background effects of autofluorescence and scattering. Work carried out by Lim and colleagues16 has predicted that NIR fluorophores with emission in the 1,100-1,400 nm range have higher tissue penetration than those near 800 nm by considering the

We address the problem of acoustic source separation in a deep learning framework we call "deep clustering". Rather than directly estimating signals or masking functions, we train a deep network to produce spectrogram embeddings that are discriminative for partition labels given in training data. Previous deep network approaches provide great advantages in terms of learning power and speed, but previously it has been unclear how to use them to separate signals in a classindependent way. In contrast, spectral clustering approaches are flexible with respect to the classes and number of items to be segmented, but it has been unclear how to leverage the learning power and speed of deep networks. To obtain the best of both worlds, we use an objective function that to train embeddings that yield a low-rank approximation to an ideal pairwise affinity matrix, in a classindependent way. This avoids the high cost of spectral factorization and instead produces compact clusters that are amenable to simple clustering methods. The segmentations are therefore implicitly encoded in the embeddings, and can be "decoded" by clustering. Preliminary experiments show that the proposed method can separate speech: when trained on spectrogram features containing mixtures of two speakers, and tested on mixtures of a held-out set of speakers, it can infer masking functions that improve signal quality by around 6dB. We show that the model can generalize to three-speaker mixtures despite training only on twospeaker mixtures. The framework can be used without class labels, and therefore has the potential to be trained on a diverse set of sound types, and to generalize to novel sources. We hope that future work will lead to segmentation of arbitrary sounds, with extensions to microphone array methods as well as image segmentation and other domains.

No abstract